Last year, in March 2024, at a gathering in Malaysia on the 10th anniversary of the disappearance of MH370, family member V.P.R. Nathan presented several slides highlighting Ocean Infinity’s willingness to conduct another subsea search for the debris field of the aircraft. The search area proposed at that time extended along the 7th arc from about 33°S to 36°S, and out to about 45 NM (83 km) on either side of the arc. Ocean Infinity also announced plans to “fill-in” areas that were previously searched that had low quality or missing data due to equipment failures or challenging terrain. The proposed search area surrounded the “Last Estimated Position” (LEP) from UGIB 2020, which was designated the “IG Hotspot” in the slide. The proposed search area also incorporated our High Priority Search Area due south of the LEP that was missed by the previous searches by the ATSB and Ocean Infinity due to the steep sloping terrain in that area.
More recently, there were the following developments:
- On December 20, 2024, the Malaysian government agreed “in principle” to a new search effort with Ocean Infinity, indicating the start of formal negotiations.
- On March 19, 2025, Malaysia’s Transport Minister Anthony Loke announced that the Malaysian cabinet had agreed to the terms and conditions of the agreement with Ocean Infinity under a “no find, no fee” arrangement, with a payment of $70 million if successful.
- On March 26, 2025, a group representing the families of MH370 victims announced that a no-find, no-fee contract was signed by Malaysia and Ocean Infinity. Despite this announcement, there has been no official confirmation from either Malaysia or Ocean Infinity.
In February 2025, with no signed contract in hand, Ocean Infinity began the search that was proposed the year before. After completing two phases of the subsea search, Armada 7806 and its team of three AUVs is now on course to Singapore. Prior to departing for Singapore, the activities conducted during the two phases can be summarized as follows:
Phase 1: After arriving in the search area after departing Mauritius, this phase of the search began on February 23 and was completed on February 28, before departing to Fremantle to reprovision and resupply. The areas searched during this phase (dotted black box in the figure above) were primarily areas previously searched by the ATSB and by Ocean Infinity. Consistent with Ocean Infinity’s proposal from March 2024, the Phase I activities focused on infilling data where challenging terrain such as steep slopes previously resulted in missing or low quality data (data holidays). As such, most of the area in the Phase 1 box was not re-searched. The areas searched in this phase included our “High Priority Search Area”.
Phase 2: After departing Fremantle, Armada 7806 began searching again on March 11 and continued until March 28, before departing for Singapore. The areas searched during this phase (dashed black box in the figure above) were never searched before, starting further southwest and wider than the area proposed by Ocean Infinity in March 2024 (red boxes in the figure above). This would suggest that Ocean Infinity intends to enlarge the search area from what was previously disclosed.
The reasons for Ocean Infinity concluding this part of the search appear to be related to worsening seasonal weather and also related to prior contractual commitments for Armada 7806.
An interesting event occurred during the return to the search area before the start of Phase 2. The course of Armada 7806 was originally towards a part of the area that was searched in Phase 1. Then, on March 10, there was a change in course towards the southwest to begin searching areas not previously covered.
If there was no change in course on March 10, the purple dotted line in the figure above shows where in the Phase 1 search area Armada 7806 would have reached. In fact, our prior article discussed the possibility that Armada 7806 was returning to a debris field that was detected during Phase 1. This area is shown in more detail in the figure below.
Shown in the figure by the black oval is an area that was possibly searched in both Phase 1 and Phase 2, which might mean a promising area required further investigation. However, as we don’t know the exact path that the AUVs followed after launching, there may have been little to no overlap between the Phase 1 and Phase 2 search areas.
So why did Armada 7806’s course change on March 10? Here are two possibilities:
- What resembled a debris field was found during Phase 1 and the plan was to revisit that area during Phase II to collect more data. Due to ongoing contract negotiations, OI strategically opted to not disclose the location of the debris field, and the course changed to the southwest. Some additional data in the area of the debris field might have been collected at the end of the Phase 2. (See the potential overlap area in the figure above.) If the debris field was already found, it is unknown whether Ocean Infinity has met the requirements to earn the fee of $70 million, which may require the retrieval of one or more parts of MH370.
- No semblance of a debris field was found during Phase 1. Originally the plan was to begin the Phase 2 search where Armada 7806 originally headed and then progressively search to the southwest. For some reason, the plan was changed on March 10 to begin the search to the southwest and progressively search to the northeast, ending where originally planned to begin.
In any event, if the debris field was not found, we are hopeful that Ocean Infinity will return to the search area in November, later this year.
Acknowledgement: This article benefited from ongoing discussions with Mike Exner, Don Thompson, Bobby Ulich, and Andrew Banks.
Tags: AUV, debris field, Malaysia, MH370, Ocean Infinity, subsea
They didn’t find anything like you told you they wouldn’t.
I wonder if you ever think back to October 2023 when I tried to tell you the truth but you were too ignorant to accept it.
I’m going to haunt you like a curse for the rest of your life.
You allude to it somewhat in your article, but if Armada 78-06 found the debris field in this phase of the search, how likely do you think it would be that they would mask that fact by continuing to search northeastward, as if nothing had been found?
Thanks in advance ….
@VictorI:
Thanks for your previous comment mention that I’m not violating physics! From fresh research, I have additional confidence in my Java anomaly candidate site. That would leave your option 2 of nothing found in the data holidays, consistent with OI continuing an expanded search. We can hope for some confirmation that a contract has been signed. Meanwhile, it would be great if OI could slow down and gather some MBES bathymetry along the unexplored 7th Arc, if they have time. And, I’ll note that the tropical Java site can be searched at any time. No need to wait until November.
@Kenyon:
The public seismic data is not hard to find. I started off a decade ago acquiring data with MATLAB/OCTAVE scripts, with encouragement from Curtin. Nowadays, Python with the Obspy library is the way to go. Acquiring Iris/FDSN data, applying instrument corrections, and polarization analysis are intrinsic functions. If you’re not up for wrangling python, then I suggest consulting with seismologists.
I’ve recently acquired public 2014 data from two very broadband ocean bottom hydrophones operated by the Australian Antarctic Survey, placed 1,500 km apart. This has opened up a whole new range of experiments and astonishing results.
So, we are not stuck with just public seismic data for confirmation. There are hydrophones, too.
@Erik Geer: If the debris field was found before the contract was signed, and announcing that it was found would in any way jeopardize the claiming the fee, there would definitely be an effort to keep that information secret.
@Ashton Forbes: Get a life. Or maybe transport yourself through a wormhole.
@Victor Iannello
You do have to feel a little sorry for Ashton, Victor. I mean, typically children start to learn that what happens in cartoons isn’t real around age five, whereas this fellow seems to have missed that realisation, and has based his entire thesis on an animation. Probably a good thing he hasn’t seen Disney’s Planes.
@Ashton Forbes
All of us have a right to our opinions. But in expressing them, I would argue that we should do so civilly, and without threatening those that we might disagree with. Please give this notion some consideration.
Thank you ….
I know next to nothing about maritime operations but I don’t think there is much significance to the change of course on march 10 other than someone making a decision based on weather forecasts for the search area. They likely knew in advance that they wouldn’t be able to cover more than half the length of the proposed area before the end of the season. By initially heading roughly for the center, they could decide spontaneously which half to go first: Center to south, center to north or – go all the way to the southern end if the weather permits it. The explanation probably lies in march 10th weather data.
@Ed,
My question wasn’t well formed. I have found multiple resources that have seismic data for thousands of events, I just can’t seem to find data for the seismometer stations listed on your website for the JA seafloor impact timeframe on March 8th. My assumption some time ago was that perhaps the magnitude wasn’t high enough for the nearby stations to report or I perhaps I wasn’t looking at the right resources.
I use python quite a bit, today found and tried ‘PyWEED’ launching it through Anaconda. Ran fine, but it reported the roughly the same info on Sage and others. Does the Sage link you provided above have the stations’ data for the date and time presented on your website? Maybe I’m not seeing the obvious?
I’m unaware of Obspy, I’ll give it a go.. Any help finding the JA seismic data or if your could publish the basic data on your website that would be most helpful. Not looking to do any sophisticated data analysis that requires a seismologist, just looking for basic data that the stations reported.
@SG: We were tracking the weather and it was fine.
@370Location
I agree with VI that your position does not violate the laws of physics. I already stated that long time ago. However, it is not the same as saying that it has high probability.
In order to reach a position so close to Indonesia and crash at so late time demands an active pilot. Not just any pilot, but an excellent one. More importantly, keeping the airplane flying for so long time means (according to the laws of physics) that there is practically no free energy left. That means absolutely no fuel, practically no potential energy, and only an absolute minimum of kinetic energy to prevent stalling.
This means that such a late crash liberates very little energy. However, the energy liberated at your location is large. It is similar to a medium size nuke (not a Hiroshima bomb).
My guess is that you most likely observed a coal chondrite meteorite falling in the sea, or perhaps a very rare type of superhot lightning strike. Most of these rare strikes happen in the intertropical convergence zone, and your position is near that zone.
Is there any chance that OI will simply refuel, change the crew, and continue the search for one more phase before November ?
I hope we can receive, at the very least, some communication from OI confirming whether the contract has been signed.
@Rachid: It seems unlikely that OI would refuel and resupply in Singapore when Fremantle is so much closer. All indications are that Armada 7806 will not return to the search area after departing Singapore.
@Kenyon:
Tom, I see why you were stymied. The PyWEED page says it is an event-based downloader, built on ObsPy. So, it cannot help with MH370. The reason I called it the Java Anomaly is because the event is among the loudest of the day even on distant hydrophones, yet it was too weak to be cataloged seismically. (Also, because even strong M5.4 quakes don’t propagate well into the SOFAR channel.) You will not find the event listed in any seismic catalog, and therefore PyWEED cannot fetch the recordings. It may have been excluded from automated cataloging because the energy content is mostly higher frequencies, which are often ignored as “anthropogenic noise” rather than quakes which typically generate very low frequency waves that travel teleseismic distances, reaching more stations for analysis.
You would need to download the data from at least those regional sites, visually “pick” the P and S wave arrivals, then use TauP model timing estimates to triangulate the epicenter. I suggest looking at rolling kurtosis for accurately picking first arrivals. I found no epicenter code at the time, so I wrote my own optimization routine. It surprisingly converged on a shallow source within 1 km of the 7th Arc. I have also gathered 48 hours of seismic records from over 4,500 seismometers globally, in case other seismic phases might reveal the surface impact. Only about 45 of the ones selected on that Sage map show a clear indication of the event. The nearest IA network geophones are restricted. GE.CISI near the Java coast has the clearest record of the event, which I sped up 60x and saved as a sound file on my website. You can use those P and S wave arrivals to estimate the distance to the event, if you just want confirmation.
I’ve sought help here at guesstimating the probability that the Java event was geologic rather then a plane impact, which is why it was originally dismissed in the first acoustic analysis as a “low level quake in the Java Trench”. The lack of a cataloged quake does shift the probability.
@Viking:
To reach the Java site while matching the BTO pings, MH370 would have been flying at a low and slow holding speed consistent with maximum endurance, not distance. I estimated it at oxygen altitude for my waypoint path. Even at fuel exhaustion, there would be quite a bit of energy in a crash, whether high speed or ditching.
Still, you seem to be confusing the timing. The Java detection is consistent with a seabed impact 55 minutes after the 7th Arc BTO, as the plane would have been sinking. Tom Kenyon has attempted to estimate the energy of a forward section of fuselage hitting the seabed, but excluded the entrained water. Surface impacts, unfortunately, do not propagate into the SOFAR channel. Here’s a report on my attempt at using lighting strikes for calibrating the hydrophone locations, thanks to a database shared with me by Viasala:
https://370location.org/2017/12/ocean-lightning-strikes-compared-with-acoustic-event-detections/
Over 10,000 of the strikes were over the Indian Ocean. Some were indeed mega-strikes, the strongest over water was -710 KiloAmperes. You can see from my map on that page that none of the storms were anywhere near the 7th Arc. Almost none of the lightning strikes were detectable in the SOFAR channel.
In his first paper on MH370, Usama Kadri thought he was detecting a meteor strike. It instead matches in direction and pattern with ice tremors. A meteor strike would also be a surface event, so not likely detectable unless it was in shallow coastal water. There were at least three surface airgun surveys (blasts every 8-12 sec) happening in the NE SIO that cluttered the soundscape. The noise is strong enough that LANL didn’t attempt to use the CTBTO Diego Garcia hydrophone array in their analysis. I instead used beamforming techniques to isolate the noise sources.
The probability of a detectable meteor strike right on the 7th Arc 55 minutes after that last ping must be infinitesimal.
Consider that Vincent Lyne studied the acoustics, but dismissed vague “talk on the internet” about a “Java Anomaly” as the result of dynamite fishing. That would also be a surface event, and the two day record contains no other “blasts” along the 7th Arc.
TLDR; I have a lot of opinions about MH370 acoustics.
@Cessi: You claim you are “informed”. I provided you with two references (my previous article and Steve Kent’s video) which give some technical explanations as to why the historical WSPR data cannot be used to track MH370 and asked you report back with any errors. You also implied on another website that I am deleting or not approving your comments, which is false.
We’d be interested in hearing your technical arguments.
@Victor
Your paper investigating WSPR only shows that you (!) are unable to track aircraft using WSPR. It is not evidence that it is impossible altogether. You claim that the reflected signals would be too weak. However, concluding that a system specifically designed to handle weak signals will fail because the signals are too weak is quite a wild assumption.
@Adi
Not every “paper” meets the requirements to be considered a proper peer review. In science, a peer review requires an objective reviewer assessing another scientist’s work. To avoid subjective bias due to personal differences, a peer review should ideally be conducted in a blinded manner. That was definitely not the case here. Therefore, I strongly question the necessary objectivity of this so-called qualitative peer review. I also suspect a confirmation bias, as the goal from the outset seemed to be to disprove RG. This is not a scientific peer review—it’s a joke.
This bias seems to be commonplace here, as I am being called “uninformed” even though you know nothing about my scientific background. Such behavior and personal attacks are not a reliable basis for discussing science, as they prevent objective debate.
@Cessi:
I have no idea about your scientific background, nor is that important. You are being called uninformed because of the technical claims you make. If you want to have an objective technical debate, you need to make arguments that adhere to the laws of physics. That is not a personal attack. Nor am I “biased”, as you claim, any more than the laws of propagation, diffraction, and theoretical noise floor are biased.
Yes, WSPR is designed to detect weak signals. But the scattered signals off of aircraft at long distances are MANY orders of magnitudes less than what is PHYSICALLY possible to detect. It’s the attenuation caused by the scattering process that makes the signals undetectable. For some reason, this detail is ignored.
Even if the scattered signal were detectable (which it isn’t at the powers and distances of interest), it would be very weak compared to the direct (unscattered) signal, and would have no influence on the values recorded in the WSPR database. Only through numerical processing that makes use of Doppler discrimination could the scattered signal be separated from the direct signal, and this spectral information is not available in the database.
Some of us have set up experiments to scientifically measure the strength of scattered HF signals off of aircraft. As predicted, the scattered signal is orders of magnitude less than the direct signal, and could only be detected when the direct signal was extremely strong (transmitter power with hundreds of kilowatts) and the aircraft was fairly close (tens of kilometers) from either the transmitter or receiver. And then, the signal was detected only by using Doppler discrimination to separate the scattered signal from the direct signal.
I have encouraged Simon Maskell and Richard Godfrey to run their own experiments to compare the direct and scattered signals from aircraft. This could provide unambiguous evidence that WSPR can be used to track aircraft. I’ve even offered to help them set up the experiment. It would not cost a lot of money as SDR receivers and omnidirectional antennas are available at low prices. Years have gone by, and they have not run those basic experiments to demonstrate the physical principles. Meanwhile, Nils Schiffhauer and I have run the experiments, and the results are exactly as predicted–the scattered signals are only detectable at high powers and short distances, and only with spectral processing to separate the scattered and direct signals.
@Cessi,
It’s not “science,” if unconflicted peers/experts cannot verify or repeat the propositions. If you are unable to grasp this fairly fundamental concept, there is no further debate necessary or warranted. Victor’s “uninformed” moniker is clearly aimed at your position on this topic, and not at YOU the individual or your background. Everyone can see that.
Regardless, my best wishes to you. Clearly your passion on this topic is driven by your desire to find MH370, and resolve the mystery. I assure you that is perfectly in sync with others on here as well. Let’s wish the parties on the ground Godspeed for their endeavors later this year and hope for the best.
Best,
Adi
@Cessi
I am more practical. I do not agree with WSPR in part due I think it is missing flight path up the Straits and beyond. I feel like I can see some of the human logic going into the WSPR flight path. Having said that, my opinion WSPR path is closer to where MH370 actually crossed Arc7 than most of the other estimates. I think MH370 probably flew some distance from Arc7, so if you rent a ship, pls see me.
WSPR arguments are great. One person repeatedly shoots at the broad side of a barn and paints targets on the place where they hit. Then a thousand people come along and say “how can you say he is not a great marksman, look at all of these bullseyes!”. And you can’t say anything to convince them otherwise because understanding a set of coordinates on a map and a “tripwire” analogy is very easy, but explaining the flaws in the process used to get those results is pretty hard.
It’s a trap that’s very easy to fall into when analyzing large datasets. You believe so strongly there is some truth in the data, it’s just buried in the noise. Like a sculptor you diligently carve away and remove all of the implausible “outliers”, tweak parameters, try new methods, until at last you finally get the result you expected. It’s beautiful, it’s clear, it’s noise free, it’s exactly what you needed, because you in fact accidentally created it yourself.
I’d encourage anyone looking into e.g. hydrophone data to be very careful of this effect too. Ask yourself, how many parameters did you have to tweak, how much “irrelevant” data did you need to discard to get your result.
I don’t blame anyone for believing in this stuff. Wishful thinking is a powerful drug. It’s very hard to get out of.
@TBill: You base your agreement with WSPR tracking on how well its “prediction” aligns with your guess. What we are discussing is the physical validity of the WSPR tracking methodology, which is not dependent on how we might guess the plane was flown.
In fact, I would even say that the plane COULD be found in the WSPR hotspot. There’s nothing that precludes it. However, WSPR tracking cannot be used to discriminate that guess from any other guess, as it provides no additional information.
Transport Minister Anthony Loke confirmed that the government has signed an agreement with Ocean Infinity last week to resume the search for Malaysia Airlines flight MH370. When asked by the media on Wednesday (April 2) whether this renewed search would result in the discovery of the plane’s wreckage, Loke stated that the outcome would depend on the progress and success of the search efforts.
https://www.msn.com/en-my/news/videos/mh370-loke-confirms-ocean-infinity-will-resume-search-efforts/vi-AA1CadFX
> Victor Iannello: “I would even say that the plane COULD be found in the WSPR hotspot”
I would call it the Godfrey hotspot instead, so nobody can misread this as WSPR being capable of discerning any hotspot.
“whether this renewed search would result in the discovery of the plane’s wreckage […] would depend on the […] success of the search efforts”
master of tautology
@Charm
It’s interesting to see how you associate wishful thinking with believing in WSPR tracking but won’t acknowledge that the same applies the other way around. This is, after all, what’s known as confirmation bias—whether it’s someone convinced that WSPR works or someone determined to prove it doesn’t, only to unsurprisingly reach that conclusion when testing it. Wishful thinking is indeed a powerful force, influencing our minds more than we often care to admit.
@Cessi
All applied science has to conform to the laws of physics, either as already understood or very occasionally by revealing a prior error in the application of those laws. There have been theoretical studies by some that that imply through prior understanding of the physics of HF propagation that WSPR is unlikely (by many orders) to be able to track aircraft on the HF anomalies they produce. This comes as no surprise to people who have knowledge of over the horizon radar which has been developed independently in a number of countries and deployed in some for more than five decades.
So far proponents of WSPR don’t seem to be able to provide examples of tracked aircraft where the aircraft track was not already known. There’s plenty of examples of “pattern matching” WSPR anomalies to known tracks. It is unsurprising that out of zillions of anomalies in HF propagation that some can be hand selected to “confirm” a known track. That is not close to even establishing proof of concept IMHO – but rather wishful thinking or, the more contemporary application of hopeium.
I am not saying that WSPN will never reveal anything useful, but it’s unclear to me (with with some experience in processing low level signals) how the current methods processing of WSPR data can extract positional data. The WSPR tracking hypothesis not been presented in any meaningful way in WSPR papers published to date as far as I can tell. On the other hand, WSPR tracking is benefiting from excellent marketing of the concept to non-scientific types.
@Cessi: If you believe my conclusions on the invalidity of WSPR tracking is based on confirmation bias, you surely don’t understand the physics of why scattered signals are greatly attenuated. That’s why I say it is futile to try to persuade the uninformed.
I have heard no valid excuse for why the WSPR proponents have not demonstrated the underlying physics with simple, unambiguous experiments as others have done. If low power HF scatter off of aircraft can be detected at a distance of thousands of kilometers, that scatter should be measurable. If my analyses and experiments are not believed, they should perform their own. They haven’t yet.
@Cessi
In a previous comment you scorned criticism of Richard Godfrey’s work, claiming it was biased and labelling it a “joke”.
Can you please point readers towards a peer review of Richard Godfrey’s work that meets the standard you outlined?
@Victor, Perhaps some of the difficulty with the [mis]understanding of WSPR that people struggle with is this …
It seems to me that we are not talking about detecting scatter from an aircraft, rather the issue is the perturbations to the “normally” received signal, in SNR or some other measure, as a result of an aircraft transiting through the signal path.
I suspect that this has the same problems, in that the “normal” signal is subject to many other influences that can affect SNR, and not just an aircraft.
I do understand some of the complications, having played with low power VHF in my long distance past. I built all my 2M gear, both transmit and receive, and had fun with long distance contacts with only 100mW on 144Mhz.
@Andrew,
I have tried that line of reasoning, and according to @Cessi, peer reviews aren’t required to meet the standards of believability in some instances – such as this, conveniently 🙂
Good luck!
Adi
@Brian Anderson: Yes, what is calculated by WSPR trackers are perturbations of the “direct” (unscattered) signal. But if scatter were to influence (perturb) the recorded values in any way, the magnitude of the scattered signal has to be similar to the direct signal, whether the signals constructively or destructively combine. In fact, the scattered signal is MANY orders of magnitude smaller than the direct signal and will have no effect on the measurement. Meanwhile, as you say, other factors such as dynamic effects in the ionosphere and multipath interference will create strong fading that will strongly influence the measured values.
There is no reason to doubt established science, but that’s what the WSPR proponents are implicitly claiming.
MH370 must be found and Ocean Infinity should be getting strong support from ICAO. It appears however that the Malaysian and Australian governments are reluctant to push the search forward. The MH370 families must be disappointed.
I lived in Malaysia for two years flying fighters from a base near Penang. Also have about 5000 hours PIC B777.
I don’t care who is ultimately successful regarding the science as long as MH370 is found.
Malaysia’s Ministry of Transport just announced that the Ocean Infinity search effort has been suspended, at least for now. It’s supposed to resume at the end of the year. I don’t know if it’s due to contractual issues or OI needing to do a vessel resupply.
A controlled final glide would put the wreckage far beyond the existing search boxes. There could be up to 10x more area to search which would make the likelihood of finding the plane exceedingly remote. I don’t think any government would be willing to spend a billion dollars with no guarantee of success.
I think almost everyone wants an end to the MH370 mystery. Science is agnostic when it comes to picking sides but, with respect, time effort and money spent on hocus-pocus is not a means to hasten to that end (with the possible exception that discussion (AKA brainstorming) around the technical feasibility of pathways to investigation may sharpen thought processes).
As far as I can tell all of the hypotheses to the end point for MH370 depend on one or more subjective assumptions which are therefore subject to innate biases. There aren’t always absolute rights and wrongs, but there are certainly differing degrees of likelihood between those hypotheses.
That there is a significant number of MH370 end point proponents who appear to be blind to their own biases in the development of their hypotheses does surprise me somewhat. The prosecution of science isn’t always as stringent as it should be. Which is not a flaw in science, it is a flaw in the human application of science.
@Byron Bailey: Of course the top priority is that MH370 is found.
@John Laurens: A controlled final glide would not put the wreckage very far away from the existing search box if we take into consideration the BTO and BFO data from Inmarsat. The seventh arc constraints from satellite data provide strong boundaries on possible endpoints regardless of piloting technique(controlled ditch OR no inputs by the pilot in command). Ocean Infinity just covered the Blelly/Marchand area, which is an area based on final glide theory, integrating both the satellite ping data and aircraft performance limitations.
The wreckage might be further south than the Blelly/Marchand area. At that point you’re right, nobody will likely go search there because it requires even more resources without a guaranteed result, especially after multiple high-cost search operations have already been conducted.
Hopefully in the upcoming days, OI will find something interesting after analyzing the raw data from the last phase of the search.
@Rachid,
You wrote: “Hopefully in the upcoming days”,
Suggest: Hope does not factor in the search.
If you look at Phase 2 of the recent search it seems like steady persistent progressive area coverage.
The only available evidence is the movement of the surface vessel.
But this may be considered indicative of progressive coverage by the three underwater craft.
If you also look at the implied total area covered during Phase 2, then there may be sufficient time available (presuming searching restarts in late October or November this calendar year) to cover a much larger further area before weather “closes in” mid 2026.
It is not hope which might produce a result, but such persistent searching until a result is produced.
@Geroge G
I agree that results will come from science and persistent and methodical searching. I’m not saying that we should rely on hope alone. My use of “hopefully” was merely expressing anticipation for what the data analysis might reveal, not suggesting that hope is a factor in the search methodology itself.
@Rachid: What’s the absolute furthest south the plane could have reached in a controlled straight glide without any turns or doubling back after the seventh arc transmission?
Would that location allow for some debris to have reached the coast of Western Australia?
@John Matheson
Thank you for that announcement, that was missing in action. We have to wait for contract details (if that can be disclosed), but the correct answer was: yes they will do everything in their power to find aircraft.
@Adi
The sudden silence from that quarter is deafening. It’s quite incredible that some people consider criticism to be a “joke” in the absence of peer review, while no such standard is applied to the original work.
I’ve now determined the reason behind the apparent data discrepancies in the latest WSPR study:
https://www.dropbox.com/scl/fi/o5ys0coxzixr9l3aem73l/Data-discrepancies-in-1_1_2025-WSPR-study.paper?rlkey=ntg21v6dwdxyfwxx9cp52t4ho&st=kmpopri5&dl=0
Cases 1 and 2 erroneously use +27h/-3h time frame instead of the supposed +/- 3h (someone probably mistyped the wrong date adding 24h). In addition, case 12 uses +3h/-6h. I’ve confirmed with multiple links that when searched with this “correction”, the group of links corresponds to the flawed SNR% presented in the study.
More interesting than the error itself is the lack of integrity displayed in the response, and what this tells about the reliability of other parts of these WSPR studies.
I also finally got “banned” from his site after sending this dropbox link over email.
@eukaryote: Thank you for tracking this down. Unfortunately, few people bother to dig into the actual data as you have. I assure you that if WSPR data seems to show any ability to discriminate the position of a plane thousands of kilometers away, it is due to bias introduced somewhere.
@John Matheson
“That there is a significant number of MH370 end point proponents who appear to be blind to their own biases”…true but it is more complicated than that. Blaming the pilot has sensitivities for the public, Malaysia, many pilots, and some aviation industry supporters. Also there is a assumption that says if there were active pilot maneuvers, then we cannot find it. Since I think there probably were maneuvers, I am a pessimist re: actually finding. But the flip side is many key stake holders accept all that above, in other words, many accept the ground rules above and that we may not find MH370, but are hopeful that it might be found.
eukaryote,
In answer to your question from Mar 27, the number 1108 at the 20:04 mark is in error (introduced when transposing from my notes to the slide) – it should be 1103. The other numbers, however, should be correct, and the conclusions are unchanged.
@sk999
What makes the ADS-B table confusing to me is the inclusion of multiple 2 min time slots with some of the cases instead of the 0/+6 min laid out in the study description. I still don’t know why that is, but fortunately this didn’t affect cases 1, 2 and 12 so I could mostly ignore it.
There also aren’t clear rules about how the 48 flights and timings were chosen, and I’m no longer sure I can trust that this was the original random selection.
@TBill
Thank you for your considered comments.
When I wrote that I had in mind those ‘experts’ who, upon for whatever reason having satisfied themselves with the veracity of their hypothesis, proceed to search for pattern fitting to support it, whilst disregarding data that doesn’t. Without pointing fingers I’m sure people can think of quite a few examples.
Unfortunately the media has limited penchant to apply critical thinking to different theories – they exist to sell content, not scientific integrity per se.
There is a chance, however small, that the plane will be found in nearest proximity to a spuriously predicted endpoint, allowing the proponent to bask in undeserved glory. Unfortunately that is a consequence of inherent uncertainties in the assumptions that all trajectory to endpoint models have.
@eukaryote
Your uncertainty with regards to the randomness of either or both of the flights selected and the times applied is particularly well founded, for when it comes to cherry picking, these fellows make the good people of Yamagata Prefecture, Japan look like clumsy amateurs.
You might recall their December 2023 (ahem) “technical paper”, How does WSPR detect Aircraft over short Distances?. That paper looked at purported “detections” of aircraft as they supposedly affected WSPR transmissions between transmitter W8AC in Chardon, Ohio and receiver N8GA in Miamisburg, Ohio, some 252 kilometres apart.
The “paper” stated,
“During the 3rd November 2023 a total of nine flights were analysed with different types of commercial aircraft including Boeing, Airbus, Embraer and Canadair flying at altitudes between 7,350 feet and 37,000 feet in the vicinity of the transmitter W8AC. Eight of the nine flights were detected with a total of 13 SNR anomalies. In two cases flights were additionally detected by frequency drift anomalies.” (p.2)
On the face of it, one might think that a raw detection rate of nearly 90 percent is most assuredly something to write about. However, it quickly becomes apparent that the nine flights selected were likely specifically curated to produce the extraordinarily high detection rate.
The nine flights addressed covered a 10.23 hour time period from 09:20 UTC – 19:34 UTC (05:20 – 15:34 local). It turned out that during that time period there were some 140-odd aircraft, ranging from Embraer Phenom small passenger jets all the way up to B777s, B767s and even USAF C-17s, that impinged on either the long or short path while WSPR spots were being recorded between those two stations. When you examine the spot data for all of those impingements, you find an actual detection rate of less than half their 8 out of 9 claim.
Basically, you cannot take anything that is produced by that trio at face value.
How does everyone really feel about the radar targets published from over the Malacca Strait? Before the FMT?
Confidence?
@370Location
In case your signal came predominantly from a seabed impact, most of its energy would have gone into seabed R-waves. I found strong evidence that was not the case. Instead, most of its energy went into waves propagating above the jump layer. That would never occur in case of a dominant seabed impact.
Instead it is consistant with a lightning strike or a coal chondrite. An iron meteroite would put more energy in the SOFAR channel and/or seabed R-waves, depending on its impact angle.
I personally beleive most in a coal chondrite impact. They are actually quite common.
@William Shea
Interesting question. I presume you mean those collected by Singapore, but later debunked or perhaps just ignored since they did not seem to fit in? Do you have technical details of them?
@William Shea: I think you are referring to the slide shown to the NOK at the Lido Hotel in Beijing in March 2014.
Either:
1. The radar targets are valid but Malaysia decided to later not release them. For instance, the radar data released to the ATSB contained no targets between 18:01z and 18:22z. Why not?
2. The radar targets are not valid. But then, why did Malaysia show them to the NOK and claim they were MH370?
For either case, Malaysia is being deceptive.
I tend to believe that (1) is more likely because the targets align with the civilian radar data and the Inmarsat BTO/BFO data.
Thank you all for response. I do not have anything to contribute. I wish I did.
Was hoping by now, there would be some newly discovered radar data or possibly new interpretation of existing data.
Thanks.
I wonder what Armada 7806 stopped to look for at 6.4S, 103.4E? It’s not like it was on their way to Sunda Strait. Obvs nothing to do with MH370 as it is so far from Arc7 / Ed’s Java anomaly candidate.
@Mick Gilbert,
“You might recall their December 2023 (ahem) ‘technical paper’, How does WSPR detect Aircraft over short Distances?”
Yep, but I never downloaded it. So important, that I now find that when I go looking for it, it has vanished – into thin air. Didn’t even leave the slightest smidgen of a recordable doppler shift on it’s progression into the ether.
Hello all,
First time poster but have been following the MH370 case for years. Godfrey has stated that the US military within 24 hours of MH370 going missing had analysed the BFO data and sent out aircraft to search around the 7th arc. This is the aircraft Captain Smith supposedly saw the day after MH370 went missing. Is it all possible they had this ability to figure out potential crash sites so soon while the rest of the world was looking in the south China Sea?
@Barry Carlson,
If you wish, using Richard Godfrey’s site you will find: “WSPR as Radar
by Richard Godfrey | Dec 18, 2023 | ”
The accompanying (there linked) “technical paper” is that to which Mick Gilbert referred.
(So you don’t miss out)
@George G,
Thanks for not letting me ‘miss out’.
The paper is also linked to from the Airline Ratings website, but the PDF file has been deleted from DropBox. I wonder why.
@George G,
I’ve read the paper, but still missed out… on the physics behind detecting a specific aeroplane, or indeed how a particular aeroplane could be identified and tracked without it already being known where it is.
It’s rather charitable to call the document a ‘paper,’ at least in the sense of scientific publications, because it’s lacking key details, like the mechanism of how the detection is supposed to work (within the laws of physics), how the theory can be objectively tested, and how the 20,000 or so WSPR sites and 20,000+ aeroplane movements per day are filtered down to track a single set of coherent anomalies from potentially many 100s of millions of anomalies generated each day. And that’s before even factoring the anomalousness of HF transmission and ionospheric behaviour, which would need to be quantified and accounted to isolated aeroplane induced anomalies, plus other sources such as marine traffic.
Chris asked: Is it all possible they had this ability to figure out potential crash sites so soon while the rest of the world was looking in the south China Sea?
It’s impossible to prove what somebody did or did not know, but it’s unlikely that the US was able to interpret the BTO/BFO data and operate on that knowledge all within around a day of the disappearance.
@Chris: Besides the Malaysian military radar data, there is also the possibility that a Singaporean airborne surveillance system, the G550-ELG550-EL/W-2085, detected a radar target or SIGINT from MH370 and forwarded that information to the US.
https://mh370.radiantphysics.com/2017/02/11/singapore-radar-and-mh370/
Captain Smith’s testimony raised an interesting point. Eventually, he confirmed that he flew on March 8 actually.
But focusing on March 7, 2014, UTC in the southern Indian Ocean, after flying around Sumatra, flight MH370 likely crossed paths with a very small number of other flights. Data retrieved from the FlightAware web pages shows that it likely crossed paths with four other flights. Using the piloted flight path reconstructed by Captain Blelly and J-L. Marchand as a reference trajectory, the shortest estimated crossing distance is only 12 nautical miles for two of these flights. Flight MH370 could not have avoided seeing them. These flights did not see it because of the dark, moonless night, because its external lights were probably off and because it was likely at a lower flight level. The flight formation of these two aircraft flying between flight levels FL320 and FL380, one directly above the other, suggests that flight MH370 crossed them from below, which corroborates the flight level FL300 calculated in the reference trajectory.
The report with its video summary and the data are available at https://www.mh370-caption.net/index.php/possible-mh370-encounters-in-the-southern-indian-ocean/
The data about the 91 identified flights is in csv format for ease of use.
@VictorI. Captain Smith in his statement refers to HIS sighting, ie in the singular. It remains a puzzle to me as to why he makes no reference to confirmation of that or otherwise by his Captain of that flight. Just no mention.
Yet so far as I am aware Captain Smith has not been asked about this or an explanation sought from his Captain, such as being absent from the flight deck at the time.
Surely though he would have been alerted to this, particularly since, as Captain Smith notes, “the aircraft came through our block”.
Besides, this having been reported to the airline, surely he would have a view?
@Chris,
According to an early report by Martha Raddatz on the US news channel ABCNews , the US military may have detected MH370 in the Indian Ocean.
It could have been a signal detected through an Over-the-Horizon (OTH) radar in Diego Garcia, or the SOSUS network to detect submarines, or the Australians lied about JORN being unavailable.
“We have an indication the plane went down in the Indian Ocean,” the senior Pentagon official said.
The official initially said there were indications that the plane flew four or five hours after disappearing from radar and that they believe it went into the water. Officials later said the plane likely did not fly four or five hours, but did not specify how long it may have been airborne.
https://abcnews.go.com/International/us-officials-malaysia-airline-crashed-indian-ocean/story?id=22894802
@Chris,
Also, there is a credible eyewitness report of a US military person attached to the TUDM who was informed that Malaysian F-18’s with external fuel tanks were sent to search the Malacca strait the same night of MH370’s disappearance.
https://www.youtube.com/watch?v=xlJBS9yRVpU
@CanisMagnusRufus
That article says nothing about the US detecting MH370, rather it talks about “an indication”. The “indication” at that time was almost certainly the evolving interpretation of the Inmarsat data.
@CanisMagnusRufus
It was reported by Reuters that a senior Malaysian military official admitted the military had tracked an unidentified aircraft heading west across peninsular Malaysia the night MH370 disappeared. Military officials also claimed that fighters were not scrambled to intercept the aircraft because it was not viewed as “hostile”.
Reuters Exclusive: Malaysia starts investigating confused initial response to missing jet
Nevertheless, it’s quite possible the military did send fighters to search for the aircraft at first light, even though the official search didn’t move to the Malacca Strait and Andaman Sea for several days. That could fit with “Gonky’s” story in the video, given he dropped his parents-in-law at Penang airport in the morning and then went to work, where he spoke to the fighter pilots after they had returned. If that search did happen, I don’t think it’s ever been admitted to the public by Malaysian officials.
On the subject of the US military detecting MH370 in the Indian Ocean, is there any evidence to support such claims, or is it speculation? The ABC News report you cited is dated March 13, two days after Inmarsat had provided investigators with analysis that showed the aircraft may have headed south into the Indian Ocean. Given that US officials were part of the investigation team, the “indications” mentioned in the report might well be based on Inmarsat’s analysis, not work done by the US military.
@Jean Luc
I greatly appreciate some of your videos of air traffic. Those are nice contributions. According to my studies, EK425 out of Perth hit Arc7 right about cash time at about 20-22s, so if MH370 had unexpectedly gone that way (eg; Ed Baker’s theory) then it could have been close.
@All on Radar etc.
I recently came up empty-handed on a MH370 Freedom of Info Act (FOIA) request of FAA. I was told all documents are beyond the retention date (discarded).
Clearly the joint investigation team (including FAA) likely had more radar data including Thailand reportedly saw MH370 in Andaman Sea (I recently posted media reference on Twitter aka X). Per Victor, Singapore either saw something or did not see something, which we’d like to know the answer in either case.
I do feel there was an effort to give researchers enough data to try to find the aircraft, but some data is sensitive, for example sim data we did not get a leak for 2 years and it was incomplete.
In other news, I’d be shocked if USA was searching Arc7 on the first day, before Inmarsat even had a chance to study their data. There is not the slightest indication of that. But I do think this break in the action is a good time to do FOIA requests…there is much data hidden still.
@Andrew, Mick Gilbert… thanks for the response. Yes it was speculation, but not mine. It was something written on some military blogs.
@TBill … I’ve seen some videos of the French passengers going through security on some French documentaries about MH370.
Can someone ask the Malaysians to release the full video of the passengers going through security? It would be really interesting to take a look at the 2 Ukrainian passport holders to answer the following questions:
– do they actually look like the people on the photos that JW released?
– do they actually have the physique of ‘marines’ as FdC wrote on her book, or is it just the imagination of a middle-aged single white woman living on a boat in HK?
– were they actually wearing identical black clothing, and carrying identical baggage, or is it just an artifact of the black and white security camera?
– what can be ascertained about their demeanors?
I approached FdC about this but she was non-committal. Looking closely at her theories, she seems to be playing upto the anti-US audience and making $$ while at it.
@All,
Armada 78-06 berthed adjacent to 53 Shipyard Road, off Pioneer Rd., at 1045 SGT
@BarryC
That is the Mooreast Offshore Base in Singapore. OI has been using the facility for some time.
@Victor
Do you have some information about which time-window(s) the Singaporean G550 airplaine made observations near the NW point? Without such information it is hard to use the NW position data to distinguish between different models.
Perhaps we may be able to infer some indirect information from the ‘curved paths’ proposed initially (if they relied on G550 data), but it will be highly uncertain information.
@Viking: The statement from the ATSB says that at 19:11z, there was NO detection from the Singaporean radar source, and that set the position of the NW point. In the article, knowing the NW point is 8° 35.719’N, 92° 35.145’E, I tried to back out what that might imply about possible locations of the G550 and further implications on possible MH370 paths, but that’s not easy to do with the limited information. In the end, the ATSB chose to ignore any information related to the NW point.
https://mh370.radiantphysics.com/2017/02/11/singapore-radar-and-mh370/
@Viking,
As noted above, I was provided a database with 48 hrs of lightning strikes in the IO that day, down to weak 10 kiloAmp strength. The was no storm near the Java Anomaly site, and even the largest 500 kA+ strikes are surface events that do not propagate into the SOFAR channel. How then could one of the loudest SOFAR hydroacoustic events within 15 hrs be from a lightning strike?
Meteorites of the size required are bolides that leave huge infrasound sonic booms as they decelerate and fragment on entry at altitudes up around 20 km. After slowing, the pieces fall at terminal velocity. The CTBTO analyzed infrasound for reports of such an event and found nothing. It would also be a surface event, unless you’re talking about the rubble drifting down to the seabed. If you are convinced the sound was made by a meteorite, then perhaps it would be big enough to be cataloged by IR detecting satellites.
You clearly have lingering doubts about the source of the Java Anomaly on the 7th Arc possibly being related to MH370. My acoustic research continues, characterizing previously undetected waves.
@370Location
My point is that the event deposited most of its energy in the upper couple of hundred meters of the ocean. That is what I call the ‘above jump layer’ mode.
Second-most energy went into the SOFAR channel, but this energy was less than in the ajl. Never the less you say that it was the largest SOFAR signal for many hours. You are probably right about that, but it just emphasizes how enourmous energy went into the ajl. Therefore it must have been an event comparable to the impact in Chelabinsk (Russia). That one was hard to find in IR satellite signals. After a couple of weeks some experts found it, but it was a small signal.
Least energy went into seabed R-waves. That is directly inconsistent with a dominant seabed impact.
Therefore I think the only way it could have been MH370 crashing would have been a very late crash. However, that means low energy, unless the airplane was carriyng a nuke (probably not realistic).
@Victor
Do you agree that NO detection at that time and place indicates that the curved paths were developed because G550 did not see anything resembling MH370 during its entire flight? That implicates a late FMT.
Initially, it would have been entirely reasonable to assume that MH370 flew slowly and along a curved path. Particularly since that also improbed agreement with the received power.
@Viking: I think we have to be careful about drawing conclusions about statements we don’t fully understand. In the article I cited, I tried to draw some conclusions about possible paths, but it was all speculation.
According to posts on other blog sites the Chinese deep sea research vessel Tan Suo Yi Hao is currently conducting activities in the vicinity of Lyne’s hotspot.
This vessel has been circumnavigating Australia for the past few weeks including passing through Bass Strait between Tasmania and mainland Australia, but apparently has recently changed focus.
Has BFO and BTO data from the same Inmarsat satellite through which MH370 communicated been analyzed for any other flights on or around March 8, 2014 whose flight paths are well known? I was wondering what analysis has been done to assess (1) how accurate/reliable the satellite’s historical BFO and BTO data is, and (2) the accuracy of the methodology for computing the relevant arcs based off that data.
The idea of there being some extra evidence the militaries don’t want to disclose to protect the secrecy of their capabilities is fun to think about. But I don’t think there will be anything interesting from in that data. There’s really only three scenarios:
1. Military assets picked up information that would significantly impact the search location. In this case, they would have likely found a way of covertly revealing it to the investigators. For example via the US intelligence practice of “parallel construction”. Say, by claiming a submarine in the area picked it up. Or just giving the investigators a nudge. Even if they were somehow unwilling to reveal it, at least one person who knew such significant information would have surely had the urge to brag about it or a moment of consciousness and leaked it. See the flight sim data.
2. Military assets picked up the plane, but it merely corroborated existing information. In this case, there is no point in disclosing their capabilities just to confirm existing data. Even if it tracked the plane slightly further out than civilian radar, there’s no point in revealing that if it matches the predicted track anyway.
3. Military assets did not pick up the plane. In this case there is no point in disclosing their lack of capabilities. If anyone believes you should have seen it you refuse to comment, claim it was offline for maintenance, you lost the data, or some other excuse.
@RazU: The BTO and BFO models were validated by Inmarsat (Ashton et al, The Search for MH370, Journal of Navigation, 2014) and the DSTG (Davey et al., Bayesian Methods in the Search for MH370, November 2015). Included were past flights for 9M-MRO and contemporary flights of MH370. The statistics for expected errors of the BTO and BFO were determined from these flights, as well as on the ground data from 9M-MRO before MH370’s takeoff.
@Charm: I tend to agree with you that the collective militaries have no further data that could help find MH370, but I see no harm in pursuing those leads.
@Victor: Exactly what I was looking for. Thanks! (And looking forward to the next weeks of reading.)
@Viking wrote:
“Second-most energy went into the SOFAR channel, but this energy was less than in the ajl. Never the less you say that it was the largest SOFAR signal for many hours.”
I get your point, but apparently am unable to convey mine. You are still conflating the undetected surface impact with the anomalously loud seabed impact. If even a small portion of the MH370 crash energy went into the SOFAR channel, it would have been detected. The hydrophones are dominated by clutter from multiple surveys using towed airgun shots. They are in relatively shallow coastal water, which allows the reflections from seabed and coastlines (upslope/downslope T-wave mechanism) to enter the SOFAR channel as H-waves. That doesn’t happen over deep water.
If we had been talking for years about the Java anomaly being among the loudest hydroacoustic events of the day with 7th Arc impact timing, there would be no question that the source was MH370.
Regarding your earlier comment about most of the energy of a seabed impact being R-waves, the CTBTO hydrophones are suspended in the SOFAR channel, so not well poised to pick up seabed waves.
@All
Just a note on the EK407 sighting by Capt Martyn Smith of a close call with an unidentified flight in the SIO on Mar 8 2014. The unfounded WSPR detections and extrapolations from that are ridiculous. If it were indeed a military plane on the correct heading from Diego Garcia, most likely it was on its way to Butterworth AFB or wherever multiple countries were converging assets to join the search effort near IGARI at the time.
@370Location
The US Navy reported that a P-8A Poseidon arrived in Kuala Lumpur on 14 March to assist with the search effort. That aircraft subsequently re-deployed to Perth when the search moved to the southern Indian Ocean.
Pacific Fleet Commander Recognizes P-8 Squadron MH370 Search Efforts
It’s hard to believe the US Navy would have deployed an aircraft to Kuala Lumpur for local search efforts if the US already had information that MH370 had gone south. That tends to debunk the notion that P-8A aircraft were searching the Indian Ocean on 8 March, as claimed elsewhere.
@370Location
Concerning the energy distribution, I am in contact with two geologists in Australia. They have data from detectors predominantly sensitive to ajl signals. They are totally baffeled which type of event could have put such a large amount of energy into that mode. It is something like a magnitude 4.5 earthquake on the surface of the ocean.
For comparizon there was an explosion in a depot for fireworks in the Netherlands while I worked there. It involvolved 300-500 ton of explosive material, and practically leveled a medium-sized town with the ground. The liberated energy equaled a magnitude 2.5 earthquake. This event was at least 100 times more intense, since the other modes also carried some energy.
That means something like a 50 kTon nuke. There is no way an airlane with empty fuel tanks can liberate so much energy.
@370Location
To be fair, I need to explain why I still do not consider your solution against the laws of physics. The reason is that the meteroite may have hit MH370 just before impacting the sea – causing the large explosion. That would explain everything.
However, the probability is low.
@Andrew, Thanks for the info on the P-8A going to KL. Your conclusion is spot on.
@Viking variously wrote:
“My point is that the event deposited most of its energy in the upper couple of hundred meters of the ocean. That is what I call the ‘above jump layer’ mode.”
“I am in contact with two geologists in Australia. They have data from detectors predominantly sensitive to ajl signals.”
“…the meteroite may have hit MH370 just before impacting the sea – causing the large explosion. That would explain everything. However, the probability is low.”
Geologists typically use seismometers on land to analyze vibrations in the Earth’s crust, which are insensitive to remote ocean surface events. AFIK, the only “detectors predominantly sensitive to ajl signals” would be sonobuoys dropped by military aircraft near a suspected source. Do tell your source for an M4.5 surface event in the SIO.
Your new scenario of a rare huge meteorite striking MH370, not at IGARI causing the turnback, but hitting the improbably impaired plane just as it crashed hours later into the SIO, is, well, stunning.
We can agree that the the probability for that is low, likely incalculable.
Where is a statistician when you need one?
For those interested in aircraft accident report quality, clearly this is thorough, well presented and its findings useful, including seat belt fitment:
https://www.youtube.com/watch?v=dum4SfnX8uk
For those with a deeper interest, the complementary written report:
https://www.atsb.gov.au/sites/default/files/2025-04/AO-2023-001%20Final.pdf
@370Location,
“There are three kinds of lies: lies, damned lies, and statistics.” (Attributed to Benjamin Disraeli by Mark Twain)
@ Viking
Re: “That means something like a 50 kTon nuke. There is no way an airline with empty fuel tanks can liberate so much energy.”
I’m not sure what the discussion is regarding ‘50kTon nukes’ is about? To offer some perspective, running a quick monte carlo (2k runs) for MH370 impact with ocean surface using EOF spiral descent data from Victor Iannello, Mike Exner, and ATSB results in a fairly wide Kinetic Energy budget range of 1.1GJ to 9.7GJ with an average of 4.8GJ.
This KE budget could generate respective Seismic Energy budgets of 9.3kJ (min), 880kJ (max) and 241kJ (avg). The weak seismic signal (if any???) is more difficult to calculate but could be on the order of ~ -0.8 to 0.8 Mw. More calcs would be required to prove out tighter.
A rough (unchecked) calculation of source level power range 205–215 dB. Gforce deceleration rough order of magnitude is ~450g, largely occurring in milliseconds.
The Chinese deep sea research vessel Tan Suo Yi Hao has now tracked westward along the Diamantina escarpment to about 20NM or so from the 7th Arc. Trajectory here (from: Ocean Infinity News and Updates Facebook page):
https://scontent-syd2-1.xx.fbcdn.net/v/t39.30808-6/490349637_1229057655890750_7957958215321867623_n.jpg?_nc_cat=109&ccb=1-7&_nc_sid=127cfc&_nc_ohc=vNpKCyC22LEQ7kNvwEbz8lp&_nc_oc=AdmnGdaOm820gXsItz4PWzJ7D3TxAa6VbkQL6N_1OobXyTxNrvaAJOrN6GszozSzNBI&_nc_zt=23&_nc_ht=scontent-syd2-1.xx&_nc_gid=ZktZISFQyNHMLw2sZTkEsw&oh=00_AfG65sdxZyOCUv6ToTSqgJsyXVIkDKcsuo8LiGo5YC6QYQ&oe=68014E5E
@John
Whether the Tan Suo Yi Hao is hotspotting or otherwise, best of luck to the Chinese Vessel on it’s venture and research at this time of the year.
@John
@Joseph Coleman
The Chinese conducted a similar mission in early 2023, exploring the Diamantina Trench and other deep-sea features in the Oceania region. The present operation is very likely a follow-up to that earlier mission rather than a search for MH370, as purported by commentators on other websites. There is some speculation the present operation may have a military purpose, but that is not confirmed.
Seabed sensors and mapping: what China’s survey ship could be up to
All,
Jeff Wise so enjoyed my video on WSPR tracking that I thought it only fair to make one on his favorite topic – Kazakhstan.
https://www.youtube.com/watch?v=iTcjjc5N-98
@sk999, nicely done 🙂
I’ll just add one point which has sometimes been under-appreciated when considering a “northern route.” The probability of a foreign aircraft cutting across the Eastern seaboard of India, and then crossing over across the Northern border into Nepalese / Tibetan / Chinese airspace unchallenged is approximately equal to zero, even in 2014.
Best,
Adi
@sk999, … Thank you for thoroughly debunking the “takeover from EE bay” hijacking scenario. JW is famous for uncovering the backgrounds of the Russian and Ukrainian passport holders, yet his unwavering commitment to the “northern route” theory left him oblivious to its numerous flaws.
However, I find it hugely problematic that both JW and Florence de Changy (FdC) shy away from commenting on whether the persons who used the Ukrainian passports to board the plane are who they purportedly are.
By piecing together the details provided by JW and FdC, it can be inferred that the two Ukrainians might have been compensated and placed into a witness protection program, while their passports were potentially utilized by others to execute the so-called “unlawful interference by a third party.”
A few countries are notorious for “borrowing” the original passports of real people and using them for nefarious political activities.
@ sk999 … BTW do you know if the request to top up the emergency oxygen for the flight deck was conveyed via EICAS?
@sk999: Thank you for creating and sharing this video. My feedback: Unfortunately, every second sentence of yours starts with “Comrade Jeff”. This derisive tone and ad hominem attack is uncalled for and unfortunately distracts from your arguments, which I find generally solid – except for the first argument you present in your video: Jeff suspected an abduction to the north very early on. What is your point there? By itself, this is not a meaningful argument in my opinion. Granted, this fact can show a cognitive bias (which none of us is completely immune to, so I don’t think it’s clever to mock this) but it can also be the result of something else entirely: As more and more facts come to light, they may all neatly fall into place, fitting the same pattern (or jigsaw puzzle) and confirming a pre-established theory or hunch.
Bottom line: Your video would benefit from abandoning the personal attacks and focusing solely on the facts and factual arguments. This would be more scientific.
CanisMagnusRufus,
According to the Safety Information Report, the topping off of oxygen was a routine maintenance task. There is no mention of it being a response to an EICAS message or other notification.
Peter Norton,
Thanks for your feedback. Sorry for the apparent derisive tone of my video. However, I was quite annoyed that Jeff chose to embed my WSPR video prominently in the middle of his own latest ad hominen attack, with which I desire to have no association whatsoever.
@Victor/all
Just a brief comment on the China research vessel, if I were going to search that area, I would first consult with Bob Ballard. I am thinking Ballard would search for debris trail in wider search zone inside but also outside BR proper. which is not what the Chinese survey is doing.
Hardly a day goes by without someone suggesting MH370 has possibly been found by OI or the China research vessel. Wishful thinking unfort, although the first group getting close to the correct crash spot could have quick success, if luck be with us.
@CanisMagnusRufus
I may have some input for you on your query of April 7 re: PAX video boarding. It is not my personal focus but I have info coming to me from others. In short, the PAX videos are NOT thought by some to corroborate FdC and JW characterization of several PAX as suspicious military muscle men.
Very well said on all counts, @Peter Norton!
Adi
@sk999 wrote
> Peter Norton,
> Thanks for your feedback. Sorry for the apparent derisive tone of my video.
> However, I was quite annoyed that Jeff chose to embed my WSPR video
> prominently in the middle of his own latest ad hominen attack, with which I
> desire to have no association whatsoever.
@sk999: Thank you for listening and responding. I was completely unaware of any background stories between you and Jeff and I can fully relate to your annoyance if your video was misappropriated.
Even if your annoyance is justified (and I am not here to judge this) I just wanted to let you know what impression an oblivious bystander like myself gets from this video.
Regardless, your technical arguments are well presented and really helpful to those of us who are less familiar with the technical details of a Northern route. So thanks a lot for your efforts in producing this video. Much appreciated. I have watched half of it yesterday and very much look forward to finishing it today. Cheers.
@sk999:
@CanisMagnusRufus:
@Andrew:
re: “the topping off of oxygen was a routine maintenance task. There is no mention of it being a response to an EICAS message or other notification.”
Famous Youtuber “Greendot Aviation” covers the topic of the oxygen replenishment in his MH370 video https://youtu.be/MhkTo9Rk6_4 at position 00:30:40:
« As for his own oxygen supply, he was at no risk of running out. His supply came from 2 large tanks underneath the cockpit which had enough oxygen to last 2 pilots 13 hours or 1 pilot 27 hours. These tanks are topped up a few times each year and luckily for Zahari – or perhaps more than just luckily – this oxygen supply had been topped up that very evening just before the plane had departed Kuala Lumpur. Take a look at this: It’s a scanned copy of the actual tech log from flight 370. This is one of the last pieces of paper generated by the flight before it left Kuala Lumpur. The tech log is the booklet where engineers and Pilots note down any faults with the aircraft and whether they need fixing before the plane goes on its next flight. There are 2 things which stick out in this document: First, under “defect description“ the engineer wrote “nil“ and “nil notes“ – in other words: no maintenance needed. He also crossed out these fields to show that nothing was needed. But then sometime later that evening, an engineer did make an entry: “crew oxygen system replenished to 1800 PSI“ – sometime between when the plane arrived at Kuala Lumpur earlier in the day and when it left again that night as MH370. Somebody decided that the pilot’s oxygen needed to be topped up. Whether it was Zahari has never been determined. »
Here are said tech log pages:
* https://archive.is/PTsYs/97e6938133f77ed337346232e48f8f41301fd996.webp
* https://archive.is/PTsYs/6afb5ec7bac3699d717d311860a4d458c7dd4965.webp
Also mentioned here:
https://www.mot.gov.my/my/Laporan%20Siasatan%20Mh370/02-Appendices/Appendices%20Set%201%20-%207%20Appendices%201.1A%20to1.9A/Appendix-1.6A-RecentTechnicalLogEntries.pdf
——————————————————
S/N 4918752 – 07 March 2014
Night Stop. Crew oxygen system pressure reads 1120 psi (EICAS).
07 March 2014
Crew oxygen system replenished to 1800 psi – EICAS.
AMM 12-15-08 refers
——————————————————
There is a related discussion at PPRuNe:
https://www.pprune.org/tech-log/583578-b777-crew-oxygen-system-2.html
B777 Captain Simon Hardy finds the replenishment very suspicious:
« There were also several last-minute additions to the jet that were shown in the technical log […]. The cockpit’s oxygen levels had been topped up before the flight […]. An extra note can be seen on the log, showing that the top-up was requested for the cockpit but nowhere else on the aircraft. Mr Hardy told the paper: “It’s an incredible coincidence that just before this aircraft disappears forever, one of the last things that was done as the engineer says nil noted[no oxygen added], then someone else gets on onboard and says it’s a bit low. Well it’s not really low at all… it’s a strange coincidence that the last engineering task that was done before it headed off to oblivion was topping up crew oxygen which is only for the cockpit, not for the cabin crew.” »
source: https://archive.is/nLm8P#selection-631.0-701.221
To a layman like me this sounds indeed suspicious. Not the replenishment itself but the fact that replenishment was not planned but than apparently overruled by someone else and the fact that (if true) only the crew oxygen system was replenished.
@Andrew: Could you maybe chime in to tell us from your professional experience whether or not this is something out of the ordinary ?
@Victor Iannello:
Was Simon Hardy’s hotspot in the Geelvinck Fracture Zone covered by the latest OI search ?
« He calculated the most likely position of the remains of the doomed flight. And he was invited to join the search with the Australian Transport Safety Bureau (ATSB) in 2015, where a team of experts were leading the hunt. But his calculations put the resting place for the plane just outside the official search area – and despite consulting on the operation, he never had the chance to prove his theory. […]
He suggests the plane’s pilot would have been in control the whole time – attempting to neatly plunge the plane into the ocean so it could never be found in a spot known as the Geelvinck Fracture Zone. The trench is hundreds of miles long meaning the pilot would have had manoeuvre room when deciding when to ditch the plane. […]
Another document of note is MH370’s operational flight plan. It which shows that an extra 3,000kg of fuel was added to the plane – the maximum amount of extra fuel that can be added to a Boeing 777 flight. The fuel would have given the pilot 30 minutes more flying time or more importantly, Simon explains, more time to ditch the plane in the ocean in daylight. He told The Sun:
“If you want to do a good ditching, you do it in daylight or at least half daylight. In the case of MH370, if the pilot has another half an hour of fuel it will be daylight. Another half an hour of flying would be another 244 nautical miles and the most important thing is that it will be dawn. […] Too much leftover fuel would leave an oil slick on the surface and reveal the plane’s final resting place. […] Even if you have tonnes and tonnes of fuel and it’s at the bottom of the Geelvinck Fracture Zone it still will be leaving a plume of oily rainbow residue on the surface for years. He wants to preserve the aircraft but he doesn’t want to save the passengers. It’s all part of it being planned meticulously for, ‘how can I make it disappear, I don’t want tonnes of fuel but I do wanna go as far as possible. If you’re of a motive to make it disappear then only one solution is to ditch it as neatly as possible, so it sinks to the bottom with all the people inside, with all the flotation devices inside, with no baggage. That’s what you want, if you want to make it disappear, you don’t crash it you ditch it.” »
source: https://archive.is/PTsYs
Personally, I think any water landing amid the significant waves and swell of the Roaring Fourties will obliterate the aircraft, no matter how hard you try. Keeping the aircraft hull largely intact like Sully on the flat Hudson River is not possible in the open ocean IMO. Or does anybody of you know a successful water landing in the open ocean ?
But even if it is impossible, this doesn’t necessarily falsify Hardy’s theory that whatever perpetrator could still have tried to do so (i.e. try to keep the aircraft in a few large and sinkable pieces).
@Peter Norton
Claims about the addition of extra fuel and oxygen, as reported by The Sun, are hyperbolic nonsense.
There’s no “mystery” about the “extra” 3,000 kg of fuel. That fuel is clearly shown on the flight plan as “COMP FUEL”, ie company fuel. Company fuel is fuel added by the airline to cover contingencies such as unplanned holding, based on the airline’s previous experience of operating to a particular port. Beijing and other ports in China are notorious for extended delays that have the potential to cause diversions and attendant disruptions to the operation. Consequently, airlines often carry “extra” fuel to mitigate that risk. According to Malaysian’s operations manual, extra fuel added by the captain is annotated on the flight plan as “SURPL”. In this case, no such extra fuel was added by the captain.
As @sk999 mentioned earlier, replenishment of the oxygen system is a routine maintenance task, normally completed at an airline’s home base during extended stopovers. In this case, the circumstances are outlined on p.47 of the MH370 SIR. Airlines typically replenish the system well before the pressure reaches the MEL minimum requirement for dispatch.
As for the notion that “extra oxygen was also added at the last minute but only to the cockpit“, words fail me. If they’d bothered to do some homework, they would have found that, unlike the crew oxygen system, the passenger system uses chemical oxygen generators that do not need to be replenished.
@Andrew:
Many thanks for your technical insights. I thought the “only to the cockpit” part may have some relevance since B777 Captain Simon Hardy heavily emphasized this point. But apparently (if I understand you correctly) this is not noteworthy because only cockpit oxygen can be replenished.
In another comment (pending publication) I asked you whether it might be suspicious that the replenishment was initially not planned but than apparently overruled by someone else (see tech log).
Maybe you can give some feedback once the full comment is published ?
Thanks again.
@TBill
No matter what one thinks of the theories espoused by JW and FdC, the evidence they revealed after investigating the PAX and the pilots is reliable. JW and FdC may be stubborn, but they’re not lying about the evidence. It’s upto us to interpret that evidence.
@sk999, VictorI, anyone
According to the Green Dot video, the request to top up the pilots emergency oxygen tank was issued after flight MH371 arrived from Beijing earlier that day. The handwriting on the Technical log notes on the left that it WAS 1180 psi, and the same handwriting notes on the right the Oxygen tank was replenished to 1800 psi.
1. is it routine to note the crew Oxygen level as per EICAS on the technical log before each departure?
2. is the technical log read by the pilots prior to takeoff? How long before takeoff? Is there enough time to replenish the oxygen before takeoff, or can it be something on a ‘to do’ list to be completed before the next flight?
3. could something have happened on the flight MH371 from Beijing to KL that prompted this request to top up the oxygen tanks?
CanisMagnusRufus,
If I am reading my manual correctly, EICAS provides the ability to display status information. Here’s a description of how it works:
===========
The status display shows information that helps the technician make a decision on the dispatch status of the airplane.
The status display shows:
– Hydraulic system information
– APU information
– Crew oxygen information
– Status messages.
You use the status display switch on the display select panel (DSP) to select the status display on the multifunction display (MFD).
=============
So it is possible that a technician made a routine pre-flight check of the EICAS status page, saw the low oxygen pressure, and had the oxygen system serviced. But I know nothing about MAS operating procedures other than what is presented in the SIR, so that is just conjecture. I also don’t know if the technical log entries are chronological, but just below the entry r.e. the oxygen is one stating that the software in the EPESC (a component of the IFE) had been downgraded. Don’t think that that would have been ordered by the Captain. In case anyone is interested, the EPESC is powered by the left AC bus via the P110 power management panel, same as the SATCOM.
@CanisMagnusRufus
This crew oxygen top-up is sadly yet another distraction put about by someone who frankly should know better.
Topping up the crew oxygen system was performed by MAS engineering staff as part of the stayover check after 9M-MRO had arrived into Kuala Lumpur as MH371 on 7 March 2014. This would have been carried out long before the Captain had even arrived for duty at 2250 hrs MYT.
Anyone who has taken the time to properly review the oxygen system maintenance records would see that MAS actioned Aircraft Maintenance Manual (AMM) task 12-15-08 Crew Oxygen System Replenishment essentially any time that the crew oxygen system pressure was reported as being at or below 1200 psi. It was a standard top up conducted entirely independently of the operating flight crew.
As to why the crew oxygen needs to be routinely replenished, Andrew and I spent a good deal of time researching this many years back. The reason for the replenishment is that oxygen is bled off from the system:
a. when the flight crew check the oxygen flow on each of the flight deck masks as part of their pre-take-off checks, and
b. on engine start-up, as part of an automated system pressure/volume check.
Draw your own conclusions as to why the entirely mundane is being promoted in such a fashion.
@Peter Norton
RE: “But apparently (if I understand you correctly) this is not noteworthy because only cockpit oxygen can be replenished.”
Exactly. The crew system has two oxygen bottles that are replenished as required. The passenger system, on the other hand, has individual chemical oxygen generators located in the passenger service units above each row of seats. When activated, the generators produce oxygen by a continuous chemical reaction between iron and sodium chlorate, for about 22 minutes. The generators are only replaced during major servicing, or if they have been activated for some reason. Unlike the crew system, the passenger oxygen generators are not ‘replenished’ during routine day-to-day servicing.
Regarding your earlier comment, these claims about the oxygen system replenishment are arrant nonsense. I’ll start at the beginning and work my way through your post.
1. Greendot Aviation said: “These tanks are topped up a few times each year…”.. My experience in close to 30 years of operating large Boeing and Airbus aircraft is that gaseous oxygen systems are replenished more frequently than “a few times each year”. The B777 crew oxygen system has an automatic safety feature that bleeds oxygen for 25 seconds during engine start. That function, together with crew oxygen checks and ‘normal’ leakage causes the pressure to reduce over time.
2. Greendot Aviation said: ”First, under “defect description“ the engineer wrote ‘nil’ and ‘nil notes’…” That statement is incorrect. The “Nil” in the “Defect Description” column would have been written by the inbound captain when he signed off the tech log after the previous flight. The inbound captain writes up the details for the flight and notes any defects that occurred in the “Defect Description” column. If there are no defects, that too is noted. The engineer then wrote up “Nil noted” in the “Action Taken” column.
3. The next entry in the Tech Log is the oxygen replenishment. Greendot Aviation said: ”Somebody decided that the pilot’s oxygen needed to be topped up. Whether it was Zahari [sic] has never been determined.” The MH370 SIR states that the oxygen system was replenished during the stayover check, because the pressure was reading 1120 psi. In other words, the engineers noted the pressure during the stayover check and topped up the system to 1800 psi. That action was not initiated by Zaharie.
4. Simon Hardy said: ”It’s an incredible coincidence that just before this aircraft disappears forever, one of the last things that was done as the engineer says nil noted[no oxygen added], then someone else gets on onboard and says it’s a bit low.” Where on earth did he get the idea this was a last minute addition to the Tech Log? The aircraft arrived from PEK mid-afternoon and was on the ground for over nine hours before it departed as MH370. At some time during that nine hour period, the engineers performed a stayover check and a software downgrade, as noted in the Tech Log. Both actions were very likely completed well before Zaharie got anywhere near the aircraft.
5. You said: ”To a layman like me this sounds indeed suspicious. Not the replenishment itself but the fact that replenishment was not planned but then apparently overruled by someone else and the fact that (if true) only the crew oxygen system was replenished.” The Tech Log entry was “overruled”. The inbound captain noted there were no defects during the inbound flight, and during the stayover check the engineers noted the oxygen pressure was 1120 psi and topped it up to 1800 psi. The reason that only the crew oxygen system was replenished is explained above.
In summary: There is absolutely NOTHING out of the ordinary about the replenishment of the crew oxygen system.
Thank you all for graciously answering my questions.
The following questions may seem bizarre (I guess most of my questions are like that, LoL!), but have you ever noticed that on some of his videos JW seems to slur some of his words and seems almost drunk? What’s up with that?
@CanisMagnusRufus
Adding to Mick’s comments:
You asked:
1. “is it routine to note the crew Oxygen level as per EICAS on the technical log before each departure?”
Normally, the oxygen pressure would only be noted in the Tech Log if the system needed to be replenished. As Mick explained, the engineers noted the pressure had dropped to 1120 psi, which is below the limit at which they take action to top it up, per Malaysian’s engineering procedures.
2. “is the technical log read by the pilots prior to takeoff? How long before takeoff? Is there enough time to replenish the oxygen before takeoff, or can it be something on a ‘to do’ list to be completed before the next flight?.
Yes, the Tech Log is read by the pilots before departure. The Captain reviews the work that has been done, notes any outstanding defects that may be allowed by the MEL, and makes sure the dispatching engineer has released the aircraft for its next flight. The Captain then signs the Tech Log to accept the aircraft. The Tech Log is reviewed when the pilots arrive at the aircraft, normally about 45-60 minutes before departure. The Captain may not be able to sign the Tech Log immediately, as the dispatching engineer often doesn’t release the aircraft until 10-15 minutes before departure.
Strictly speaking, oxygen replenishment can be deferred if the engineers are pushed for time, provided the pressure is above the limit required for dispatch, per the MEL. However, airlines normally prefer doing it at their home base, rather than down route. In this case the engineers noted the pressure was low during the stayover check and had plenty of time to service the oxygen system.
3. “could something have happened on the flight MH371 from Beijing to KL that prompted this request to top up the oxygen tanks?” If something had happened during the previous flight, it should have been noted as a defect by the inbound Captain. In this case, the Captain wrote “Nil” in the defect description column, so I think it’s safe to assume that nothing untoward happened during that flight.
@Andrew:
Thanks again for your detailed insights and answers.
You said: « The Tech Log entry was “overruled”. »
I assume there is a “not” missing after “was”.
I explain what I mean by “overruled”:
In the tech log there is a table. The left column header reads “defect description” and the right column header reads “action taken” (as you know of course). There are 3 rows (or horizontal sections) shown in the picture, each denoting a defect/action pair: row 1 (D1/A1), row 2 (D2/A2), row 3 (D3/A3).
My point of concern is the section A2:
It has been crossed out and then subsequently* overwritten (with “crew oxygen system replenished […]”).
*The reverse order would make no sense.
Could this matter?
My thinking is that normally this section is crossed out after all checks are done and no action (A2) was taken. I assume the section is crossed out to void it. To my laymen eyes it seems noteworthy that despite the section being voided, there was a subsequent action taken and noted in A2. To me, the crossed out section means that this action (oxygen replenishment) was initially unplanned. The person overwriting the crossed out section A2 probably was the same person who crossed it out initially: MAS 0053.
The question is: What prompted this change in the course of action ?
Maybe a change like this occurs quite frequently in practice and is not at all something out of the ordinary. But I would like to ascertain this.
The reason why I am focusing on this seemingly insignificant detail is the combination of
(a) the crew oxygen playing a key part in the mainstream/mainline theory about MH370’s diversion and
(b) B777 Captain Simon Hardy and Green Dot Aviation both finding this detail very suspicious.
In light of the extraordinary circumstances around MH370’s disappearance, such a detail may matter and I wanted to ascertain whether or not there is something out of the ordinary in the tech log.
BTW, is anybody able to decipher the vertically overwriting text in capital letters in A2 ?
@Peter Norton
My apologies – yes, I meant to say “The Tech Log entry was not “overruled”.
RE: “My thinking is that normally this section is crossed out after all checks are done and no action (A2) was taken. I assume the section is crossed out to void it.”
If a defect is entered in the left column, it is always accompanied by an explanation in the right column for any action that was taken, even if only to say “Noted”. The right column is never “voided” with a diagonal line if no action is taken.
The diagonal line may not be related to the writing on that page of the Tech Log. Paper Tech Log pages have multiple copies and are self-carboning. If you rotate the image 90° to the right and then flip it horizontally, you will find the “vertically overwritten text” appears to say “MRO Enter”. That makes me think the page depicted is actually a carbon copy and not the original, and that someone has written on the back of one of the pages. That writing has been transferred as a mirror image to the front of the page.
As I said previously, I don’t believe there is anything out of the ordinary about the replenishment of the crew oxygen system. People are making things up to justify various theories, without properly considering the evidence.
All: I am puzzled by all the speculation re the reason for topping off the O2. As Andrew and others have explained, it was SOP. Besides, Z had all the O2 he needed for his plan, even if the pressure had been only 1120 psi. He did not need more than an ~1 hour of O2. If he flew to the end, he probably repressurized the cabin circa 18:25. Alternatively, he could have simply removed his mask once the plane was headed south. Either way, there was never any reason requiring the O2 to be at max.
@Andrew
When looking for the reverse “MRO Enter” you mention I was immediately struck by the impression that the diagonal slash and the “MRO Enter” were written by the same hand/writing implement. Both certainly do not look to be written by one of the hands/writing implements as the service notes. That was just my impression – I am not in anyway a handwriting expert.
The impression came from seeing a ‘focus’ not unlike seeing the hidden 3D image in a “magic eye” photo, namely the the diagonal slash and the “MRO Enter” characters stand out from the rest of the form as though they were written on a different plane (no pun intended). Is it a coincidence the letters MRO are the identifier of the plane in question?
@Andrew:
Thanks for your follow-up. That was very clever of you to flip the image so that it reads “MRO Enter”! (I had only rotated it and wondered why I could not decipher the letters.)
If “MRO Enter” appears as mirrored letters, you are right that this mirrored writing must be a carbon copy from someone writing on the backside. This opens up the possibility that the diagonally crossing line “belongs” to “MRO Enter” (i.e. it was also written on the backside along with “MRO Enter”) — particularly if you say that in practice, the right column is never voided. In this case, it (the diagonal line and “MRO Enter”) probably has nothing to do with the maintenance report and the crew oxygen.
re: “People are making things up to justify various theories, without properly considering the evidence.”
Fortunately I am exempt from this problem since I don’t have any particular pet theory. Just to let you know where I was coming from: I interpreted the diagonal line as crossing out the A2 (action) section, thereby voiding it. Forms are often voided in this way if a particular section is left empty (and/or the person filling out the form wants to signal that he/she has duly noted the section but it does not apply and thus is left empty). I am therefore not ashamed of this interpretation, considering that professionals much more knowledgeable than myself (B777 Captain Simon Hardy and Green Dot Aviation) both interpreted the maintenance report in the same way I did:
« under “defect description“ the engineer wrote “nil“ and “nil notes“ – in other words: no maintenance needed. He also crossed out these fields to show that nothing was needed. But then sometime later that evening, an engineer did make an entry: “crew oxygen system replenished to 1800 PSI“. Sometime between when the plane arrived at Kuala Lumpur earlier in the day and when it left again that night as MH370, somebody decided that the pilot’s oxygen needed to be topped up. »
Thanks to your discovery that the diagonal line likely belongs to “MRO Enter”, we all probably have misinterpreted the line as voiding the section. So, based on your feedback I no longer find the maintenance report suspicious.
But honestly, before your discovery and feedback here, I viewed the scenario exactly as Green Dot Aviation. Picture this:
(1) The inbound captain writes in D1 (“defect description”): “NIL”
(2) Engineer MAS 0643 subsequently works through his checklist, discovers nothing to be fixed and thus confirms in A1: “NIL NOTES” and crosses out the remainder of section A1 along with A2 (diagonal line) so as to void them since no more checks and actions are planned and scheduled. So the checks were already finished.
(3) Despite this, now all of a sudden the crew oxygen is replenished. This is noted in the already crossed out A2 section! To a laymen’s eyes this doesn’t look like proper procedures were followed, given that the paper output is confusing (A2 notes that crew oxygen was replenished but the section is crossed out!? This looks slightly suspicious or at the very least not right.)
(4) Add to this that the replenishment was not even mandated by MAS guidelines – the report states:
“The minimum pressure for dispatch as per the MAS Minimum Equipment List (MEL) is 310 psi at 35°C for 2-man crew and with a 2 cylinder configuration (as installed on MAS B777 fleet).”
(5) Now a once in a lifetime black swan event happens to 1 out of millions of flights. And in this black swan event the crew oxygen just so happens to play a central role (unproven but that’s the mainline working theory here so far). And the crew oxygen was initially not planned to be replenished (crossed out section) but then there was a change of plans and despite the crossed out maintenance action section, crew oxygen was replenished although oxygen pressure amounted to the quadruple of the MEL requirement.
Can you understand that viewed from this perspective, the entire chain of events looks suspicious ?
Your feedback now really sheds a different light on the situation, particularly because
(1) you say that in practice, the sections are never voided with a diagonal line like this (but there needs to be a verbal annotation) and
(2) you discovered the “MRO Enter” to which the diagonal crossline probably belongs to.
Based on these 2 facts, I no longer view the maintenance log as suspicious.
But do you see why one may have deemed it so if oblivious of these 2 facts ?
I think we all (Captain Hardy, GDA, myself …) erroneously interpreted the log in the way I described here.
I hope this explains where we were coming from and thank you Andrew for clearing things up.
It’s sad in a way, because I thought maybe we found a smoking gun that could help narrow down what happened and help in the search. But alas, as so often with MH370: smoke and mirrors …
> @airlandseaman: “there was never any reason requiring the O2 to be at max”
Sorry, but I don’t find this a good argument. The replenishment was also not necessary for the MEL (according to the quote above), yet it was done.
Sometimes things are not strictly necessary, but you want to be on the safe side.
The same logic potentially applies to a meticulous perpetrator.
And the perpetrator indeed seems to be a very meticulous one, considering the plane went totally dark at IGARI, not even sending any log-off message.
> John says:
> When looking for the reverse “MRO Enter” you mention I was immediately
> struck by the impression that the diagonal slash and the “MRO Enter” were
> written by the same hand/writing implement. Both certainly do not look to be
> written by one of the hands/writing implements as the service notes.
> That was just my impression – I am not in anyway a handwriting expert.
@ John: Sorry, but how do you determine whether a straight line was or was not drawn by the same person ??
I am sure not even a handwriting expert can determine this.
> Is it a coincidence the letters MRO are the identifier of the plane in question?
MRO certainly refers to 9M-MRO.
@Peter Norton
Thank you for your comments. In my mind’s eye the line and text in question float off the page together in the foreground. The line is not straight, has a pen on/off at each end and is drawn with the characteristic force of the writer and is consistent with the written characters IMHO.
I accept that is not proof itself which is why the word impression features in my post. I posted my observation because it reinforces the notion that the line was not voiding the comments box on the original maintenance sheet.
@CanisMagnusRufus
@Peter Norton
Gents,
Further to the previous discussion, have either of you looked at the leaked Royal Malaysia Police report, Folder 5 Aircraft Record & DCA Radar Data? It contains a selection of maintenance records relating to the execution of Aircraft Maintenance Manual (AMM) task 12-15-08 Crew Oxygen System Replenishment for not only 9M-MRO, but also the ill-fated -MRD, and -MRQ, over a period of about 3 months.
If you follow those records, you will note a few things:
1. Replenishment of the crew oxygen system per AMM 12-15-08 occurred routinely roughly every 30 or so days. Those records put paid to the notion that the “tanks are topped up a few times each year.
2. The trigger for executing AMM 12-15-08 appears to be when the crew oxygen pressure read 1200 psi or less (ie following the depletion of about 600 – 650 psi from “full”). Given that normal depletion of the system is driven by flight cycles (viz pre-flight crew mask checks and engine start-up), the frequency of the top-up varies depending on the sectors the aircraft has been operated on (at the time MAS were operating their B777-200ER on sectors as short as 3 hours (KL – Denpasar) and as long as 13 hours (KL – Amsterdam and Frankfurt).
3. The crew oxygen system pressure check appears to be part of the
transit and night stop maintenance check lists.
4. The check and consequent replenishment are routinely recorded as being the D2 defect and A2 defect rectification respectively.
Bottomline, when viewed within the context of two other aircraft in the MAS fleet operating similar routes around the same time, there is nothing even vaguely unusual about the replenishment of -MRO on 7 March 2014.
Beyond that, if you look at the sequence of tasks recorded in the -MRO Tech Log, does anyone seriously entertain the notion that D2 was added at the instigation of the Captain, who didn’t arrive for duty until a couple of hours before the flight departed, and then, only an hour or so before the aircraft was meant to depart, someone decided to add the D3 software downgrade task. That makes little to no sense at all.
According to Simon Maskell as reported in The Diplomat:
By analyzing this [WSPR] data, Maskell explained that his team had been able to largely discount most of the more outlandish theories about what happened to MH370 – and narrow down some of the theories that remained.
“The analysis we did indicated that there are three explanations that appear to be approximately equally consistent with the information we had at the time: there is a chance that a freak accident occurred and the crew were unable to communicate or land the aircraft elsewhere,” he said.
“[Or] it was a murder-suicide with the murderer alive when the descent occurred, [or] a murder-suicide with the murderer no longer alive when the descent occurred.”
https://thediplomat.com/2025/04/mh370-the-devil-is-in-the-data/
Despite Maskell’s many public statements, we’re still waiting to see a paper from him explaining how WSPR data can be used for anything related to MH370.
> John says:
> The line is not straight, has a pen on/off at each end and is drawn with the
> characteristic force of the writer and is consistent with the written characters
> IMHO.
@ John:
As far as I can tell, we can see the handwriting of 4 or 5 different persons in the lower half of the image:
1. the inbound captain writing in D1
2. employee MAS 0643 writing in A1
3. employee MAS 0053 writing in D2+A2
4. employee MAS 2259 writing in D3+A3
5. “MRO Enter” (the handwriting seems to be different from persons 1-4 but it’s hard to say that for sure)
IMO the handwriting of persons 1-4 are clearly distinct.
How can you tell that the diagonal line was drawn by person 5 and not by any of the persons 1-4 (or a 6th person for that matter) ?
@Peter Norton
Thank you for your explanation. I can understand why a layperson might be suspicious, but people who hold themselves out as ‘experts’ should know better than to make such claims without checking to see if there is some reasonable explanation. There are reasonable explanations for both the ‘extra’ fuel and oxygen replenishment that have been discussed previously, yet these issues continue to be raised.
@ Mick Gilbert: Many thanks! This is very helpful information indeed, which collectively tips the balance even more to the side of “regular maintenance task” (as opposed to something suspicious) IMO.
Some thoughts:
> 1. Replenishment of the crew oxygen system per AMM 12-15-08 occurred
> routinely roughly every 30 or so days. Those records put paid to the notion
> that the “tanks are topped up a few times each year”.
I don’t see a big difference between a dozen and a few times per year. But it’s true that this information makes clear that it is a regular maintenance task. That’s very helpful, thank you.
> Bottomline, when viewed within the context of two other aircraft in the MAS
> fleet operating similar routes around the same time, there is nothing even
> vaguely unusual about the replenishment of -MRO on 7 March 2014.
It may not be unusual per se, but the timing is still noteworthy in my opinion.
You say replenishment of crew oxygen occurs roughly every 30 days on average. Assuming daily operation, the chance that this occurred right before the fateful flight is only 3% !
Sure, this can be a mere coincidence. But a good detective in a murder investigation would take duly note of such “coincidences”. Particularly in the case of MH370 where there are so many other “coincidences”. At one point you reach the point where you say “surely these cannot be ALL just coincidences” …
> Beyond that, if you look at the sequence of tasks recorded in the -MRO Tech
> Log, does anyone seriously entertain the notion that D2 was added at the
> instigation of the Captain, who didn’t arrive for duty until a couple of hours
> before the flight departed, and then, only an hour or so before the aircraft
> was meant to depart, someone decided to add the D3 software downgrade task.
> That makes little to no sense at all.
Why does it make no sense to you ?
Yes, I happen to think we should “seriously entertain” all notions and give them some thought instead of discarding them outright.
For the sake of argument, let’s assume the Captain was the perpetrator – since this appears to be the mainline working theory here. Why is the following scenario impossible? The Captain arrives “a couple of hours before the flight departed” and tasks maintenance with D2 and D3. It seems logical that a software downgrade had to be done by a a different engineer (i.e. software engineer) than the oxygen replenishment. (Thus, A2 and A3 may have been done simultaneously rather than consecutively.) If we put ourselves in the shoes of the perpetrator, it’s obvious that he would like full crew oxygen reserves if his plan involved depressurizing the aircraft (as is speculated here). I don’t know about the effects of the EPESC software downgrade. But since it is a component of the IFE which apparently also handles cabin satphone calls, could the software downgrade disable cabin satphone calls ?
If yes, this would also clearly be something a perpetrator would want.
If no, can you think of any other reason why the software downgrade could be helpful to the perpetrator ?
@ Andrew :
Yes, I can agree with that.
As I stated above, I am not a firm believer of any particular theory and I think it is important to think through all possibilities wherever they may lead us, since this flight was so out of the ordinary. For this reason, I also like to think through the mainline working theory which suspects the Captain. I find it hard to believe that an apparently meticulously planned crime would leave not the slightest clue, trace or paper trail. Therefore I find it interesting to study such details, because I think that if the Captain truly did it, there must be a tiny clue hidden somewhere. No crime is perfect.
Victor,
WSPR has truly become the modern Ouija Board for divining cause and location of all flights that ever disappeared into the Southern Indian Ocean.
Either that or the author of the article may have misconstrued some of the information that she was given by Simon.
As for waiting for a paper from Maskell, we might as well be Vladimir and Estragon waiting for Godot.
@Peter Norton
Well, I believe it was the oxygen cylinders that ruptured that caused this accident. So if there was a flaw in the bottles, or the oxygen system it would more likely occur after the pressure had been topped up.
@Peter Norton asks: How can you tell that the diagonal line was drawn by person 5 and not by any of the persons 1-4 (or a 6th person for that matter) ?
I don’t understand the point of this question. Whether the person who wrote the reverse image lines and text also wrote on the other side of the form has no particular relevance to the veracity of observations about the text in question.
@Tim
Apologies if I have missed something, but what evidence exists that indicates oxygen cylinders ruptured?
>> @Peter Norton asks: How can you tell that the diagonal line was drawn by person
>> 5 and not by any of the persons 1-4 (or a 6th person for that matter) ?
>
> I don’t understand the point of this question. Whether the person who wrote the
> reverse image lines and text also wrote on the other side of the form has no
> particular relevance to the veracity of observations about the text in question.
@ John: You are of course not obliged to answer, but you said that “the diagonal slash and “MRO Enter” were written by the same hand”. I was just interested in how you think you can tell this from a simple slash/line, in case you care to answer.
I extensively explained the “point of this question” in my previous comments above. Prior to Andrew’s discovery of the mirrored “MRO enter” text, various people (including the 2 abovementioned pilots and myself) thought the diagonal slash was meant to cross out and void the action sections. To us it appeared suspicious that initially no oxygen replenishment was planned, the action section was crossed out, but subsequently the already crossed-out section was overwritten due to an apparent change in events, followed by the perpetrator probably using the replenished crew oxygen. I have already laid this out in great length above.
For this analysis, it is of interest who crossed out the section and when.
@Peter:
I have already addressed your questions. You don’t accept the answers. What hasn’t changed is what I see, and revisiting the evidence it continues to appear clear to me. So be it. Time to move on because this conversation isn’t usefully progressing the subject of this web site.
I respectfully disagree on all counts.
@john,
I have always maintained that MRO suffered a crew oxygen cylinder rupture. The evidence points to local electrical failures and a subsequent decompression.
Power interruption to L transponder, Satcom, comms, L AIMS, and autopilot/autothrottle failure would all occur if there was local damage to the left side of the avionics bay.
The pilots became unconscious shortly after the IGARI turn and the aircraft just meandered on, with no pilot and no autopilot in control.
@ Andrew:
As a follow-up to my comment above, in which I hypothesized:
« The Captain arrives “a couple of hours before the flight departed” and tasks maintenance with D2 and D3. It seems logical that a software downgrade had to be done by a different engineer (i.e. software engineer) than [the one carrying out] the oxygen replenishment. (Thus, A2 and A3 may have been done simultaneously rather than consecutively.)»
I just checked: According to Appendix 1.6A (“Recent technical log entries”) of the safety investigation report, defects D2 (crew oxygen) and D3 (EPESC downgrade) both carry the same serial number S/N 4918752, while corrective actions A2 and A3 were carried out by different engineers. This would fit the hypothesis above, right?
What does the shared serial number mean? Does it suggest defects D2 and D3 were flagged by the same person (for example the Captain) and at around the same time?
(If so, this would be technically in line with the hypothesis above.)
@Andrew wrote:
> That makes me think the page depicted is actually a carbon copy and not
> the original, and that someone has written on the back of one of the pages.
> That writing has been transferred as a mirror image to the front of the page.
Was this maintenance log page printed in one of the reports ?
If yes, this would beg the question why investigators were investigating carbon copies instead of the original pages.
I looked through the reports but couldn’t find this page as an image anywhere, only its plain text content (in Appendix 1.6A).
@Peter Norton
It’s quite obvious from the Tech Log the two tasks were signed off by different engineers. We’ve already discussed the oxygen replenishment at length – it was clearly NOT ordered by the Captain. Please also note that pilots do not task engineers to downgrade aircraft software – they have no authority to do so.
The shared serial number simply means the two tasks were entered on the same page of the Tech Log. Each page has a serial number, located in the top right corner.
I don’t recall seeing a copy of the Tech Log page in the investigation reports.
@Don Thompson
@Andrew
@Mick Gilbert
I looked through all previous discussions about EPESC (Enhanced Passenger Entertainment System Controller) here, but I don’t think this has been discussed yet:
(1) Does the EPESC software downgrade disable cabin satphone calls and/or other air-ground communication available in the cabin (given that EPESC manages the air-ground messaging function for the IFE system) ?
(2) Does the downgrade disable the “moving map” on PAX displays ?
(3) Why was the EPESC software downgraded? Does the report state any reason?
(4) What non-nefarious reasons for the downgrade are likely ?
Reason for asking:
It is obvious that a perpetrator would like …
• to disable the moving map on PAX displays to conceal the diversion from PAX.
• to disable all air-ground messaging from cabin.
• full crew oxygen reserves if his plan involved depressurizing the a/c (as some here think).
This would all fit together nicely:
The Captain orders crew oxygen replenishment (D2) and EPESC downgrade (D3), therefore D2 and D3 share the same serial number. This ensures full crew oxygen and disables the “moving map”, the cabin satphone and all-air ground messaging from the cabin.
As experts, can you please fact-check this scenario ?
@Andrew:
Sorry, we were writing simultaneously.
• Yes, I missed the serial number in the top-right corner. This demystifies it.
• re: “Please also note that pilots do not task engineers to downgrade aircraft software – they have no authority to do so.”
I understand that it would be strange to give such an order and it would probably be hard to find an excuse for it. But in the end, doesn’t the Captain have the authority to demand any maintenance action he deems required for accepting responsibility for the aircraft?
• re: ” the oxygen replenishment […] was clearly NOT ordered by the Captain”
Sorry if I missed something, but why is that impossible ?
@Peter Norton
No, a Captain does not have the authority “to demand any maintenance action he deems required…”. A Captain, in consultation with the engineers, might request some maintenance action or other be performed, but ultimately it’s up to the engineers whether or not they do so. If the Captain isn’t satisfied, they can refuse to accept the aircraft, but they can’t demand the engineers perform certain work. Besides, what legitimate reason could a Captain possibly have for requesting an EPESC software downgrade, which has no relation to the aircraft’s airworthiness? Such a request would have raised a big red flag to the investigators, yet it has never been mentioned.
Why do we keep going round in circles about the oxygen replenishment? The SIR describes the circumstances that led to the oxygen servicing, yet you keep trying to infer some nefarious intent.
You are barking up the wrong tree on both counts.
@Andrew:
re: “Such a request would have raised a big red flag to the investigators, yet it has never been mentioned.”
Exactly, that’s why I am enquiring about it.
I find it an interesting line of inquiry, particularly in case the downgrade conveniently disables all air-ground messaging from the cabin (including the cabin satphone) and the “moving map”. A perpetrator would want that.
If you (or @Don Thompson, @Mick Gilbert …) could confirm or refute this, I would be thankful, because if the EPSEC downgrade doesn’t disable satphone, messaging and the PAX display map, it would end this line of inquiry right there.
Is the real reason for the EPESC software downgrade mentioned somewhere ?
If not, what non-nefarious reason is likely ?
You wrote:
« A Captain […] might request some maintenance action or other be performed »
This is what I would expect.
So in principle, the Captain could have requested the maintenance actions A2 and A3.
re: “what legitimate reason could a Captain possibly have for requesting an EPESC software downgrade ?”
I already conceded that “it would be strange to give such an order and it would probably be hard to find an excuse for it.”.
But we are all human beings and I assume that in these situations human factors are involved. The Captain was a very distinguished pilot, type-rating instructor and a type-rating examiner for the company. According to various interviews (including cabin crew), he maintained good relationships with other company employees. Furthermore, Malaysian society and culture is said to be more hierarchical than the West.
Therefore, all in all, I think he could have gotten away with an unusual request. The engineer would probably find it strange, but maybe there was no time to discuss it (airline schedules are always tight). And if it’s the Captain’s last flight, he won’t have to justify this demand later on.
This is my take as a layperson, but of course I defer to your hands-on experience and insights.
As far as the crew oxygen is concerned:
Your discovery of the mirror-inverted “MRO enter” text was excellent. But I don’t remember that we have yet established that “the oxygen replenishment […] was clearly NOT ordered by the Captain”. Apologies in case I have missed that. How can you exclude this possibility with absolute certainty?
re: “You keep trying to infer some nefarious intent. You are barking up the wrong tree on both counts.”
I don’t infer anything. Investigators pursue always pursue all avenues. This is one of them. It’s part of investigations that most lead nowhere until one leads to a break-through. I would like to first establish the facts before reaching that conclusion.
As many will have seen, MH370 families relayed an OI statement giving underwater search coverage during the period 25-28 March 2025 – presumably corresponding with the period under contract. This yields a coverage rate of 452sq.km per (full) day. Applying this to the 17.0 days spent in the zone 1200Z 11.03.2025 to 1200Z 28.03.2025 I surmise that they have probably covered 7500+ sq.kms. Additional are the 4.25 days apparently examining discrete data holiday spots between 23-28 Feb.
With reference to the presumed target zone of 15000sq.km outlined in March, I guess that all of the priority “holes” plus most of the planned swathe south of the arc have already been searched. Given the stated intent to return in November I think it is safe to assume that nothing has been found.
Where next??
A very common general reason for a software downgrade is that a recent upgrade intended to add a new passenger feature introduced an undiscovered bug: the system would typically be returned to the last stable version until the the bug is resolved. An example of a new feature might be personalised preferences associated with passenger’s frequent flyer membership.
The inflight entertainment equipment is an optional addition specified by the airline’s passenger experience team, and normally housed in a compartment in the passenger cabin. It would be gobsmacking if a stand alone passenger entertainment system software affected mission critical communication systems in any way whatsoever.
That is all I have to say on the topic.
@Peter Norton
Peter, for a bloke who purportedly is not a firm believer of any particular theory, you seem to be doggedly trying to make facts comport with one particular theory.
I am working on the basis that you are largely unfamiliar with airline operations, particularly as they obtain to servicings and maintenance. The notion that two unscheduled and entirely unrelated engineering tasks, neither of which were necessary for safe dispatch – crew oxygen system replenishment, and IFE software maintenance – could have been requested and completed inside of 70 minutes on a Friday night, an hour or so before midnight is preposterous. Airlines simply do not have maintenance crews of varying disciplines sitting around waiting to be called off the bench to undertake unscheduled, non-essential tasks at the best of times, leave alone at 11pm on a Friday night.
The reason each of the crew oxygen system replenishment, and IFE software maintenance tasks were initiated is detailed in the relevant tech log entry.
The crew oxygen system was checked as part of the “N/STOP CHECK” ((over) night stop check-list), and replenished in accordance with the relevant AMM instruction as the recorded pressure fell below 1200 psi.
The IFE software downgrade was initiated as a “- MAINT -” (Maintenance) item, and carried out in accordance with technical service instruction 77/SR/14092.
If either of those tasks had been requested by the Captain (or Operations), that would have been noted in the tech log entry.
In addition to Mick’s comments above:
Regarding the EPESC software downgrade, note that p.48 of the SIR lists the deferred defects from the MR2 section of the Tech Log. One of those defects relates to the Airshow system, deferred since 31 Oct 13: “In-Flight Entertainment (IFE) Airshow does not show arrival time/time to destination logged time & problem still persists.” I have no specific information, but the EPESC software downgrade may be an attempt to troubleshoot that problem. The action taken section states the downgrade was “carried out IAW TSI/77/SR/14092“. The reason for the downgrade is likely to be mentioned in that document. Note the Tech Log does not say the downgrade was carried out at the request of the Captain.
You might also like to note that Tech Log page S/N 4918752 isn’t the final page that was actioned before the aircraft disappeared. Appendix 1.6A of the SIR shows there were two subsequent pages of maintenance actions completed that day, S/Ns 4918753 and 4918754. Those pages show the additional work was written up in the Tech Log after the oxygen replenishment and software downgrade. That work includes terrain database loading, lubrication of the forward cargo door, an AD inspection on the lavatory waste compartment doors and flaps, and servicing of the potable water system. It is highly unlikely all those tasks were completed after the Captain arrived at work, rendering your scenario totally implausible.
@Peter Norton
My previous comment was also for you, in case there’s any doubt.
I can’t disprove your ideas to the standard of absolute certainty you seem to require. As I think Victor has said on several occasions, there are many theories about MH370 that we can’t prove or disprove; we simply don’t have enough evidence. All we can do is weigh up the evidence we do have, consider expert opinion, and assign probability accordingly. In my opinion, the weight of evidence puts your ideas at the bottom end of the scale.
Regarding the expert opinion you quoted, Greendot Aviation’s YouTube channel claims they have a background in psychology and are training to become a pilot. They don’t claim to have any particular expertise in airline operations. I don’t know much about Simon Hardy, but the evidence clearly does not support his claim about the carriage of ‘extra’ fuel, so how much weight should we give his other claims?
I continually forget to ask this question: MH370 would be the first flight in history to have all passengers turn off their cell phones. Doesn’t that reality mean there could be hundreds of additional pings that could at least further narrow the search zone??? How is it humanly possible that no one has attempted to look at this, particularly when after one of the most important clues being the FO’s cell phone? And how is this never addressed at all in either the ATSB or Malaysia reports?
There are others here who are more qualified to comment, but all the cell phones in the world would make little difference if there were no cell towers to detect their signals. The cell tower on Penang Island that detected the FO’s phone was very likely the last such tower that could have detected signals from phones onboard the aircraft. Consequently, there was no way to use those signals to narrow the search zone.
Cell phone base station towers are designed not to waste energy radiating signals upwards above the horizontal plain to where planes fly, although antenna technology does mean that there are some spikes of signal going in different narrow directions above horizontal. Whilst it varies, maximum cellular range is typically less than 40km in the horizontal plain and less above that.
An area that can now handle millions of calls would only support a few hundred mobile phones prior to cellular technology. Limiting the viable communication distance from a mobile phone to a cellular base station is a critical element of the “cells” of cellular phone technology, because it allows the reuse of frequencies within close proximity, which in turn is the key to allowing the explosion of simultaneous connections available today.
@Andrew
“The cell tower on Penang Island that detected the FO’s phone was very likely the last such tower that could have detected signals from phones onboard the aircraft.”
An aircraft at cruise rapidly passes through the unintended narrow fingers of cellular signal above the horizontal plain, meaning that phone connections, if made at all, are transitory lasting only a few seconds.
@All
I have followed Victor Iannello’s rational and informative blog here for years before being brave enough to make a post. An observation I might make is the word “hypothesis” is increasing being used incorrectly in place of “conjecture”. I lament a transition from the sublime to the ridiculous and hope that common sense (AKA freedom from logical fallacies) prevails.
paul smithson asked: Where next??
From my perspective, areas further out from the arc and to the north and south along the arc have lower probability density, which means it will take a lot more searching. In my opinion, the most promising areas were already searched.
@BRS
Your comment makes more sense in the Malaysia flyover area, where there could have been additional brief tower connections (in addition to the known FO connect at Penang). Years ago, Malaysia had reportedly planned to search cell records for possible add’l connects by pilots and cabin crew phones, but not passengers. To my knowledge we have never heard the results of that study.
Cabin crew would have a SatPhone, that could connect via Inmarsat system, but no outgoing sat calls were made according to the detailed satellite logs.
PS- re: Phones- one thing I’ve wondered about for future, if the cabin crew had access to a back-up emergency StarLink phone system that could be a way to give global access independent of cockpit. But anyways, that is the type of system needed to do that remote connection anywhere, based on the thousands(?) of near earth satellites launched by SpaceX.
@TBill Yes; I was referring to tower connections over Malaysia. It shocks me this wasn’t even looked into. It strikes me as almost improbable that only the FO’s phone attempted a call and/or that no one attempted an SMS message. Do we know for sure none of this was explored? Is it possible this info was held back for some reason?
Assuming the search continues to be unsuccessful upon resumption, at what point does the weight of incorporating these negative results start to materially sway a purely neutral (but logical and open to all rational possibilities) observer away from the existing search strategy? In other words, what additional data would a lack of success provide to the leading theories, in helping to refine them? I then wonder where the tweaked search regions would be, and whether they start to diverge quite a bit from each other (as opposed to the current situation, where there is meaningful overlap, or some semblance of adjacency.)
Probably a meaningless (and hopefully ultimately unnecessary) exercise at this stage. However, we should be prepared for this eventuality. Like I stated earlier, our collective understanding of probabilities continues to be rudimentary WHEN dealt with a starting situation whose probability was extremely minuscule to begin with, but has still happened (the loss of AND the 10 year inability to find a wide body civilian aircraft.) You don’t want to speculate for the sake of speculation, you want to carefully analyze the data that is trusted, and you want to use the common-sense smell test. Yet, if the search continues to be unsuccessful, where does careful logical thinking take us?
I feel it is more important to fully internalize these questions than to quickly attempt an answer.
Best,
Adi
First, I wish a Happy Easter to everyone.
@paul smithson @Victor Iannello @all
Probably the best and smartest thing OI can do is to go back to ATSB recommended area and search data gaps and everything what has been left.
I am repeating occasionally on this blog since the search for ARA San Juan about the necessity to do this first before moving the search somewhere else. In my opinion the start should be Geelvinck Fracture Zone which passes near the 38S (37-39S, ATSB’s hotspot). It is a very challenging terrain for underwater search and the plane could have been easily missed at the bottom of some slope. If the plane is not found there, then the rest within 25NM from the Arc should be searched next. I don’t want to elaborate more because probably no one wants to read that anymore.
The plane has actually been located, but not physically found, we are just not aware of that yet.
I have been saying for many years that the best prospects for finding Malaysian 370 are in unsearched locations wider from the 7th arc in the UWA area between 28.3°S and 33.2°S, most likely at or near Broken Ridge between 32°S and 33.2°S.
@Adi
@Marijan
Both of you make very good cases for alternative thinking, well said.
All efforts hitherto, have effectively been ‘anchored’ in a ‘totally devout devotion’ to ‘the ‘rigid gospel’ according to INMARSAT / ATSB / DSTG, of a ‘ghost flight’, from the FMT (or very soon thereafter) to the seventh arc. Not even the relatively recent concession by the IG / ATSB that the possibility exists of a consciously piloted flight to fuel exhaustion at the seventh arc, and even further conceding that it is plausible that a glide of up to 120 nautical miles from the seventh arc ‘may have occurred’, doesn’t substantially change anything, not really.
The assumed ‘sanctity’ of the INMARSAT data, and the professed ‘infallibility’ of the exhaustive analysis conducted by many, has effectively ‘collared the thinking’ of the majority here to the seventh arc.
You could liken the situation that we find ourselves in after eleven long years of frustrating circular deliberations, to that of a group of neighbors agreeing to walk their dogs together, same time every second Saturday morning, taking the same path every time, discussing the glacial pace of the ‘forever redevelopment project’ that they pass both outbound and inbound, with the only difference being the loudness and repetitiveness of each individual dog’s bark, and the variable length of each owner’s leash.
Just like the dog walkers, we chew over the same issues, time after time, marching up and down the arc, with rarely ‘a new insight’ in sight. Albert Einstein famously said, “Two things are infinite: the universe and human stupidity; and I’m not sure about the universe,” and, “The difference between stupidity and genius is that genius has its limits.” He also noted that “Everybody is a genius. But if you judge a fish by its ability to climb a tree it will live its whole life believing that it is stupid.” It is time to change our thinking, lest we be eventually condemned for our collective stubborn, inflexible, stupidity.
Perhaps serendipitously, it is Easter Sunday, which celebrates the resurrection of Jesus. I think it is a fitting day to formally resurrect ‘logic’. I think it is way past high time to acknowledge that there is a growing probability, that 9M-MRO is most probably NOT ‘immediately proximate’ to the ATSB’s ‘hallowed’ seventh arc. The Inmarsat data (specifically the seventh arc BFO’s) have been heavily and consistently portrayed as the unassailable primary clue that defines fuel exhaustion, and thus the case for ‘proximate to the seventh arc’, but it is not an unassailable fact that fuel exhaustion actually occurred at that point. If we refuse to deny the distinct possibility that any other interpretation is valid, then we have to acknowledge, that in police investigation terms, we are going nowhere fast (nothing new) we are dealing with ‘an intractable cold case’.
“Heresy” they cry. How dare he disparage the accepted BFO doctrine ? “Boil him in oil” they cry (or impart severe pain by some other means). No other ‘faith’ has hitherto been welcomed here, let alone taken seriously (although, thankfully, up to this point, at least some of the more plausible ones have at least been tolerated). But none of them seriously challenged the ‘defining inferences’ derived from the BFO analysis (other than DennisW (R.I.P.)) and were thus no threat to the prevailing view of the ATSB and their supporters. But I can hear “indignant offended” screams rising now. Tolerance evaporates.
Whether we like it or not, we have to acknowledge, that regardless of Malaysia’s protestations to the contrary, Adi’s ‘common-sense smell test’ of the available evidence and clues make an overwhelming circumstantial case that the vanishing of MH370 was almost certainly a deliberate act by the Captain, and it was clearly meticulously well planned, and very skillfully executed, with the specific intention to ‘vanish without a trace’.
Furthermore, Marijan reminds us that there is no certainty in ‘one pass scans’ even with AUV’s, let alone Fugro’s towed side scan sonar. Ocean Infinity’s first attempts to find ARA San Juan failed, because they initially scanned ‘across the trench’ (that it was eventually found in), instead of scanning ‘along the trench’. Peter Foley (formerly of the ATSB) assertively stated during his presentation of the eve of the 10th Anniversary at ADFA in Canberra on the 7th March 2024, that 9M-MRO is definitely not in the GFZ (Geelvink Fracture Zone), presumably based on Fugro’s scans.
Given the ‘confidence’ (perhaps misplaced in hindsight) that the ATSB had in the technology at the time, I could understand Peter believed that in 2017. But now, with clear knowledge of the difficulties subsequently experienced in finding the ARA San Juan, I find it rather, unsettling that Peter should so forcefully reject the possibility that 9M-MRO could have been missed in the northern part of the GFZ which was scanned by Fugro (the small portion which was within the DSTG’s PDA).
Personally, I believe that the ATSB should have searched Captain Simon Hardy’s position, (it was highly localized and was only a few tens of kilometres outside what they did scan) but they did not, because it was derived as “a deliberate act” which Malaysia would not (and still will not) accept. In fact, I agree with Captain Hardy that Captain Shah did deliberately fly 9M-MRO to, and ditched in the GFZ, but my position is 40 nautical miles further south than Simon’s.
Whether or not Simon or I eventually prove to be correct or not is not the point.
The point is, that endless ‘number crunching of electronic signals’ is not the answer. If it was, we would have found it years ago. This was a criminal act committed by a highly skilled human. It should become a human factors investigation. The only way to find 9M-MRO is to work out what Shah’s motive and OBJECTIVE was, then we can work out his method of achieving his objective, and that will ultimately lead us to the crime scene.
@ventus 45 Your comment is very interesting. But the problem is (and this is also going in circles) that we don’t know which way the captain chose when flying over SIO. Was it avoiding and confusing the tracks with a constant change of course (including how it is described by Richard Godfrey and WSPR, but not necessarily just this way. I mean, even if the WSPR is incorrect, Zachary really could have chosen the path according to this logic), was it the maximum possible path to the south, and assuming this option, the OI would have to turn on the search area at 38S-40S (as far as I understand, 41S was already out of range 9M-MRO?).
If we take the flight simulator data, they only talk about direct flight until fuel is exhausted, but the course does not lead directly to the south, and for me it is still a bit strange why Zachary chose this particular point 45S 104E, when it would be possible to do it even further from Australia in the middle of the ocean (if the goal was to crash as far away from inhabited land as possible). Thus, it cannot be concluded that Zaharie would definitely aim to fly as far south as possible to the southern part of the 7th arc (38S-40S), but this is not physically impossible (with the exception of some drift analysis questions that do not indicate the location of MH370 south of 36S).
This is the main problem in the search for MH370 – we do not know which way the captain chose to hide the plane – to fly as far south as possible to the most remote part of the Indian Ocean, or it was not an end in itself to fly as far south as possible, or he did a lot of maneuvers and hid the plane in an unpredictable place (again, not necessarily in the WSPR location, but in any).
And it’s very good that we have a lot of expert analyses on this topic, but we absolutely cannot exclude any part of the 7 arc+ max. planning range.
So if the OI does not find the aircraft either on the 33S-36S or at the WSPR point, then it remains to expand the search area to the maximum planning range of 9M-MRO or move to the southern part of the 7th arc again, perhaps (if they have the opportunity and desire), they could scan the Geelvinck Fracture Zone in October 2025 and, if there is nothing there, resume the search in November 2025 according to the original plan. It would be useful and logical as an option. However, so far there are no signs that they are going to look in this area.
@Ventus
I can answer those questions based on my flight sim work and the data. The apparent goal was to hide the aircraft. The BTO/BFO are consistent with a best-fit 180CTH flight that was approximately maneuver-less until about Arc5 (from ISBIX/Arc2). At Arc5 there must have been a sharp descent with slow down, which by definition probably also requires a heading change to cross Arc7.
Another way to state this, looks like MH370 flew straight 180S and then simply merged onto the home flight sim path towards 45S. This interpretation requires re-assessment that home sim path was probably not a LNAV ghost flight to NZPG, instead the alternate interpretation is the path was 180CMH to the Magnetic south pole.
Actually interpreting the final destination as NZPG vs. 180CMH is probably immaterial. The crux of the problem was the assumption of ghost flight after Arc2. In other words, we mistakenly did not look at the flight sim work as the actual flight path that the pilot was planning to merge onto. Rather we assumed the pilot’s intent was to be dead and/or passively fly from Arc2, which I assert in 20/20 hindsight was probably the wrong way to look at the flight sim data.
@Ventus45
Exactly who do you think believes that the INMARSAT data has ‘sanctity’? I don’t see them in this blog. What does have ‘sanctity’ for many are the laws of physics that constrain what is possible and what is not.
The INMARSAT data produces an incomplete picture of the circumstances and interpretation must thus include human generated assumptions. If hypotheses that interpret INMARSAT data have not pinpointed the position of MH370 it is not because the INMARSAT data is faulty, or that the laws of physics are broken. It means that one or more of the assumptions built into that interpretation is invalid.
You say that “The only way to find 9M-MRO is to work out what Shah’s motive and OBJECTIVE was, then we can work out his method of achieving his objective, and that will ultimately lead us to the crime scene”. The many people who have taking this approach since the very event have not formed a consensus or drawn a successful conclusion, even after 11 years.
There is a human tendency, called confirmation bias, to draw from available evidence conjectures that support existing beliefs. (Confirmation bias is innate in the human condition and was likely evolutionarily important for survival allowing fast and timely decision making.)
Discarding attempts to understand real data observations in favour of conjectures of convenience or gut feelings is illogical and frankly silly. If not before, once the location of MH370 is discovered the INMARSAT data will be fully understood, although assumptions made generating hypotheses may well be invalidated.
Happy Easter!
@ventus45 said: The only way to find 9M-MRO is to work out what Shah’s motive and OBJECTIVE was, then we can work out his method of achieving his objective, and that will ultimately lead us to the crime scene.
You’ve had 11 years to do this. You complain a lot about what was searched, but other than recommend searching near Simon Hardy’s previous hotspot (which he himself has abandoned in favor of the Blelly hot spot, which was just searched), I haven’t seen you propose a rational plan for conducting the search. People have latched onto the math because without that we are left with a series of hunches that are widespread, cover large areas, and are impractical to search.
@Adi asked: Assuming the search continues to be unsuccessful upon resumption, at what point does the weight of incorporating these negative results start to materially sway a purely neutral (but logical and open to all rational possibilities) observer away from the existing search strategy?
That’s a very good question with no easy answer. Before the search started, I put the probability at 50% that the debris would be found. But with the highest probability areas already searched, I would reduce that probability. So more likely than not, we will find ourselves asking the question you asked.
I’m actually a lot more willing to challenge the integrity and/or interpretation of the BTO and BFO data than most people realize. I simply assign a lower probability to that than accepting the interpretation is correct and we have not yet found the debris field. But it is undeniable that each unsuccessful search increases the probability of low probability scenarios.
@All
Comments from several contributors during the last few days indicate a break-up in the solid belief about a crash within the proposed search zone. In the following, I present a slightly modified input written at the end of February this year, discussing my expectations about an unsuccessful search.
There are no signs Ocean Infinity found anything significant yet. I know that many of you think the highest probability area is still ahead, but I must admit that I find it most likely that they find nothing related to MH370 in the proposed search area in 2025, or the beginning of 2026.
Why do I say that? The main reason is that choosing a correct search area based on the official data is tricky. Partly because the problem is close to being under-determined, and partly since there is an unusually high sensitivity to changes in the weighing of the available data.
I think everybody agree that the BTO data is precise and reliable. The only way to compromise it, would have been a rapid hack of Inmarsat. I do not believe anyone outside the company could have done that, particularly since nobody knew where to find it and how to change it in a self-consistent way within the first few hours/days after the airplane disappeared.
However, the sad thing is that BTO only tells us the airplane came down somewhere near the seventh arc within the maximum fuel range of a starting point (plus roughly 200 km additional distance in case of an active and determined pilot). That is an intractably large area.
The BFO data is much less reliable than BTO. The problem is that it is sensitive to height changes during the handshakes, to temperature changes in the cabin, and to on-going turns. Smaller problems may also come from turbulence, decompression, and internal technical issues. Never the less, it is practically certain that the BFO data prove that the airplane crashed in the southern hemisphere, not in the northern hemisphere.
Any conclusions beyond these two rest on guesses or models of what actually happened. Some guesses are clearly better than others are, but at the end of the day, the airplane can in principle be anywhere near the seventh arc, from roughly 40 degrees south to the Indonesian coast.
This calls for including other information than the Inmarsat data end the maximum fuel range. There is plenty of data but most of it is typically ignored or treated incorrectly.
The first example is debris. Around 20 pieces found across the Indian Ocean are from MH370 with almost 100% certainty. A comparable number is likely from the airplane. Unfortunately, the dominant current in the Indian Ocean is largely parallel to the seventh arc in the relevant area, so the beaching pattern gives little information on the crash position.
Only the beaching time for the piece found in South Africa provides solid information. It eliminates any position near the seventh arc and simultaneously south of 37 degrees, and make positions between 30 and 37 degrees increasingly unlikely the further south they are.
That is one reason I am pessimistic. Of course, it is always dangerous to put all the weight on one piece of debris, but the first photo with fresh local biofouling still on it is strong evidence, and hard to fake with technology available 10 years ago.
More importantly, the biofouling on some of the other debris has valuable information. With exception of the piece from South Africa, all other biofouling is of tropical origin. In addition, chemical and isotopic analysis of it tells a story of a hot beginning, followed by a cooler period, and then a reheating to around 24 degrees.
That points to the northern part of the seventh arc from the northwest corner of Australia to the coast of Indonesia. Only very few experts are willing to accept this conclusion. Long ago, a clever person stated ‘If you want to find MH370 you should listen to the clamps’. I think it is time to listen to those words. Particularly since climate research shows many types of biofouling from temperate and subtropical regions is able to survive in tropical climate if the temperature increases slowly as it would when drifting along the seventh arc. Therefore, the absence of non-tropical biofouling is strong evidence for a crash in the tropics.
Practically nobody is willing to consider contrail evidence collected from weather satellites. Never the less there are intermittent signs of a contrails from the Malacca Strait down along Indonesia, past Christmas Island to a position around 370 km southeast of the island. There are no connected contrails after that point. Instead, there is a clear sign of a mushroom cloud at the end of the contrails. I find that strong evidence for an airplane crash. I admit the individual contrail segments are weak, but all the pieces line up nicely, and two seismic detectors confirm passage of an airplane at the correct time, speed, direction and distance from Christmas Island.
In addition, the most southern of the seismic detectors received four small peaks through different routes and propagation modes at the correct times consistent with an event causing the mushroom cloud. The arrival headings also fit with predictions. Using advanced triangulation methods developed for cosmology, it is therefore possible to back-trace the event to the area covered by the mushroom cloud.
Most importantly, a re-analysis of the data from HA01 proves that this detector also recorded the same signals (and a couple more) from the mushroom cloud position (and time). Together with the signals from Christmas Island, it is therefore possible to make classical triangulation of the event. The fit is excellent – only 1 km from the center of the mushroom cloud. That is where the search should take place. It will take no longer than an afternoon to find a debris-field. Most likely from MH370.
Happy Easter to everyone indeed 🙂
So interesting to see how logical analyses can diverge massively when you are analyzing the proverbial singularity (my very loose characterization of this event at the current stage.) One way to visualize this situation is the graph of the function sin (1/x) as x approaches 0 from either side of the x axis. The poor function just jumps up and down, more and more vigorously, not knowing what to do! The trick to understand this, of course, is to go sufficiently close and travel with the function precisely. Then you know that the results are quite logical. But, you need to know where to look, and you need to know the properties. One also needs to be comfortable existing at all possible values, which isn’t natural to human thought as it is to math functions, and to the laws that govern nature.
Anyway, sometimes we need “lucky” breaks to then be able to explain things, and Easter is as good a time as any to wish for one 🙂
Best,
Adi
@Viking
The elephant in the room is we are discounting evidence we have: namely flight sim data (which I now see as the actual flight plan map at least to Arc7), and I would probably add some discounting of debris drift to allow consideration for south-westerly crash sites.
BFO looks pretty darn good to me, with probably some minor drift as expected. The minor BFO drift is probably in the direction of making estimated flight paths look a little more westerly than it probably was in reality.
Many of us have looked for other clues (contrails/etc) but right now that’s all we have. I do indeed favor FOIA efforts as a means to broaden our public understanding.
Since you continued the discussion in the direction of what’s next, I want to add a few words regarding the search plan. The 7th arc BTO and BFO are the best data we have about the possible location, everything else is much more uncertain. Therefore, for me, the best search plan for the next Southern Hemisphere summer would look like this:
1. Search areas defined in the contract with the Malaysian government. If there is actually any obligation regarding the location that should be searched (or there is only a requirement regarding the area)
2. Start from the IG hotspot and gradually move along to the south as the weather improves and fill in data gaps ±25 NM from the Arc
3. If there are remaining larger areas not searched before but within the ±25 NM limit, search them as well concurrently with data gap fill-in
4. If the plane is not found by reaching the furthest southern latitude searched by Fugro, extend the search to virgin seafloor to 40°S, which was also ATSB’s southern limit
5. If the plane is not found by that time, consider expansion to ±40 NM
@Victor
Victor, you have some really nice maps which show data gaps further south which you were posting on X (Twitter). Do you mind sharing them here at some point?
Happy Easter
Unfortunately the recovered debris, by itself, seems to lack indisputable evidence to discern between the various End of Flight (EOF) scenarios. Each scenario below (and all in-between) has differing debris fields to be detected by underwater scanning:
1.0 Highest Speed Impact: bits and pieces (including majority of engines), Main Landing Gear (MLG) set or components, and significant small light weight floating debris.
2.0 Unsuccessful Ditching: airframe portions, wings, subcomponents, and smaller pieces, moderate floating debris.
3.0 Successful Ditching: Largely intact (but damaged) airframe with subcomponents, smaller pieces, significantly less lightweight floating debris. (However, the recovered In-flight Entertainment (IFE) internal frame is a challenge against this scenario)
Several underwater searches have been conducted, each with debris size scanning resolution limitations and gaps of coverage.
Scenarios 2.0 and 3.0 suggest a relatively dense debris field with larger debris pieces and seems detectable by scanning.
Scenario 1.0 has small size debris and if in deep waters suggests the least dense debris field. This scenario seems that it could be potentially missed by previous searches but I can’t assign a likelihood or probability, perhaps @Don Thompson or others can provide insight on why it wasn’t likely missed or could be missed…
The ability to have a high speed decent after fuel exhaustion is also something to consider. Without question Scenario 01 can have high speed decent (and impact). What are the B777 flight control considerations, capabilities, or constraints for a high speed decent AND recovery for a ditching event? This has been discussed before in detail but a summary reminder of challenges could be thought provoking.
So, Marijan. Are you happy to ignore the drift studies which all suggest the crash location is north of the IG hot spot?
@Godfrey Jack: Which drift studies are you referring to that suggest the crash location is north of the [UGIB] hotspot? David Griffin (CSIRO) puts the point of impact (POI) at 35S and Ulich and Iannello estimate a POI around 34S.
https://mh370.radiantphysics.com/2023/06/12/improved-drift-model-and-search-recommendations-for-mh370/
Meanwhile, Chari Pattiaratchi predicts a POI around 32S-33S. Unfortunately, he hasn’t released details about the hydrodynamic effects in his model such as whether he accounts for Stokes Drift. This contrasts David Griffin, who has released detailed information about the assumptions in his model and also has released the actual tracks from his simulation of virtual drifters, which Bobby and I directly used in our drift analysis. I suspect that Chari has ignored Stokes Drift, which is estimated to be about 1.2% of windage in the CSIRO model. If Stokes Drift is neglected, it would tend to reduce the transport speeds and predict POIs further north along the 7th arc. In any event, it is difficult to reconcile the differences between David’s and Chari’s models without more information from Chari’s models and his results, which he seems reluctant to provide.
Re: Debris Drift
New barnacle evidence may disrupt those previous drift studies. Most came after the flaperon was found, and incorporated false assumptions about barnacle growth rate and the arrival time of the flaperon. My report on barnacles was posted here in Mar 2024 to no response, but there was a lot of other activity around the anniversary.
https://370location.org/2024/03/barnacle-growth-on-mh370-debris-is-consistent-with-a-7th-arc-crash-site-in-tropical-waters/
Note carefully the following high resolution image of barnacles and the leading edge of the flaperon:
https://370location.org/wp-content/uploads/2024/03/NPR-inboard-hinge-rot-crop-closeup.jpg
The inboard leading edge is clearly abraded through the paint and fiberglass. The screw heads are worn down to a nub. This is not normal wear or damage from drifting in the open sea. It most likely came from grinding against sand/coral in surf.
Most importantly, the barnacles are growing on top of the abraded paint, fiberglass, and shredded nylon seals. Thus, the barnacles must have attached after beaching.
Please do read through the report to understand why barnacles grow much faster than assumed in early MH370 flaperon analysis.
David Griffin had concerns about the length of time it would take the flaperon to reach La Reunion from the consensus search areas. Only with stokes drift accelerating the drift, and an arrival at the end of July 2015 did his drift model fit. The barnacle shell growth analysis matches sea surface temperatures near La Reunion in early May 2015, months earlier than previously allowed. This definitely disrupts barnacle and drift studies by several authors.
The “Roy” debris at Mossel Bay in South Africa was the second piece of MH370 debris found in late 2015. It traveled the farthest and the fastest, through cooler waters, and was photographed covered with barnacles. Most other pieces of debris had no barnacles, except for a cabin panel found at Rodriguez Island a year after the flaperon. This is consistent with most debris never leaving tropical waters. Tracking Griffin’s published particle tracks, the cutoff crash latitude for debris not crossing into barnacle reproductive temperatures is around 21S.
I see no point in debating the intent of the pilot/crew based on the possibly false assumption of a flight to oblivion, especially when it conflicts with factual evidence. The scientific method is all about testing a hypothesis for flaws. (By null hypothesis testing against random input, I’ve just found that months of work verifying an impact event was contaminated by phantom signals. WSPR developers might do the same.)
Getting a negative result doesn’t necessarily kill a hypothesis, but it means it’s time to check the base assumptions.
Number one would be that MH370 flew on a constant heading shortly after leaving radar coverage, which showed previous navigation between waypoints. Number two would be that the paradox of barnacle growth starting in warm waters was due to some improbable predation, die-off, or lack of nutrients in their natural cool water environment.
@TBill
It is interesting that you mention the flight simulation data. I have never put much weight on it for one particular reason. The actual flight simulated took place roughly 12 hours shifted compared to MH370.
To see why that may be important some special mathematical knowledge is needed. The reason is that a 12-hour shift roughly inverts the relative impact of the satellite Eigen-motion. I am sorry to use such an unusual word (of German origin), but in several disciplines of physics and mathematics it is the offical termonology. In particular in quantum mechanics.
Why is that important? It is a long and complicated story, but the short (and slightly wrong) version is that one must convert the flight simulation data to take this effect into account. Doing that leads roughly to my solution near Christmas Island. For experts, I am willing to give the full explanation, but for those who are not experts, this short explanation makes more sense.
Initially I found the solution with Topology Optimization, but once I had it, I quickly realized the link to the simulation data.
@All,
Between now and search resumption time, I feel it is imperative for an informed and diverse group such as this to create a spreadsheet that lists the various proposed search / re-search areas. To be effective, organized, to capture all the details, while still being able to compare at a high level, here’s what I propose this spreadsheet should contain in terms of columns:
1) Name of the lead owner
2) Simple name for search / re-search area
3) 50 words (or less) characterization of the rationale
4) 50 word keyword summary of the methods used
5) Link to a summary document containing the ABSTRACT to the proposal
6) Link to a summary document containing the full detailed work already done (this should ideally NOT be wheel reinvention)
7) Link to a “comment” document with ONLY 2 sections., where peers succinctly provide their scientific views on why they support or don’t support this proposal
If this group can organize ourselves to create this summary spreadsheet, I’ll ensure that it reaches the right people, and the key stakeholders – either directly, or covertly. No one needs to get to work under any assumptions about this prior sentence, since I hope you can see that the above framework would provide a useful and alive document that actually is tolerant of as many divergent viewpoints as necessary.
And yes, I would love for WSPR and Kazakhstan to be on this spreadsheet, since both ideas result in specific “searchable” areas. The document behind Column 7 will provide sufficient space for rebuttals. This is the only way I can see this work gaining dynamic legitimacy across future time.
We should order the rows by author last name, and not by a measure of scientific accuracy, which is unfortunate, but I feel, needed.
I would urge folks to consider this effort, and I’m hopeful that what I’m proposing here is mainly organizational effort of work already done, and comments already provided. I hope folks can see that no row will exist unless the work points to a specific search area.
Best,
Adi
@Viking
The home sim data is now thought to represent take-off time for MAS flight MH150 to Jeddah, based on ATSB’s report of additional sim data in Oct_2017. The ATSB interpretation (that the sim data represents a MH150 flight) is accepted by many of us, based on fuel loads and take off time.
I will plan to document my interpretations of the sim data and implications for actual flight, meanwhile here is short summary.
https://docs.google.com/document/d/1To66Jk2jphmzRtywflZI_xJIArwgqHPSGtymxdaFZ84/edit?usp=sharing
We have Eigen vectors over here, not that I remember how to use them.
@Kenyon
Thank you very much for that.
I can go with either Crash option, but I feel the debris evidence is supportive of Option-2, which is quite similar to Jean Luc Marchand’s scenario (CAPTIO et al). They have some nice analysis and Powerpoint slides. I don’t know if I’d call it unsuccessful ditch or successful deliberate break-up for fast sinking and minimal debris.
@TBill
Great, you have obviously learnt some of the basic mathematics. That is a good starting point.
Just to clarify my main point on the simulation data. I suspect they were left to confuse the investigators. If I had done that simulation of a realistic case, I would have destroyed the external harddisk afterwards. Leaving it next to the simulator would have been a ‘present’ for the investigators, unless it was deliberately made to cause confusion.
@Viking
There have been some new sim data learnings since 2020 from ATSB guidance under agreement. We now know the recovered files are a type of temporary file that MicroSoft flight sim makes, that most users are unaware of. This implies these data are not “case files” saved by the user and later deleted, as originally thought. There is now less suggestion of false evidence planted, which in the first place was only a highly speculative secondary possibility.
It is important to note that the original (probably wrong) speculation was caused by incomplete “leak” of partial data by Malaysia. The original 2016 leak deleted a number of key data lines, whereas Line#1 is file name.type, had we known that, which we now do. But we only know that verbally from ATSB who has the complete data including the data lines that were not originally leaked. Malaysia has still not released unredacted complete sim data files, but I think some of us now understand from ATSB most of what was missing.
@All
During the last two days there has been a lot of interesting news on contrails in DK news. Many details are new, and support my satellite observations concerning MH370 (including contrail shadows, long contrail lifetime at night, etc).
All the new information is originating from a contrail conference in Copenhagen at the end of March. I think it may be worth for many of you taking a look at the homepage of the conference:
https://copenhagencontrails.org/
Unfortunately I did not know about the conference before it was over. I would certainly have participated if I had known about it.
@TBill
I am not fully updated on the latest news you mention concerning the files. However, I presume the original information that the relevant files were on an external harddisk (not on the simulator computer), and that the previous flight was shifted 12 hours are still valid.
Particularly the 12 hour shift is of paramount importance.
@Viking
I am not following your beliefs about the sim data. I personally assume the recovered sim data is candid capture of some runs the pilot did. Interpretation of the limited data is the question.
@TBill: I don’t think the Malaysians redacted the sim data files. Rather, the data was extracted separately by the Malaysians and the FBI, and the fragments recovered were not identical.
@Adi: Why don’t you take the lead in compiling the summary spreadsheet?
@TBill,
Considering my understanding of the failure mode of the recovered Flaperon hinge systems and the EOF BFO data I think a high speed descent is integral to all three (3) milestone Scenarios (and any variation in between or beyond). I agree that the term ”Unsuccessful Ditching” needs improvement, I got hung up on it myself as well, but was tired and decided to just move on.
How does a period of high speed descent, fuel exhaustion fit, and recovery into your favored Scenario 2.0 or even Scenario 3.0?
There are several contributors here that can speak to the many challenges associated with recovering a B777 from a high speed dive, exhausted fuel, and completing a one-chance ditch attempt. Critical comments from contributing pilots and those with deep understanding of the B777 flight capabilities on Victor’s blog help us stay on track with reality.
@Mick Gilbert, @Andrew, @all,
RE: Simulator Dates – Feb 2 (MH150 on Feb 4) & Feb 3 (MH370 on Feb 21)
Q1: Could the reason for NOT going ahead with the diversion to the SIO on these dates (Feb 4 & Feb 21) have been the rough sea state or bad weather conditions in the target area in the Southern Indian Ocean?
Q2: If so, does this suggest that the perp had a specific target area he was aiming for to ditch the aircraft?
Q3: Was ZS’s duty schedule for the next 5/6 weeks prepared and ready by Feb 2nd, 2014?
Q4: in the Feb 2 sim session, the fuel required for MH150 would be different from that required for MH370. Therefore, does setting the fuel to 0 and gliding from 37,000 ft, and also from 4,000 ft suggest that each time the simulation aircraft reached the target area, it had different fuel loads remaining?
I’m sorry if this information was already here, but what data do we have on the stimulation of MH370 on February 21? Is there any route data in this simulation?
@All,
Seems like the Chinese survey ship is loitering near the 7 arc again. Looks like it’s near a deep sea trench and perhaps near the location of the original discounted black box pings.
Anyone got more in-depth information on this?
@ Godfrey Jack
Victor already replied, so I really don’t have much to add. According to the CSIRO model, the discovery of debris 500+ days after the accident on Reunion Island and the East African coast is consistent with the crash location between 32°S and 40°S. According to the same study, a location around 35°S was favored because of the low probability of debris floating to Western Australia, which was consistent with the absence of any aircraft parts being found on beaches of that region.
Maybe others can correct me, but the second in line of locations of interest was roughly 37-38°S, which also showed a high probability of the flaperon going to Reunion around the time it was found, but also a higher probability of debris being close to the WA coast. The other day I heard Richard Godfrey quoting that “The absence of evidence is not the evidence of absence,” so the fact that debris was actually not found on WA beaches does not necessarily imply that it was not close to the shore. Furthermore, I am not sure of how much drift models capture the mechanics/physics of “washing ashore” or “beaching” (due to the shallow water and a sort of boundary condition between liquid and solid phases), but the fact is that actually all debris was found onshore, not offshore, and that debris can be washed back in the water (like it was the case with the Roy piece found at Mosel Bay) certainly adds to the complexity of prediction. In other words, being in the vicinity or close to the shore is a prerequisite to being washed ashore but does not necessarily imply that you will actually be washed ashore. Anyone who knows more about it, please correct me if I am wrong with this conclusion.
The bottom line is that, at least in my view, BTO/final (7th arc) BFO as an actual measurement conducted real-time has much more weight as evidence in locating the plane and drift models are there to complement it, not overrule it.
@ventus45
Thank you for your earlier comment about sonar coverage. I was not aware of Peter Foley’s presentation, but I found it in one of your previous comments. I am reposting the link you provided if this is also new to some:
https://vimeo.com/997685457?&login=true
I didn’t find the part where he specifically comments on the search in the GFC Geelvinck Fracture Zone, other than that it was challenging terrain. He does mention later that they launched AUV operations during the southern hemisphere summer to fill in missing data based on 200×200m criteria. The same information is presented in detail in the Final Report.
I was aware of the information he presented, because it is also available in the Final Report. During the search ATSB used 200x200m criteria, i.e. they were rescanning the area if the missing was larger than 200x200m, but not all. They have chosen it based on the analysis of debris field areas of previous crash sites. However, in the final report they opted to present the map using 100x100m criteria, i.e. every “data holiday” larger than 100x100m was shown on the chart. Therefore “data holidays” larger than are the subset of “data holidays” larger than 100x100m. I don’t know why they used this, using their terminology, more conservative, criteria.
During the search, ATSB used 200x200m criteria, i.e. they were rescanning the area if the missing part was larger than 200x200m, but not all. They have chosen it based on the analysis of debris field areas of previous crash sites. However, in the final report they opted to present the map using 100x100m criteria, i.e. every “data holiday” larger than 100x100m was shown on the chart. Therefore, “data holidays” are larger than the subset of “data holidays” larger than 100x100m. I don’t know why they used, using their terminology, more conservative criteria. Maybe there were debris fields smaller than 200×200m?
@Kenyon
Flight End- If it is an active pilot with intent, I suggest a whole different paradigm. I envision double or triple descent going from FL350 down to say FL150 between Arc5 and Arc6. At Arc7 the BFO apparently captured another deliberate descent, I envision to fly in the cloud layers as low as FL050, with fuel.
I interpret the sim studies, if one allows the PSS777 sim flight to keep going after 45s with the fuel loaded in the sims, you end up with fuel exhaustion at Magnetic South Pole. With APU management the aircraft makes descending circles around MagSP into the water. Therefore right now I propose the game plan might have been to end with a descending circular Hold pattern at a selected location. This implies I think flaperon might have been in upwards position acting as a aileron (but it is conjecture since MS flight sims do not show flap position, as far as I know).
As an aside, I derived from this logic the IGARI tight turn might have been a LEFT HOLD at 30 deg bank, which I could confirm in PMDG flight sim is an easy command to execute.
@Edward
My understanding is we do not have any extra data for other sim runs, except we do have initial point on KLIA runway for some runs, I believe on FSX/PMDG777. My understanding one of the runs take off time could be consistent with MH370.
@CMR
Weather is possible factor, but possibly other considerations. What I would simply say, from FBI perspective, I speculate FBI advised Malaysia that they have to assume MH150 was under serious consideration. This is quite controversial for global security and could explain the delay admitting to the sim data. What? 3.5-yrs later Oct_2017 we hear from ATSB it was MH150 flight in the sim cases. No wonder Anwar got the cool reception in Saudi. Think about it. Do you think ATSB was smart enough to make that MH150 interpretation, or was that FBI guidance?
@Victor, it has crossed my mind, and maybe I will be able to dedicate some time towards this task in the summer. These days and weeks are tough. I guess it’s easy to throw out suggestions than to actually do it oneself 🙂
Adi
@TBill: I’ve often wondered if the ATSB independently linked the sim data to MH150, or if they considered that only after reading one of my blog comments:
https://mh370.radiantphysics.com/2017/04/23/important-questions-about-most-likely-mh370-crash-site/#comment-2803
There might have been less analysis by the ATSB and the FBI than many believe.
@Adi: It is much more difficult than it seems to produce something that does not inherit the biases of the organizer.
@Thanks TBill,
@Victor, I’ll repeat my last 2 questions
Q3: Was ZS’s duty schedule for the next 5/6 weeks prepared and ready by Feb 2nd, 2014? How far in advance, in weeks, was ZS’s schedule planned?
Q4: During the simulation session on February 2, the fuel requirements for MH150 would differ from those of MH370. Thus, does setting the fuel to zero and gliding from altitudes of 37,000 feet and 4,000 feet imply that the simulation aircraft had varying fuel levels remaining upon reaching the target area each time?
@TBill, while you envision possibilities, there is ‘out of fuel‘/APU/SDU logon sequencing, BFO/ Inmarsat time stamps, and debris evidence that all need to align with EOF elevation and time constraints. Do you think flying in the clouds at 5000 ft. fits?
I have no clue what low speed or high speed descents (if any) occurred around Arc5 and Arc6. As stated above, I think a high speed descent need to exists for all scenarios.
@Victor
You have an important point concerning confirmation bias in your last contribution.
@All
After more than 11 years of unsuccessful investigations, confirmation bias is an increasing problem for finding the correct solution. It comes in many forms. One example of positive confirmation bias is WSPR. For those few among us who have tried WSPR to look for airplanes in the past, it is completely obvious that it cannot work with transmitters of a few watt over distances of several thousand kilometers. It is simply against the laws of physics, and any ‘signal’ certainly drowns in many orders of magnitude stronger noise. However, some Bloggers still believe in it, and one or two have managed to find a pattern that mimics an airplane flying towards the seventh arc along a ‘correct’ route near the correct time. Personally, I am sure it is simply noise combined with strong confirmation bias in the interpretation.
There are also many examples of negative confirmation bias. For instance, the main problem for my solution near CI is that absolutely no airplane accident investigators or other experts believe in this solution because of three main issues:
1) The mathematics linking the Inmarsat data to this solution is advanced and unusual.
2) The solution is most plausible if the person controlling the airplane attempted to parachute out near Aceh, leaving a door or hatch open, thereby systematically changing BFO values due to falling temperature in the cabin. That also explains the strange signal to noise ratios observed by INMARSAT (against the laws of physics with S/N improving with longer distance between satellite and airplane).
3) The main purpose of the long flight was hiding the airplane, so motive and methods remain unsolved by the investigators. Such an extended and dedicated effort to hide the debris and motive is completely unusual – also for previous cases of murder-suicide.
As far as I know, there are no known cases of similar nature in the entire history of commercial aviation. Typically, we solve accident investigations using experience from previous cases. There is no way to solve this particular case using these normal methods.
However, transportation by ship has a much longer history than aviation, and it contains a couple of similar cases. For most of those cases, the motives were either insurance fraught, mutiny, piracy, war, politically motivated murders or terrorism, or a mix of some of these. I guess the motive for MH370 belongs to this group, but I have no idea which one it was.
My attempts to corroborate the CI solution by contrails ending in a mushroom cloud near the 7th arc are rejected as biased, because it is not usual to use weather satellites to chase contrails. Research presented at the recent conference on contrails in Copenhagen clearly indicates that contrails often remain visible in infrared weather satellite pictures for a couple of hours at night. This also explains their large climate-impact (the main subject of the conference). I strongly recommend those with interest in contrails to look at the conference homepage.
CMR: I don’t know the answer to your first question.
Regarding the second question, the final two coordinates, 45S1 and 45S2, are in the SIO near 45S, 104E, and are separated by a distance of about 2.5 NM. Despite the short distance, the altitudes at 45S1 and 45S2 are 37,561 ft and 4,000 ft, respectively. The data set for 45S2 shows unmistakable evidence that the altitude was manually changed from 37,654 ft to 4,000 ft just before the data set was saved, which is consistent with the large change in altitude over a short distance.
@CMR
All the internal MAS stuff about flight staffing is part of a large body of evidence that we the public do not know much about. For example, I wonder if Jeddah flight time was changed (they were going to add a redeye version) is why MH150 was not taken. There has been hearsay that the pilot’s future schedule/plans was a blank sheet.
The pilot is active pilot in the sim studies. I believe the fuel load at N10 is consistent with reaching Magnetic South Pole, however the pilot then appears to use the fuel jettison valves to empty the fuel, then he then drags the aircraft down to 45S for some apparent gliding. Mick Gilbert gets credit for recognizing apparent fuel dump in the sim cases after N10.
@Victor
Yes we certainly owe you credit for recognizing flight sim Jeddah similarity.
For readers less aware, the original 2016 leaked flight sim data was missing a lot of key information (that was later made available) such take off time/date, simtime (showing the runs were indeed contiguous), file type (Line#1 of the data), flight path, etc.
I am strictly a follower of Victor’s on flight sims. Everything I do is guided by Victors work and example, although after about 4 years on the learning curve, about 2019, I got to a point I am trying to contribute new ideas.
For me, the most logical explanation for the data indicating a steep descent is the captain’s unwillingness to wait for a slow descent. I’m sure he lost height on purpose, probably in a few “rounds”. An experienced captain who knows that he plans to land on the water does not need to wait an extra 15-20 minutes, because this will not help hide the plane, unlike an hour-long flight to the southern Indian Ocean, which made sense to disappear. He probably didn’t know about handshakes.
In my opinion, the simulator data can be interpreted in support of this hypothesis.
So I mean that MH370 is 40-60 miles away from the 7th arc to the east, in my opinion.
I know this is a very difficult task, but only an extensive search (especially if MH370 is not on 33S-36S) will make the success of the OI quite likely.
At least personally, I am convinced that MH370 most likely did not fly slowly descending and did not fly 100 miles or more. This increases the likelihood that the wreckage can be found this time if all hot spots are checked and a sufficient distance from arc 7 is searched.
@Edward
Close but there is potential for greater distance after Arc7 at say 30-32s. I feel, and Peter Waring recently told me on X that Simon Hardy also feels, the true stealth portion of flight path started at Arc7. Actually maybe Arc7 reboot was a mistake so that helps a bit to see what was happening (descent).
@Adi
@Victor Iannello
I have followed this page since the crash. I have an infinity for all of the efforts for missing aircrafts. Despite not knowing the all of the technical approaches / research terms and such I would be happy to contribute by launching such a spreadsheet that @adi has proposed. Perhaps we can connect briefly off thread so that all helpful pertinent intake categories / data are included?
I still cannot get my head around the idea that someone would let the aircraft run out of fuel, let it go into the steep descent, then recover it and try landing on the sea surface. I would rather challenge the accuracy of BTO and BFO. It is a shame that investigators didn’t experimentally test the whole scenario in flight, with the B777, by turning off the left and right AC buses, powering back on after an hour, and subsequently measuring BTO and BFO (of course, without the depressurization and rapid descent part), but it is fully understandable why they haven’t done that. Maybe there are some pieces of the puzzle that were missed and that were not replicated or predicted during the investigation.
@TBill: For the hypothetical scenario in which there was fuel at the 7th arc and the plane flew for some time after crossing the arc, how would you suggest that a search area is defined?
And what is the evidence that fuel was jettisoned in the simulation rather than the fuel level was manually reduced in the same way that altitude and position changes were made?
@JP: My suggestion is that you solicit whatever information you need here. I really don’t want to have any more input than anybody else.
@Viking:
Regarding your statements on bias…
“The mathematics linking the Inmarsat data to this solution is advanced and unusual.”
I guess you could characterize it as “advanced”, but it is certainly not unusual. Actually, it is math that has been in use since Sputnik, Transit, and dozens of other satellites, but not previously applied to Inmarsat data. The BTO and BFO math is 100% solid. Stop casting doubt on the best data we have.
“The solution is most plausible if the person controlling the airplane attempted to parachute out near Aceh, leaving a door or hatch open, thereby systematically changing BFO values due to falling temperature in the cabin.”
This is pure nonsense. There is zero evidence that anything like this happened.
“That also explains the strange signal to noise ratios observed by INMARSAT (against the laws of physics with S/N improving with longer distance between satellite and airplane).”
There was absolutely nothing “strange” about the signal to noise ratios. They were all consistent with normal and expected operations for the Inmarsat links and transponder behavior. Statements like this only show your ignorance of how the the GES, transponders and AES work. Hint: There are multiple factors that determine link S/N. The distance between the s/c and the aircraft is only one of the factors, and it is not the dominant factor.
@airlandsman
Concerning the mathematics you point out yourself that it was unusual for INMARSAT data. In addition, it involves a U-turn. Those are exactly my points.
Concerning an attempted parachute jump, there is clear contrail evidence that an airplane coming from the north decended just before Aceh and continued below a low-hanging hazy cloud cover (projecting an infrared contrail shadow up on it). The general pattern fits with Kate Tee’s eye-withness report except for the time which is roughly one gybe off. However, the time and position fit with her GPS log to great precision. I trust that more than her memory of the exact time and position.
The main problem with the S/N ratio for the signals is that it shows a steady increase over the remaining part of the flight (except at 7th arc, where lots of other things happen). A gradual decrease would be expected. This deviation is so systematic, that it raises a red flag. I agree with you, in case only one point was off, but here we see a completely unusual systematic behavior.
@Viking:
I wrote: “…it is certainly not unusual….”, the opposite of what you attribute to me. Again, the BTO and BFO data are rock solid and the analysis thereof is also solid…verified by many officials and independent investigators.
Re: “The main problem with the S/N ratio for the signals is that it shows a steady increase over the remaining part of the flight …”.
This is NOT a problem. As previously stated, S/N is a function of several factors, not just the A/C to S/C distance. The inbound C/N as recorded at the Perth GES is also a function of the instantaneous AES Power output, AES antenna gain in the direction of the spacecraft, spacecraft L band antenna gain in the direction of the aircraft, transponder gain (which can vary due to other traffic at the time), C band downlink path loss, etc. All these factors cause more variation in C/N than the 1 dB change in AES to S/C path loss due to increasing distance from ARC 1 to ARC 7. This has been extensively covered. For example, see the following: https://bit.ly/3413OX9
@Victor
Search area needs work. I propose a fairly well defined problem but lots of variables: Time and extent of descent1 after ~Arc5? descent2 at ~Arc7? Heading? which can substantially shorten distance to Arc6, etc. Basic question what is max possible distance after Arc7? There is fuel at Arc7 but low altitude limits fuel efficiency: basically taking descent credit before Arc7, but also the upward northerly swing of the Arcs gives potential for greater dist from Arc7, ruled out previously in part due to the “fuel not-exhausted at Arc7” argument. Perhaps less focus on a hot spot (I would consult with Bob Ballard). In my tests, up to 125-150nm from Arc7 seems possible, but I am not sure if actual case could get out that far, or further.
Re: Sim Cases/Fuel Jettison Inference at N10
It takes about 45-mins to jettison the fuel at the N10 fuel load. This turns out to be consistent with simtime elapsed until the icon drag to 45s. Simtime not public info but verbally confirmed by ATSB to researchers (Mick Gilbert) to be in the range of 45mins. I have a 2021 paper on some of this, because I felt disclosure was important for update:
MH370: New Interpretations of the Pilot’s Home Flight Sim Studies
https://www.mh370search.com/2021/02/04/guest-paper-by-bill-tracy/
or here:
https://docs.google.com/document/d/1cpaWvW861hpKoy8IHK3NACuw-NyeW9qIIQ6S4Fd8GOA/edit?usp=sharing
@TBill: It’s those “lots of variables” that will make it very hard to define a manageable search area without introducing bias, i.e., guessing.
I am not persuaded by the case for jettisoned fuel when it is trivial to change the fuel quantity as part of the “jumps” between positions, although we can’t definitively rule it out.
At this point in the search results and analysis to date, it seems to me the options from here are really very limited.
1. Unsearched data holes – but are there any significant holes remaining in high probability zones?
2. Further out from the arc – requires assumption of lengthy glide. Not impossible, but contraindicated by multiple aspects of the evidence – and also implies an unfeasibly large search area
3. Close to 7th arc but further north than searched (N of 27S?). Highly improbable: requires a contrived path model and involves very large BFO errors
4. Close to the 7th arc but further south than searched (39.5-40.0S).
As many here are aware, I favour #4 because this end-zone is predicted by parsimonious (ie minimal manoeuvre) LNAV path models that best-fit the BTO data. The relevant search area is tightly-constrained. I would certainly advocate #5 before embarking on #2 (if the latter were under serious consideration).
correction to my last sentence:
I would certainly advocate *#4* [39.5-40.0S] before embarking on *#2* [further out from arc]
@paul smithson wrote:
“3. Close to 7th arc but further north than searched (N of 27S?).”
“Highly improbable: requires a contrived path model and involves very large BFO errors”
Your four options for future exploration are strictly limited to BTO/BFO evidence from 2014. Newer evidence needs to be taken into account.
The Java Anomaly at latitude 8.32S on the 7th Arc is an undisputed epicenter not just close to, but directly on the 7th Arc within the measurable accuracy of the arc and the epicenter.
The path to that point may seem “contrived” when focusing only on the SATCOM evidence. The proposed waypoint path to the JA site is an exact match for each BTO timing. BFO was not factored in, as it was considered less reliable. The CAPTIO group found a path up to Cocos Island airport that was a very good match for BFO. Rather than “very large BFO errors”, there are multiple paths that could exactly match BFO and BTO from Cocos to the 7th Arc Java Anomaly. These paths might deviate from the simplistic low and slow constant altitude flight proposed, which when assessed had a low BFO error better than many 7th Arc candidate sites.
The contrivance may be trying to fit BTO+BFO into a narrow assumption of a flight to oblivion.
Acoustic evidence from the II.COCO airport seismometer is consistent with a flyby crossing the atoll from the NNW and departing to the ENE, at the exact timing between the 4th and 5th arcs. An additional doppler shifted detection from Christmas Island seismometers is consistent with a sixth arc timing flyby of the airport there, but it was clouded over at the time without IFR, so not place to safely land a B777.
I acknowledge that your assessment of “Highly improbable: requires a contrived path model and involves very large BFO errors” fits a consensus path based on BTO+BFO, but it omits newer acoustic epicenter and barnacle evidence.
@370Location
How did you come to the conclusion that Christmas Island “was clouded over at the time without IFR, so not place to safely land a B777”?
On X, @TBill said: See what I am saying. we know where the aircraft probably went. IG saying not interested, they want a reasonable spot for OI to certify equip, preferably where MH370 is not located. Don’t want no controversy, just a search spot.
Is this what you really believe?
@Andrew:
The METEOSAT7 images are not high resolution, but we have them every half hour in three bands = VIS, WaterVapor and IR:
https://drive.google.com/drive/folders/1R_Kokhpe5nY_rgYp7GKmAGpciPtZBRd8
I plot the YPXM flyby time interpolated between arcs as around 2352Z, with dawn at around 2330Z judging by the sat images.
The first higher resolution color MODIS Terra image is four hours later at 0356Z showing broken clouds.
I recall reading that Christmas Island YPXM has the most diversions of any AU airport, with low visibility as often as 25% of the time due to its elevation above sea level, and presumably worse in the morning. I inferred that flights take off with a forecast showing that the clouds should clear by arrival time, but sometimes don’t. If there were a glide slope, I expect commercial flights and ambulance service would land IFR.
Odd that the airport lacks IFR, as there was once a proposal to make it a space port.
You’re right though, I can’t be sure that YPXM was socked in. Perhaps someone could pull up more detailed weather for that morning.
@370Location
YPXM metar data is apparently available – but at a price
https://metar-taf.com/archive/metar/YPXM/2014?station_id=WIHH&__cf_chl_tk=iEAVew0pUpXag3egXzPkLZ_0Bg9LJOan.MWET5k66Wc-1745657888-1.0.1.1-YLR7Pdk34hQyiIIGQErTtPyTV5ysPOg2jJWJFkdabvM
@370Location
These may be of interest.
http://www.bom.gov.au/aviation/data/education/ypxm.pdf
http://www.bom.gov.au/aviation/data/education/ypcc.pdf
@370Location
Try this one:
https://en.tutiempo.net/records/ypxm/8-march-2014.html
@370Location
The YPXM METARs show there were layers of scattered cloud at 1,800 ft and 2,600 ft above the aerodrome at 0000Z. The cloud was reported as broken an hour later. The visibility at the time was 10 km or more.
YPXM suffers the vagaries of maritime weather, like many small island airports. The airport doesn’t have an ILS approach, but it does have instrument approaches that will get aircraft down to 393 ft AAL on RWY 36, with a minimum visibility of 2,200 m. The RWY 18 approach has minima of 418 ft AAL/2,200 m. I dare say that significant upgrades would have been required if the spaceport proposal had gone ahead.
@370Location
This is turning into an interesting hunt !
This one has graphics.
Pretty good.
https://weatherspark.com/h/d/149132/2014/3/8/Historical-Weather-on-Saturday-March-8-2014-at-Christmas-Island-Aerodrome-Christmas-Island#Figures-Temperature
@370Location
The link I posted earlier doesn’t seem to work. Try this:
https://www.dropbox.com/scl/fi/bsb9a4iimptkce7oru36c/IMG_2669.jpg?rlkey=9ib6t0vp8oz492bwniuap3oqo&st=1iaazj0i&dl=0
@370Location
Lots of MH370 weather data
https://www.weathergraphics.com/malaysia/index.shtml
@370Location: To be clear, Christmas Island YPXM has both RNAV and VOR instrument approach procedures (IAPs) for runways 18 and 36. A scattered layer at 1800 ft does not require an instrument approach, and a broken layer at 1800 ft would have posed no problem, i.e., weather was not a factor.
@370Location
@Victor
After a lot of digging, I have found a site with WMO coded SYNOPTIC METAR and RADIOSONDE data.
https://www.ogimet.com/synops.phtml.en
I selected three days for both Christmas Island and Cocos Island, the 7th 8th and 9th March 2014
https://www.ogimet.com/display_synops2.php?lang=en&lugar=96995+96996&tipo=ALL&ord=REV&nil=SI&fmt=txt&ano=2014&mes=03&day=07&hora=00&anof=2014&mesf=03&dayf=09&horaf=23&send=send
The METAR and TAF data are available at:
https://www.ogimet.com/metars.phtml.en
https://www.ogimet.com/display_metars2.php?lang=en&lugar=YPCC+YPXM&tipo=ALL&ord=REV&nil=SI&fmt=html&ano=2014&mes=03&day=07&hora=00&anof=2014&mesf=03&dayf=09&horaf=23&minf=59&send=send
Radiosonde data is only available for Cocos Island
https://www.ogimet.com/sond.phtml.en
https://www.ogimet.com/display_sond.php?lang=en&lugar=96995+96996&tipo=ALL&ord=DIR&nil=SI&fmt=txt&ano=2014&mes=03&day=07&hora=00&anof=2014&mesf=03&dayf=09&horaf=23&send=send
@370Location
@Victor
I posted a chart and links to the WMO encoded data in three PDF’s here.
https://x.com/Ventus_45/status/1916101673610838093
@Andrew, @ventus45, @Victor, @Viking:
Thanks for all the WX info. I suppose the airport elevation of near 1000 ft gets subtracted to determine minimums. I don’t know that it’s significant, because we know that the plane didn’t land there, and may not have had even NAV ability.
We only have a weak doppler shifted seismic infrasound detection at the right time.
There was a transition at the time to a newer Christmas Island coastal radar installation facing Java to spot immigrant boats, so there may have been two in operation, reportedly derived from navigational (ship?) radar gear. My attempt at a request did get a response saying that any data would be military and classified. Perhaps an FOIA might get a better result.
Speaking of weather maps and infrared, I was once curious to see what might have been visible on a MODIS Aqua infrared satellite capture at 1910Z as MH370 was approaching the 2nd Arc. It would be just after BEDAX, depending on speed. I have MH370 flying slower, but there is an interesting short streak just due south of my estimate, right on the BEDAX-ISBX path. Here’s a rough map, contrast enhanced:
https://drive.google.com/open?id=15OOjh7F36KgNxNYYHsv53uaNMnpiwnPl
Also a link to the source, zoomed way in to pixel level to pinpoint the possible contrail endpoint:
https://go.nasa.gov/42vWd48
I don’t see any other likely contrails in the area, but I believe @Viking does. The SSE angle seems slightly off for a flight due south, and the upper winds would have blown the tail west. Still, it’s an interesting coincidence. Getting the raw imagery might give better contrast if it’s worth pursuing.
@370Location
RE: ”I suppose the airport elevation of near 1000 ft gets subtracted to determine minimums.”
No. Cloud height is reported as hundreds of feet above the airfield elevation. There is no need to subtract the elevation to determine the height of the cloud above the airfield.
@Andrew, @Victor,
Thanks for the corrections and info (including how altimeters report).
Your expertise here is much appreciated.
@370Location: It’s a bit confusing. Altitudes in weather reports and forecasts associated with aerodromes, e.g., METAR, TAF, ATIS, AWOS, ASOS, are in AGL. Altitudes in other weather reports and forecasts are in MSL. Because the plane’s altimeter indicates MSL altitudes, pilots aviate and communicate using MSL altitudes. Instrument approach plates typically have decision altitudes (precision approaches) and minimum descent altitudes (non-precision approaches) listed in both MSL and AGL altitudes so the ceilings reported in the aerodrome weather reports can be directly compared to the minimum AGL altitudes for a given approach, and the pilot can “bug” and descend to the minimums using the MSL altitude on the plate.
In any event, weather was not a factor for a landing at YPXM. A ceiling of 1200 ft AGL would pose no challenge for any proficient, instrument-rated pilot.
@paul smithson
After the no.1 I am for no.4 option definitely.
Regarding your question related to option no.1, I can try to partly answer it. There are still areas with significantly large data gaps and OI searched them in first phase of the current search around IG’s hotspot. The thing which is still puzzling is why they in the Phase 2 continued to search further from the Arc and skipped (for now) the data gaps around 35-36°S. If you zoom in the image at the beginning of this post you will see there are areas marked dark green and outlined in white color, which seem quite large in size so they can easily fit the debris field.
The ATSB’s final report, “The Operational Search for MH370” on page 96, specifically states that:
“It is worth noting that filtering out all data gaps and LPD areas less than 200 m by 200 m decreases the individual data gap count by 78 per cent, and LPD count by 60 per cent. Using this same metric but filtering by square kilometres reduces the total square kilometre area by 24 per cent and seven per cent respectively. This highlights that many of the data gaps and LPDs are smaller than 200 m by 200 m.
Historical analysis has shown that aircraft debris fields typically cover areas larger than 200 m by 200 m. Therefore, areas less than 200 m by 200 m can be discounted with a moderate to high degree of confidence as they are not large enough to fully contain the aircraft debris field.”
LPD stands for Lower Probability Detection area.
What I can read from this is that data gaps and LPDs larger than 200x200m account for the majority in terms of area but not in terms of number. This is all related to the total area searched by ATSB, 120 thousand square kilometers. If I am not interpreting their words in the wrong way, it means that data gaps which account for 0.5% of total area searched (600km^2) and LPDs for 2.1% (2520km^2), then the all data gaps larger than 200x200m have a total area of are about 456km^2 and LPDs of about 2344km^2. I don’t know the total number of those “larger than 200x200m” areas or where are they located, os it is hard to answer your question completely.
NoK Jacinto Goncalves has passed away.
https://www.thevibes.com/articles/news/107573/jacquita-gonzales-widow-of-mh370-crew-member-dies-at-62
@CMR: That’s really a shame. She was a strong advocate for the MH370 families. She was well-liked and respected. Rest in peace.
@CanisMagnusRufus Heartbreaking news, prayers to their family. Judging from many articles over the years, she seemed like a very strong person, and like so many MH370 survivors, endured more than we can fully appreciate and understand.
@370Location – As a reader, very interested in your contrail observation but definitely not easy to discern for the ordinary eye.
Pasting link here to @Viking’s paper that details some of his observed contrails. Here again don’t have the expertise to comment but do think others here might be able to provide valuable inputs on what they observe.
https://arxiv.org/pdf/1811.09315v3
Fuad Sharuji was crisis manager at MAS in 2014 when both the MH370 and MH17 tragedies happened. He has been interviewed many times by media about both events.
At the beginning of the 2nd episode of Netflix documentary MH370: the Plane that Disappearedy the episode which deals with MH17, he makes the following statement:
I just went numb. Completely numb. And I just couldn’t believe it.
We kept thinking, “There is probably some kind of plan to destroy our credibility.”
A series of actions trying to bring us down.
I’ve highlighted the personal pronouns to emphasise that this is not just the personal opinion of the speaker.
This is a serious admission from someone at the highest ranks of MAS that, in July 2014, despite the captain’s simulator sessions, MAS leadership believed that MH370 had disappeared not at the captain’s hands, but due to a third party with malevolent intentions—both against MAS and possibly Malaysia’s international reputation.
Armada 78 06 is back in Singapore.
@VictorI et. al.
Journalist Caro Meldrum-Hannah of Four Corners (ABC Australia) did an interview with Anwar Ibrahim back in 2014 as part of a documentary Lost:MH370.
http://mpegmedia.abc.net.au/news/fourcorners/video/20140519_4c_ibrahim_iv_288p.mp4
At 14:50 she asked this question.
CMH – “What we do know is that Air Traffic Control in Vietnam made multiple attempts to contact Civil Aviation in Malaysia, and apparently Civil Aviation didn’t respond…”
AI – “Yes, there was a complaint launched by Vietnamese authorities, and to an extent that at a later stage they even threatened to cease all operations because of the failure of the Malaysian authorities to respond.”
I don’t recall seeing this ‘tense’ exchange between HCM ATC and DCA Malaysia being mentioned in any of the reports that were released. Are the interlocutors mistakenly conflating DCA Malaysia with KL ATC, or does this refer to a distinct incident that was omitted from the reports?
I agree with the summary from @John Matheson on WSPR
My analogy would be an old analogue tv with no signal when displaying random white noise or “snow”. Consider one was convinced that when a plane flys overhead an image of a plane is hidden in the noise. And sure enough if look hard enough, join the “random” dots in a specific manner – and bingo – suddenly the shape of a jet shape appears – thus “proving” your scientific theory.
“So far proponents of WSPR don’t seem to be able to provide examples of tracked aircraft where the aircraft track was not already known. There’s plenty of examples of “pattern matching” WSPR anomalies to known tracks. It is unsurprising that out of zillions of anomalies in HF propagation that some can be hand selected to “confirm” a known track.”
Im interested to dig into the nuance around the “deleted” sim data. Some media reports that “only” the suspicious flight data was deliberately deleted. I want to see if this is 100% accurate.
After some quick searching, i found @Victor Iannello blog and some other information – https://mh370.radiantphysics.com/2017/10/12/simulator-data-from-computer-of-mh370-captain-part-1/
Heres what i understand
2 different flight sim programs on captain computer
FSX sim on MK26 (main) drive
FS9 (FS2004) sim on MK25 drive (external)
Suspicious flight sim session recorded on feb 2 2014
Session data forensically retrieved from windows shadow copy volume
Shadow copy snapshot date feb 3
Flight sim session had similarities to MH150 2 days later.
Feb 20 FS9 uninstalled from MK25
What is the specific evidence of the delete ? was it simply from the user FS9 uninstall ? Or some user manual deletion. Was anything else deleted ?
I don’t have access to the leaked RMP reports
@Marijan,
The Armada 78 fleet is continuously ‘on the move’, repositioning between projects, logistics bases, and shipyards where additional equipment is mobilised on the vessels.
At present, six of the eight 78m vessels are in European waters. Armada 78 01 has undergone, what I suspect is a temporary, addition of an over-the-side A-frame to operate a modified Schilling ROV. A78-01 then positioned to the Norwegian North Sea to conduct a project in the Breidablikk field in which it’s likely to be delivering an innovative approach to reservoir monitoring with seafloor seismic technology.
A78-08, claimed in error by some commentators as the vessel intended to conduct the recent MH370 search, also recently underwent additional modifications to extend its capabilities. After its voyage from SE Asia it entered Damen’s yard in Amsterdam for fitment of rigging for towed seismic arrays as has previously been carried out to A78-03 and A78-05. After departing Amsterdam -08 conducted two short operations off Brittany, NW France, before conducting ROV trials off NW Spain. Notably, the Blas de Lezo, a Spanish Navy Frigate held station in close proximity to A78-08 as it operated on site. The Spanish Navy and Ocean Infinity are both operators of Saab Seaeye ROVs.
@Don Thompson
Thank you for your extensive reply. I thought there was a glimmer of hope that Armada 78 06 might return to the search area (a slim chance of returning before winter was mentioned on the same day 7806 departed the search area for Singapore) while transiting to another location, but now it seems it is heading straight to Mauritius.
@Maverick
In more recent years, we now understand (from ATSB), who have verbally provided additional sim data, that the home simulator data (to the SIO) were a type of temporary file created by the MicroSoft flight sim software. Therefore it now looks more like the runs were not intentionally deleted, rather it looks like it was a type of temp file the flight that the software itself creates and deletes (for the purpose of resuming a flight after certain user changes are made like fuel loading and other times when the user pauses a run). I suppose I can only say my personal opinion now is that the runs are completely candid capture of a certain flight.
ICAO’s council has found Russia responsible for the shootdown of MH17.
“In reaching its decision, the ICAO Council has upheld the fundamental principle that weapons should not be used against civil aircraft.”
https://www.foreignminister.gov.au/minister/penny-wong/media-release/russia-responsible-downing-flight-mh17
What if weapons were not used, such as in MH370? No fundamental principle for that?
Meanwhile, Malaysian PM AI is set to visit Russia over the next few days, his second visit as PM.
https://www.freemalaysiatoday.com/category/nation/2025/05/12/anwar-set-for-second-visit-to-russia
What could be the reason for this fawning over Russia?
“Anwar has become “cynical over the West’s harsh criticism of Russia but its muted response to Israel’s military actions in Gaza.” So cynical, in fact, that he felt visiting Russia and praising Putin would be a finger in the West’s eye.”
https://thediplomat.com/2024/10/anwar-ibrahim-in-russia-opportunism-or-hatred-of-the-west/
@CanisMagnusRufus
You asked: “What if weapons were not used, such as in MH370? No fundamental principle for that?”
Of course there is.
The fundamental principle behind the ICAO Council’s finding that the Russian Federation failed to uphold its obligations under international law is Article 3 bis of the Convention on International Civil Aviation (Chicago Convention). Article 3 bis states:
The contracting States recognize that every State must refrain from resorting to the use of weapons against civil aircraft in flight…
The fundamental principles that prohibit other forms of interference against civil aircraft may be found in other UN Conventions.
For example, Article 1 of the Convention for the Suppression of Unlawful Acts against the Safety of Civil Aviation (Montreal Convention 1971) states:
Any person commits an offence if he unlawfully and intentionally:
(a) performs an act of violence against a person on board an aircraft in flight if that act is likely to endanger the safety of that aircraft; or
(b) destroys an aircraft in service or causes damage to such an aircraft which renders it incapable of flight or which is likely to endanger its safety in flight; or
(c) places or causes to be placed on an aircraft in service, by any means whatsoever, a device or substance which is likely to destroy that aircraft, or to cause damage to it which renders it incapable of flight, or to cause damage to it which is likely to endanger its safety in flight…
Article 1 of the Convention for the Suppression of Unlawful Seizure of Aircraft (Hague Convention 1970) states:
Any person who on board an aircraft in flight:
(a) unlawfully, by force or threat thereof, or by any other form of intimidation, seizes, or exercises control of, that aircraft, or attempts to perform any such act…commits an offence…
The following article provides some background to the ICAO Council’s finding on MH17:
The Conversation: Does the UN aviation body have the power to punish Russia for the MH17 downing? An aviation law expert explains
@CMR
If Malaysia pokes USA in the eye too hard, maybe USA will take a more proactive stance towards releasing MH370 secrets that the FBI are holding as confidential for diplomatic reasons.
A new BBC TV series dramatizes the Scots-US investigation into the Pan Am 103 bombing over Lockerbie, Scotland, the effect it had on victims’ families and how it impacted Lockerbie’s locals. The lessons learnt went on to improve how victims’ families were treated in the aftermath of 9/11, and lead to adoption of good practices in the response system for victims’ families of aircraft disasters.
MAS seems to have incorporated some of these valuable lessons into its action plans for crises. By April 2012, the airline had signed Memorandums of Understanding (MOUs) with at least three NGOs, two of which focused on providing emotional support for staff, crew families, and passengers in the event of an ‘unprecedented air crisis disaster’.
https://www.perkamainternational.org.my/wp-content/uploads/2019/07/Article-1_April-2012.pdf
What’s interesting is the timing of the signing of these MOU’s. They appear to coincide with the Olympic games.
2007 – MoU signed with Tzu Chi (Taiwan), for Mandarin speaking volunteers
2008 – Beijing Olympics
2012 (April) – MoU signed with Malaysian International Counselling Association (PERKAMA Intl)
2012 (July) – London Olympics
Due to hightened threat levels, both Beijing 2008 and London 2012 Olympics took precautionary measures to counter 9/11 style hijacked airlines crashing into Olympic venues. These MoU’s with NGO’s may have been signed by MAS in anticipation of potential air disasters that may happen during these Olympic Games.
In the case of MH370 in 2014, Tzu Chi volunteers were called up and sent to KL & Beijing to provide counselling to NoK. It appears MAS promptly implemented an already established action plan involving Tzu Chi, a plan perhaps originally meant for the 2008 Beijing Olympics.
MAS crisis manager Fuad Sharuji spoke about his experience handling the MH370 & MH17 crises, and the role of Tzu Chi in helping MAS in 2014.
https://youtu.be/cXfAabiO7Ys?si=uCemscI6CvsWkaEC
I’m curious about the reliability of BFO data that led investigators to conclude MH370 was in a steep dive during the last handshake.
I had a read of the great document – MH370 Burst Frequency Offset Analysis and
Implications on Descent Rate at End-of-Flight.
To me, it seems the BTO ping data and measurement of 7th arc is very reliable – I’d guess above 99% confidence.
However the BFO data I’m less confident that the ONLY explanation is a step dive. The weird BFO at 18:25Z – was it a turn ? a dive ? was it startup drift ? bad data – a combination ? Or some other factors. Inmarsat initially thought the weird BFO after take off was “possible turn”. Perhaps a turn near end of flight turn could affect BFO – not only a dive.
To me it seems just so many variables affect the BFO and the chance of erroneous data is high. For end of flight – fuel exhaustion, satcom restart, then final dive – are probably the most likely sequence of events – however the confidence in the steep dive based only on 00:19Z BFO – is more like 50% confident that its correct.
Also regarding the satcom “hack spoof BFO” theory. If you are that smart – why not simply just disable the satcom terminal ? In my view – if the satcom terminal was disabled – then washed up wreckage and drift analysis would be the only clue to its location.
Maverick: Relatively speaking, you are correct to place more confidence (certainty) in the BTO data than the BFO data. From the BTO data, we know the location of the 7th arc to ~ 5nm. However, the BFO data pattern circa 18:25 has been well understood for years now. We know with high confidence the BFO time series circa 18:25 was the result of a cold start of the OCXO, matching identically the cold start behavior earlier in the day on the ground (12:50?). Thus, we should be highly confident that after 3-4 minutes, the BFO offset calibration was back to what it was before the power outage…give or take 1-2 Hz. At 00:19, the OCXO was NOT cold when the AES restarted. It was already at the set point temperature, perhaps 1-2C below. Thus, the BFO offset might have been off 10-20 Hz at most, but the “signal” was 10 times that uncertainty. Thus, we can be highly confident in the conclusion that a rapid descent was underway way at 00:19:37.
thanks for the info @airlandseaman – what you explain makes sense – and the puzzle piece seems to fit.
is there any information on how a turn or roll affects BFO ? I’m drawn to the very early guess from inmarsat that an unusual BFO could mean a turn.
Maverick: These 3 charts show the effect of a 360 degree turn at constant altitude on BFO values and the effect of vertical motion on the BFO values.
https://bit.ly/43I66fw
@airlandseaman: Those graphs look good, but really should be labeled with track rather than heading.
@Maverick
In addition, in years past IG members and others such as DrBobby Ulich developed handy Excel spreadsheets that can be used to calculate MH370 BTO/BFO as a function of time, date, location, heading, speed and descent/ascent. I use DrB’s Excel, but I am not sure if it is still posted.
One of the most interesting BFO changes is the first unanswered SATphone call at 1840 which implies a definite maneuver (turn south or descent or both) was made before 1840, but after 1825 when the SATCOM was rebooted. At the end of the flight, when the aircraft is further from the satellite, I believe BFO value is less impacted by heading changes.
The 1840 BFO explains why the ATSB/DSTG original Bayesian hotspot was at 38South, when prioritizing maneuverless straight flight after that.
Victor: Thanks for catching the labeling errors. Corrected version here: https://bit.ly/4kkLy2z
@Maverick, airlandsman, VI, TBill
Sorry for my delayed input to your discussions. I was busy with other things.
It is great reading that others are beginning to address possible problems with the BFO values. I think you discussed most of the ‘standard’ options for problems above very nicely. I also agree with your probability estimates, but I think the BTO values are better than your estimates.
However, I think you are still overlooking the worst possible systematic option for BFO problems, namely a temperature drop in the cabin. The response factor (from oscillator drift) is 0.3Hz/K. It is totally unimportant under standard conditions, but may become dominant if the cabin temperature falls to outdoor values at typical flying height.
Viking: Where are getting this info:
1. “The response factor (from oscillator drift) is 0.3Hz/K.”?
2. “…temperature drop in the cabin”?
Unless a OCXO is broken, it will maintain the crystal and oscillator circuit at a constant temperature (~ +70C), independent of the ambient temperature. The change in frequency dues toi changes in ambient temp should be minimal.
Besides, there is no evidence the cabin temperature dropped to OAT. Maybe for a short period when the cabin was depressurized, it dropped some, but it is very likely the cabin was repressurized circa 18:25 UTC.
Regardless of the long term drift (if any; very unlikely), the very rapid change at 00:19 could only be reasonably explained by a rapid descent.
@airlandsman
The drift factor comes from INMARSAT. Several contributors downloded it from their homepage. I do not know if the value is still there, since they may since 2014 have changed the temperature stabilization to a new and better model.
I do agree that most of the change at 00:19 was not due to temperature changes, but I do not rule out that up to 30% might be from temperature changes.
The elephant in the room is if the temperature did indeed fall significantly during the last hours of the flight. There are two indications of a possible temperature drop:
1) The unphysical increase of the signal to noise ratio with increasing distance from the satellite. For those who think that increasing distance is insignificant, please note that the simple diffraction contribution is small. However, there are two other and much larger contributions. (a) The effective area projection of the airplane antenna. (b) The polarization projection factor. Both are 3-8 times larger than the simple diffraction contribution. All 3 contributions have the same sign.
2) Issues related to differences in the statistical distributions of BFO values during the two attempted phone calls.
For semiconductor-based oscillators and amplifiers (microwave as well as optical) there is a general trend of increasing signal to noise factor increase with decreasing temperature. The exact factors depend on design aspects and the quality of the temperature stabilization.
Using typical values for microwave equipment from the 1980ies (where the SDU box was designed and produced) I find agreement with the observed values for MH370. That means a temperature drop may explain the increasing S/N with distance.
@Viking
Based on my practical experience with flight sims/flight paths, I do not see any problems with MH370 BTO/BFO. Just a little drift in the BFO, which is the expected behavior…4 maybe 5 unit drift in BFO during the flight, so maybe an actual 248 BFO reads as 252 BFO in the official data base, at the very end of the flight. The flight paths which directionally go approx. 180 south show good matches. Unfortunately the good BFO/BTO “match” still leaves an enormous potential search area, lets say 30-38s up to potential 200nm distance from Arc7. I am pretty sure MH370 is within those vast limits, although outside chance for even greater distance from Arc7 when we talk about 30s region (due to shorter flight path to Arc7 due to the way the Arcs curve).
Viking:
First, regarding your statement:
“The response factor (from oscillator drift) is 0.3Hz/K.”
Perhaps you are confusing the OCXO change in frequency vs. ambient temperature with the carrier frequency change with ambient temperature. The average 10 MHz OCXO temco from -30 to +50C is about 0.002Hz/C. At the carrier frequency (~1.6 GHz), this results in a change from +50C to -30C of about -20 to -25 Hz. See actual test data here: https://bit.ly/2VAL4cJ. Thus, any change due to temperature could not have caused a large change in BFO offset.
Regardless of the ambient temperature during the the flight south, the 254 Hz change in BFO between 00:11 and 00:19:37 had absolutely nothing to do with ambient temperature changes. Even if the BFO calibration “drifted down” 20 Hz, that would have happened early in the flight, not between 00:11 and 00:19. The change over that 8 minute period could only be explained by the high descent rate. Temperature had nothing to do with it.
Second, regarding this and related statements about S/N:
“The unphysical increase of the signal to noise ratio with increasing distance from the satellite. For those who think that increasing distance is insignificant, please note that the simple diffraction contribution is small. However, there are two other and much larger contributions. (a) The effective area projection of the airplane antenna. (b) The polarization projection factor. Both are 3-8 times larger than the simple diffraction contribution. All 3 contributions have the same sign.”
Sorry to be blunt, but you clearly have no idea what you are talking about. As previously explained (many times over the years), the change in path loss due to changes in distance over the whole flight was only about 1 dB. Meanwhile, changes in S/N can be caused by several factors other than path loss, up 10 dB or more (instantaneous antenna gain, transponder loading, etc.) Thus, you can’t infer anything about cabin temperature from the S/N values.
Lastly, regarding your statement:
“For semiconductor-based oscillators and amplifiers (microwave as well as optical) there is a general trend of increasing signal to noise factor increase with decreasing temperature.”
You are clearly conflating oscillator and amplifier noise with the noise in the combined inbound up/dn link S/N. In the case of the Inmarsat links, AES oscillator and amplifier noise contributes nothing to the link S/N. The link noise is dominated by thermal noise in the s/c transponder antenna system and the Perth GES antenna system. Noise from all the AES components combined has zero measurable effect on the link S/N reported in the Inmarsat data.
@Blain Allan Gibson, VictorI, DonT, MikeX, et al…
Blain was interviewed by Geoffrey Thomas on this video of Airline News.
https://www.youtube.com/watch?v=gipBaQTAco4
After watching this video, I hang my head in shame and embarrassment for continuing to harp on about the “hijacking” of MH370 by third parties for so many years! I read too much into what Jeff Wise and Florence de Changy wrote and became convinced of the hijacking theory.
I offer my sincere apologies to the families of the 2 Ukrainians on board MH370 for having cast aspersions on them.
I offer my sincere apologies to Victor, Don Thompson, Mike Exner, and Bill Tracy for taking up sooo much space on this precious blog with what MikeX would describe as “horsesh1t”
But … why couldn’t the photos of the 2 Ukrainians boarding the plane have been published earlier?
Why did Florence de Changy get away with making such outrageous claims about them for so long and no one thought to correct her?
Thank you Blain Gibson for clearing up this particular matter. But the main question still remains unanswered: who flew MH370 to the SIO?
@airlandsman
Out of your many points, I only agree with one. Since I wrote my contribution too quickly, I forgot to put decimals on the exact times. That made the remark on an up to 30 % contribution from thermal effects extremely non-pedagogical. I am sorry for that mistake.
To phrase things correctly, I was only talking about the time interval from 00:11:00 (decimals not needed, but I add them anyway to avoid further confusion) until 00:19:29. During that period flight conditions were still under control (by a pilot or more likely the autopilot), and the total BFO change was 70 Hz out of which at least 5 Hz was expected by extrapolation. That means that my 30 % maximum estimate equals 19.5 Hz. I stand with that estimate. Most likely the number is somewhat smaller, but not insignificant.
Thermal effects have no influence on the short dramatic interval from 00:19:29 until 00:19:37.
Concerning the 0.3 Hz/K, I could not quickly find the originally published number. However, it is close to expectations for equipment developed and produced at the time of the SDU in MH370. It corresponds to a relative uncertainty in frequency of 2*10**(-10). I was involved in high precision spectroscopy in those days and that number would place INMARSAT among the best commercial products in the market.
The numbers you quote (relatively around 10**(-11)) fit much better with modern technology today. Most of the improvements are achieved by double-stabilization methods. Around 1986 we only had prototypes of such systems, and they were not sufficiently reliable to install on an airplane. Precision was fine, but they often lost track and decayed to the quality of single stabilization or worse.
Concerning the S/N, most of the increase with falling temperature comes from increased power output. That is the reason that I sometimes only discussed power. However, the noise is also falling. Indeed, larger changes can occur for shorter periods, but they always go towards worse S/N. Examples are transponder load (as you mention). However, that is typically at its lowest at night and gets worse during the morning (causing worse S/N). Another example is turbulence. That also gets worse, particularly towards weather fronts in the arctic or morning-triggered thunderstorms (for the two solutions respectively). A third example is the tail of the airplane getting closer to the direction of the satellite for large distances. That also gets worse towards the end of the flight.
However, INMARSAT data tells us that S/N is systematically growing. That is against the laws of physics.
@Viking
“However, INMARSAT data tells us that S/N is systematically growing. That is against the laws of physics.”
Of course it is…
In the 2007 “AMS(R)S Part III-ICAO DRAFT SARPS FOR AMSS” is the requirement for the GES to control the output transmit power of the SDU.
“4.2.3.5.2 EIRP control. The EIRP per carrier in the direction of the wanted satellite shall be adjustable over a range of 15 dB in steps of 1 dB by command from the GES.”
Later the document “icao_doc_9925_-_manual_on_amss-route_-_1st_edition_-_2011.pdf” does not explicitly mention that the GES controls the SDU transmit power.
However footnote 11 to Table 4-1 “Principal antenna performance requirements” (pg 148) States:
“11. Minimum range must be at least 15 dBW, with a maximum value of at least 13.5 dBW. EIRP for R and T-channels must be adjustable under GES command over a range of at least 15 dBW in steps of 1 dB”
@George Tilton
Thank you for your cogent depiction of the processes occurring, which appear to elucidate the observed data.
@CMR: FdC has made so many incorrect or unverifiable claims that most informed investigators just ignore her. I have long suspected that she was deliberately misled by French authorities on a number of matters, and she blindly trusted her sources. She evidently does have high level contacts, as the RMP report was given to her with no restrictions.
JW is a different story. He has the capability to understand a lot more, but once he published his theory on the Russian diversion to Kazakhstan, he cherry picked information that supports that theory in a classic case of confirmation bias, and false accusations were part of that. I’m probably one of the few here who does not completely dismiss the theory that the plane was diverted north and the BFO data was altered. However, at this point in time, I estimate that scenario to be much less likely than the debris field is in the SIO and simply not yet found.
@CMR
I also think that was a good interview with Blaine. I believe Blaine’s (will appreciate your comment) contribution to the podcast has been well received and informative. Down to earth, honest concerned citizen and good defense that the debris finds are real and prove an SIO crash site. If we could find the crash site I’d nominate Blaine and Inmarsat for Nobel prize. Unfortunately that may take a while. As a concerned citizen myself, I can tell you concerned citizens do not have to be correct about everything, but we should be honest and willing to learn if we are wrong. The big picture, I think it is possible that Blaine and Chari could be correct that MH370 lies in the vicinity of the drift calculations that allowed Blaine (after several years effort) to finally get a feeling about where to go to find MH370 debris. Sort of an incredible leap of human effort to do that as an individual adventurer.
@George Tilton
I have two questions for you:
1) Given that the only communication between the SDU and the satellite was a few simple and basic handshakes after 18:40 (and two of them only concerning attempted phone calls), I would like to know how you think GES could tell the SDU to change its output unless its permanent memory contained a preplanned recipe or look-up tabel covering the entire globe (everything in temporary memory would have been deleted during reboot, and ACARS was turned off)? In particular since no increase in output was needed. It was within normal range. That goes for the measured as well as the expected values.
2) The deviation in received S/N (mostly power) from the expected values developed smoothly with an exponentially decaying speed starting abruptly shorty after 21:00 (as expected for a thermal effect). How could that originate in a series of discrete 1dB steps, which would have had an increasing probability with distance from the satellite position (projected on the earth surface) if it originated in a preplanned recipe or tabel?
@Viking
The SDU has a table of satellites and their positions.
After power-on and completing BITE tests the SDU using the Lat/Lon information from the IRU received over the ARINC bus chooses a suitable satellite from a table in non-volatile memory and tunes to its P-channel beacon. The logon request is on a R600 channel using the low-gain antenna with EIRP set to 12.5 dBW. The logon-request includes information on the SDU capabilities. The GES sends a logon-confirm with initial EIRP value to be used for Rd and T channels. The SDU then sends a logon-ack on a R1200 channel using the High Gain Antenna and the GES sends a logon-ack on a P10500 channel.
The pings/handshakes are a logon-interrogation “Are You There?” on the GES P10500 and are responded with a logon-ack R1200 by the SDU.
The GES can send EIRP broadcast messages in the P600 stream which likely will not appear in the un-redacted Inmarsat logs because they were retrieved by searching for the ICAO address of 9N-MRO.
For telephony the GES during the call-setup can send EIRP settings specifically for the C-channel according to the AMS SARPS. I haven’t looked at the logs for the two telephone calls for some time now…
I wouldn’t believe that the GES would use a pre-planned recipe to decide what EIRP levels to set the SDU to. Most likely it would maintain a minimum S/N margin on the Rx channel associated with the aircraft to control the BER.
@ALSM. Your earlier post stated “Thus, the BFO offset might have been off 10-20 Hz at most”. As I understand it, the “noise” characteristics of BFO are thought to contribute a sigma of ~4Hz. DSTG used a more conservative 1-sigma of 7Hz to allow for possible change in bias.
Two questions:
1) Could you elaborate the basis for your “10-20Hz at most” expectation?
2) How can the bias error documented by DSTG be accounted for? (Bayesian methods, p28 with fig 5.4 which “shows an example example the the BFO measurement errors”):
“This [constant bias] is not the case for the BFO bias term. The mean bias is different between flights and even within a single flight there is evidence of structured variation. …The residual error is clearly not zero-mean, and the mean varies with time. Substantial effort was made to characterise this
structured bias. It was found to have a geographic dependency but it has not been possible to determine a quantitative function to compensate for this change in bias. The variations in bias shown in figure 5.4 happen over a timescale of minutes rather than hours.”
For ease of reference, fig 5.4 shows BFO residual drifting from ~0 to about ~-15Hz and back to ~-5Hz within an hour. The text makes clear that this is not a unique/isolated example.
Paul: Re: “Thus, the BFO offset might have been off 10-20 Hz at most”. I was not referring to the “BFO Noise”. Your figure of “sigma of ~4Hz” is about right for the noise. I was referring to the BFO error resulting from the brief power off period of about 90 seconds between MEFE and the reboot at 00:19:27. My point was that for such a short power off period, the OCXO internal temperature could not have changed significantly, thus the frequency transient would be small (say, 10-20 Hz) compared to the what we saw at 12:50 and 18:25. This is important when looking at the BFO change between 00:11 and 00:19:37, which was 10X any possible power on transient.
Viking: Further to George’s points, the AES eirp is controlled by SDU in 1dB steps. It is changed as required by the Perth GES to maintain the desired C/N at the transponder. Too much AES eirp robs transponder capacity from other users in the transponder. Too little eirp results in a high BER. Thus, the GES sends commands to the SDU to adjust the eirp for the “sweet spot”. The 2 phone calls provide the best illustration of how the eirp is reduced to the minimum needed for a reliable BER. See for example: https://bit.ly/4kj7efy
@Paul Smithson
“For ease of reference, fig 5.4 shows BFO residual drifting from ~0 to about ~-15Hz and back to ~-5Hz within an hour. The text makes clear that this is not a unique/isolated example.”
DSTG didn’t make clear how frequent this effect occurred, nor did they specify typical moments and locations. How do you look at this?
@Niels, you ask “How do you look at this?”.
TL/DR version: the empirical data from DSTG cast doubt on the accuracy of BFO for purpose of granular terminus latitude discrimination.
As long as it is accurate within, say, 20Hz, then it is still valid to use BFO as a much coarser discriminator to a) rule our Northern path solutions b) infer a significant course change between 1825 and 1840 c) infer ongoing descent at 00:19:xx.
In further detail, I look at it as follows:-
1) From the text, it would appear that bias change *between* flights is the rule rather than the exception (“The mean bias *is different between flights*”,p28). This might reflect “retrace error” each time the SDU is power-cycled. So we should anticipate a bias step change (magnitude – few Hz?) between tarmac-calibrated BFO bias and post 1825 BFO bias.
2) The “structured variation” during flight defeated DSTG’s efforts to characterise despite data for 20 flights and privileged access to information on satcom equipment & operations. Their text is unclear whether this was a unique example or not. Fig 5.5, p29 (histogram of BFO residuals, 20 flights, 3392 in-flight datapoints) contains a only a handful of datapoints in each 1Hz bucket between -10Hz-20Hz, consistent with the number of similar datapoints the Mumbai-KL Figure 5.4. So my best guess is that this was the *only* example of larger scale “structured variation” in their 20-flight sample of 9M-MRO data, though they also had data from other aircraft.
Without speculating further on possible cause, it is safe to conclude that the accident flight would (1) *likely* be affected by BFO bias step change after power-cycling (2) *might have* been affected by a larger unexplained structured variation in order 0-20Hz. (3) in light of this, DSTG’s choice to go for a “conservative” BFO 1-sigma of 7Hz looks a bit arbitrary and is possibly too small.
I note that in their discussion section, DSTG do say that they played around with different BFO sigma values and found that it made little qualitative difference to their probability density function conclusions.
Remember, BFO Bias is also a function of Channel Types and Channel Unit. Details follow.
Summary: In 2014, the IG undertook a collaborative effort to extract additional information from the Inmarsat log. To keep it simple, ATSB chose to use only R channel data for their path analysis. We were interested in the potential use of all the available data in the log. However, it was evident that the BFO bias was slightly different for different Channel Types and Channel Units where: “Channel Type” refers to the different transmission protocols used by the SDU (R, T and C) “Channel Unit” (CU) refers to the specific frequency and protocol hardware units at the Perth GES used to demodulate the data packets (SUs).
The full paper from Dec 2017 is here: https://bit.ly/41jEOZn
@airlandseaman
That reinforces what I see with BTO differences between Channel Units.
The ground truth for MH370 at KL was on IOR-R1200-0-36E3/CUID-8 and the flight was on IOR-R1200-0-36ED/CUID-4. There is a significant offset in measured BTO between the two and I would not be surprised if the BFO wasn’t also affected. The GES MITEQ demodulator PLO in 2014 would have used a VCO in the Costas loop and not a numerically controlled DSP that appeared in smart-phones and software defined radios later.
@Paul Smithson
Thanks, Paul, for elaborating.
I agree on your point (1). Regarding (2) and (3):
Indeed by looking at fig. 5.5 it seems the type of larger BFO error sweep (below – 10 Hz in fig. 5.4) does not occur frequent. When DSTG writes they take a 7 Hz noise SD “to be conservative”, I read that as “being on the safe side”, so I tend to disagree that “it is possibly too small”.
George: Re: “The GES MITEQ demodulator PLO in 2014 would have used a VCO in the Costas loop and not a numerically controlled DSP that appeared in smart-phones and software defined radios later.”…
The Perth GES demodulators (AKA Channel Units) were manufactured by Square Peg, not Miteq. (Miteq supplied other equipment in the GES.) They were state of the art at the time, incorporating FPGAs and DSP technology. Data sheet here: https://bit.ly/4dMdtWU
@airlandseaman
Thanks!
You and Don Thompson have been invaluable whenever I hit a brick wall trying to track down information. 😉
@Niels. Is 7Hz “conservative”?
Table 5.1 p29 (Statistics of BFO errors for 20 flights of 9M-MRO prior to MH370) derives a noise sigma of 4.3Hz for BFO in-flight, excluding outliers. In addition, we have the unknown offset change associated with each power-up. If this is (say) ~3Hz, that already gets you to a sigma >7Hz without accounting for the (apparently rare) occasions when BFO bias exhibits “structured variation” excursion to ~15Hz.
It seems to me that “7Hz” might be reasonable if it is *only* intended to cover noise and retrace error. But I cannot see how it could possible allow for the larger bias excursions.
It would be less disturbing if the odd BFO values were isolated outliers but they are not. With a BFO sigma of 7Hz, an entire segment of flight affected by structured variation would be rejected at the 2-sigma (95% probability) limit.
In short, if we are to allow for the possibility of “structured variation” occurring on the accident flight (plus bias offset due to power-cycling), then we must be ready to entertain BFO residuals of up to ~15Hz+. A 7Hz sigma doesn’t do that.
@Paul Smithson
Ok I see. I typically separate out the retrace error and apply normal distribution to the remainder. Then 7 Hz is a lot for SD, see fig. 5.5 and the fit therein. Again, the larger “structured variations” seem infrequent.
@All
Many thanks for providing a lot of detailed information on BFO errors during the last few days. I need more time to digest everything to the finest detail, but below this introduction is my prelimenary conclusion.
The BFO error is composed of two conceptually different contributions. A statistical error of 4.3 Hz, and a number of possible systematic error sources. During a statistical analysis these cannot simply be added (certainly not in quadrature, and most likely not linearily).
The known (relatively rare but certainly possible) systematic errors are turns, height changes, electronic errors (most likely related to reboots), and thermal effects. When one of these things occur, it typically makes a systematic shift of 10-20 Hz. The propability of two of these occuring simultaneously is negligible, but the probability of one such effect is significant.
The first two types will typically only affect one data point, but may occur more systematically if the PIC (pilot) tried to deliberately confuse the investigators. The last two are systematic effects with influence on all data points following the time of the original ‘perturbation’ causing them, until the next reboot.
Re: MH370 BFO
Part of the reason that the 38-South hot spot was the first ATSB/DSTG/Bayesian search location solution, is that the BFO match is almost perfect to Arc5 (Arc2 BFO is a little “off” as many of us speculate a maneuver in progress at Arc2). It is hard to imagine that BFO near-perfection and still take a stand that BFO is somehow a number that flight paths projections can take much liberty. I do personally feel the near-perfect (to Arc5)BFO match for 38 South (187deg) could be due to a few numbers drift from a true path more like 180 South.
P.S. to above
This older essay explores the BFO match of 187 and 180 South paths:
“On the Straightness of the MH370 BFO vs. Flight Time Trend”
https://docs.google.com/document/d/1d5_ZM6lfcgjVelP6k6RjPIvsZeOgDYtJXELYHDewg3k/edit?usp=sharing
@TBill,
I don’t think it is analytically useful to speak of a single BFO datapoint being a good fit.
Take the example of Victors “great circle” paths that fit the data across Arc 2 to 6. These range in initial azimuth from about 175 to about 190. The BTO residuals are a bit smaller (better) as track angle increases.
Now when you run the BFO numbers you find that at 190 you have steeply trended residuals – with large errors at beginning and end. The 175 track angle version is trended in the opposite direction. BFO errors are minimised when the residuals aren’t trended – which occurs for track angles around ~180 that reach 7th arc between 34 and 35S, so a BFO probability distribution is centred around that latitude (see UGIB).
Meanwhile, BTO fit is optimised much further south. In my analyses – based on single turn LNAV path solutions that cross Arc 1 on N571 – you find the probability distribution is centred near +/-39.5S, see here.
Now, what happens if you try to simultaneously optimise fit to BTO and BFO – whose distributions are divergent? You will end up taking paths where the tails overlap – i.e. sub-optimal examples from either distribution. Which is how DSTG ended up with a pdf based on normalised, combined BTO+BFO probability at 38S. As any statistician will tell you, if you observe two such divergent probability distributions (BFO and BTO) you should ask yourself why, rather than combining them to create a centaur.
Incidentally, this is the same reason that ATSB originally noted a fundamental divergence between “autopilot” solutions (straight, furthest south, best BTO fit) and “data optimisation solutions” (further north, slower/curving) that produce best BFO fit.
To illustrate how divergent the DSTG’s pdf was based on BTO only vs. BTO+BFO combined, I digitised their figure 10.3 here.
Most of the “primary peak” in the red BTO-only distribution has not been searched. After the 2014-15 search of the pdf signified by the blue line came up empty, the “first principles review” recommended searching further north rather than south. That might be a better fit with CSIRO’s drift models, but it was a demonstrably poorer fit to the BTO data – which is the most precise and unequivocal data we have got.
If we are going back to basics, I say we should be guided by the BTO data and search 39.5 – 40.0S.
@paul smithson
“…the BTO data … is the most precise and unequivocal data we have got.
If we are going back to basics, I say we should be guided by the BTO data and search 39.5 – 40.0S.”
I could not agree with you more Paul. The BTO are the only solid data we have, the BFO’s are too problematic and only cloud the issues, and the multiple drift models are cloudier still.
Someone please tell OI to get a bigger boat than the Armada 78 (get Seabed Constructor back (or similar)) and go back to 40 south and search up to 120nm south of the arc.
@TBill
I disagree with you for the same reasons as paul smithson.
@paul smithson
I appreciate two great contributions from you during the last two days. Now we are getting into a discussion we should have had a long time ago.
Before going to the discussion, I would like to point out an extremely small detail where we disagree. For semantic reasons I do not like that you use the word divergent about the data. From a purely mathematical point of view, they are not divergent. That takes larger discrepancies. However, in the more common sense of the word, it is correct.
Concerning the rest of the discussion, it is too long for one contribution, so I will start slowly. For the data point at the second arc, I suspected from the beginning that it was perturbed by a turn. This could either be a small one, as assumed by INMARSAT, TBill and many others, or it could be something approaching a U-turn. The analysis I made ‘in’ and ‘around’ Figure 2 of my paper (and never changed since version 1), concluded that it most likely was close to a U-turn.
It is all related to the unusual statistical distribution of the BFO values (mentioned in my last contribution). When I started my analysis in 2016, I knew nothing about the details behind the BFO, but my gut feeling from many years of data handling was that there was something fishy about the second arc point and the general distribution of BFO values (particularly during the two attempted phone calls). It all looked like a so-called tail-heavy distribution (many people prefer it the other way around ‘heavy tail’) rather than a usual Gaussian distribution of errors. In such cases one must be particularly careful during the data analysis.
Let me start with a funny example of what can go wrong when dealing with such distributions (taken from one of the most prestigious scientific journals in the world). What is the most common death reason for humans? I guess most people would answer heart disease or cancer. Depending on the finer details, one of these would be correct if we were dealing with Gaussian distributions. However, the real distribution is far from Gaussian. There are two reasons. The number of people on planet earth is exponentially growing and has been so for most of the last 10000 years, so the humans present at any time during the last 10000 years make up a significant fraction of all humans ever existing on the planet (before and at that time). Before 10000 years there were extremely few, so they can be ignored in this context. The other reason is that most humans live near the coastlines of the world, where the risk of flooding and tsunamis is large. Because of these two reasons combined, the most common (non-Gaussian) death risk for humans is a hit by a large meteorite causing global tsunamis of several hundred meters. Other things can cause global tsunamis (Putin’s nuclear superbomb or the outbreak of a super volcano), but the meteorites are more likely. After all, that was what killed most types of dinosaurs 65 million years ago and caused most other known mass extinctions on earth.
However, it may change in the future. Deliberate genocide may become a larger risk.
@Paul Smithson
Can you explain the two rather narrow peaks (BTO only) in the fig. 10.3? Do you know their assumptions regarding thrust settings?
I get much wider distributions for BTO only, and not necessarily agree that BTO only would mostly indicate these southerly latitudes.
Furthermore: fuel limits?
Hi Niels. Re narrow BTO peaks.
The primary peak is a good match to what I get for single turn, constant M LNAV models, For example, my FL350 model optimum is a 39.45°S with 90% CI 38.28-40.0.
By my reckoning, there should be another smaller peak representing single turn, constant M TT models, centred at about 38.6, albeit with slightly lower p-max. DSTG’s BTO-only pdf has a “bump” at this latitude, albeit a bit smaller/less prominent than my modelling produces.
Single turn, constant M TT paths don’t explain their secondary peak centred at about 36.8. I don’t know what autopilot/thrust mode this represents and I haven’t tried to model/reproduce it. My best guess is that it is a “TT or LNAV” nav mode but maybe some kind of LRC/Cost Index speed mode? Or maybe paths that flew north of N571 to Arc 1 (since this BTO-only pdf model was initiated at 1802). Or possibly paths with a second manoeuvre.
You said: “I get much wider distributions”, This is because your latitude at 1941 is loosely constrained and rotational symmetry means you can obtain a reasonable BTO fit across a wide range of track angles. In my models, the requirement of contiguous/no loiter flight from Arc 1 near N571 constrains latitude at the top of the straight path – which constrains latitude at the bottom.
Fuel limits. Not included in DSTG’s pdf, though they did constrain M to 0.84 max. I believe the only reason their BTO-only pdf extends beyond 40S is due to the small number of paths that turned early/crossed Sumatra.
@paul smithson
You said:
1. (DSTG’s) ..“BTO-only pdf model was initiated at 1802”
(which is near the island of Pulau Perak – which has a lighthouse):
AND:
2. “I believe the only reason their BTO-only pdf extends beyond 40S is due to the small number of paths that turned early/crossed Sumatra.”
I have a few questions:
Q1. Are you saying that the DSTG did model Sumatra overflight(s) beginning from near Pulau Perak ?
Q2. If so, I missed it, so can you provide a reference please ?
Q3. Have you considered a Sumatra overflight in your own modelling ?
Q4. If so, what path, and with what results ?
@Paul Smithson
Thanks Paul; I agree that possibly the southerly peak is related to constant M setting. Indeed we should understand the secondary peak as well.
Through the link you can see the results I get when fitting to a dataset containing a mix of ECON, LRC, MRC, M078, M080, M082, M084
Based on the measured BTO data “as such” the southerly latitudes have a higher probability density. The drop-off south of S38 deg and north of S28 deg is caused by the 19:41 latitude constraint that I imposed (between S2 and N10 deg).
However, if I apply random error correction (the random errors can be estimated as we have an over-constrained problem while assuming straight and level paths) there is a (weak) max around S34 deg. If I then also apply my fuel model, I find a max around S34 deg both with and without random error correction applied.
https://www.dropbox.com/scl/fi/ns13ngvbm5n75vl7fjsk0/BTO_only_fit.pdf?rlkey=oumzdayfnfjszvctrscbjkb0a&st=doxymeqm&dl=0
When including BFOs in my analysis, the BFOs actually pull the max a bit to the south (towards around S35.5).
I could repeat the BTO only analysis and focus separately on constant M and LRC.
@paul smithson
A quick feedback on the latest issue.
The asymmetric shape of your curves is probably mostly due to non-optimum flight height. As you can see, the curves get icreasingly symmetric with higher maximum for lower flight height. Did you try FL320? As far as I remember that is better (maximum peak amplitude and more symmetric peak). I may be a bit off, since I did that kind of studies in 2017 before going to the complete model.
@Niels, – as we acknowledged earlier, the reason that your BTO only distribution is relatively flat across a broad swathe of terminus latitudes is that it is largely “agnostic” with regard to latitude at 1941. What follows is a general commentary rather than a critique aimed at you.
The rotational symmetry of BTO-fitted solutions across Arcs 2-6 means that you can obtain fits that are nearly as good as one another as long as you permit free movement along Arc 2. This produces an unconstrained problem/solution that can only be resolved by invoking an independent latitude discriminator (like BFO, fuel, drift, etc).
By decoupling the path models from the aircraft’s last known time/position, you end up discarding information that is otherwise (arguably) our best latitude discriminator before invoking more imprecise/uncertain latitude qualifiers. This is the key problem that I have with the “Arc 2 onwards” modelling approach.
Now, it is POSSIBLE that there was an unexplained loiter/detour after 1828 that allowed the aircraft to meet Arc 2ff at a much more northerly latitude, but does it look plausible? It requires the aircraft to have:
1/ Followed a fast (M0.87, M0.84) and direct path between diversion and Arc 1
2/ “fiddled and diddled” (terminology borrowed from sk999) somewhere between Arc 1 and Arc 2 for reasons unknown/speculative
3/ Resumed a fast and straight path from Arc 2 for another ~5 hours
#2 above amounts to inserting a superfluous manoeuvre – which path modellers are otherwise loathe to do for fear of Occam.
A second difficulty that I have with the Start-at-Arc 2-school-of-modelling is that you run out of degrees of freedom. You only have 6 datapoints, of which the first [Arc 2] is fixed by virtue of starting at 1941 on/close to the arc. 5 datapoints left. Now you allow the following (5!) parameters to vary: latitude at 1941; thrust control mode and speed [if applicable]; initial azimuth; altitude [and thereby cross/tailwind exposure]; lateral navigation mode. Hardly surprising that a BTO fit can be found – and then require rely on triangulation with independent datasets of questionable precision to discriminate among solutions.
Now, consider the converse. Fitting a path from 1830ff is more difficult its initial latitude is highly constrained and it needs to hit an additional data point (hit Arc 2 rather than starting from it). The geometry of the problem means that any path needs to thread a very narrow gap to meet Arcs 2 and 3 (the right azimuth), get to Arc 6 on time (the right speed) AND happen to hit Arcs 4,5 with precision along the way. For those familiar with the problem, it ought to strike us as an enormous coincidence that: 1/ this can be done from a single turn soon after last known position 2/ the turn timing required matches the timing predicted by BFO (between 1828 and 1840) 3/ all analysts agree that LNAV path models ending far south (consistent with single turn path model) have the best BTO fit of all.
It would be one thing if such a simple path model fit was unobtainable. That would tell you that there HAS to have been an additional manoeuvre. But – against the odds – a simple path fit is obtainable. I hope folks now understand why I am so adamant that the “simplicity” of the far-south solutions cannot be regarded as a coincidence.
Incidentally, what I am saying above was mainstream doctrine before the first search came up empty and undermined faith in the “simple” scenario. The problem is that the baby was thrown out with the bathwater.
@PaulS
fyi I believe I saw some minor reported-BFO corrections in the xls from that older memo, so the delta-BFO’s may need slight adjustment.
Otherwise I am not following above exactly. I feel BFO was quite accurate for the flight (except Arc2 probably maneuver in progress). After Arc2 some drift is allowed.
@paul smithson
Good observations. I came to similar conclusions in 2017. However, one important detail slipped my attention. The Singaporean no-observation. It actually puts further constraints on the problem.
Did you consider that issue?
By the way, I should have written FL330 in my previous contribution.
@paul smithson
Two other (very tiny) issues:
1) Did you include the small height change for the INMARSAT sat. due to its Eigenmotion?
2) Did you consider the small gravitational anomaly over the central Indian Ocean changing the sea level?
Actually the two issues are somewhat related.
@Viking.
1/ I don’t think a symmetrical p distribution is inherently superior/desirable. If the solution exhibited perfect rotational symmetry, your BTO errors would grow at a similar rate if you moved off the optimum Azimuth in either direction. However, that’s not the case because the greater your track angle, the further south you travel and the more (intensity, duration) headwind you encounter.
2/ I didn’t try FL less than FL340 because the fuel sufficiency problem is already pretty insurmountable and would only get (much) worse the lower the FL.
3/ Singapore non-observation. Not sure that this really adds any value. The BTO optimisation tells you that you need to commence your turn around 1831, less than 3 mins after crossing Arc 1 so you never get anywhere near IGOGU/Andaman Is.
@Viking – re “tiny issues”
1/ the satellite ephemeris used is the “PAR5” model (Oct 2014) derived by Henrik Rydberg from the ECEF positions provided in Table 2 of Ashton (Inmarsat) et al, Journal of Navigation, 2014.
2/ Flight model uses geometric height above WGS84 earth so doesn’t include “lumps and bumps” in the geoid.
@Paul – One question that comes to mind when reading your 10:14am post: are there similar solutions possible by delaying the turn significantly and then turn south and fly south “straight” and at a constant thrust setting, while hitting the arcs properly, So how unique is your proposed path solution?
So I’m afraid that even by limiting yourself to these “direct” paths scenarios there is still a wide range of 7th arc latitudes possible.
Another question: is the best BTO fit to the measured data necessarily related to the best estimate for 7th arc latitudes, knowing that the data is noisy?
@Niels, thanks for the question. I’ll try to explain with the help of a diagram. The single turn path models are depicted here.
1/ We presume that the aircraft passed MEKAR and crossed first Arc on or close to N571 at the designated time of 18:28:15.
2/ We then model a series of turn possibilities, separated by 1 minute intervals. At the end of each turn, we tune speed and initial azimuth to obtain the best possible fit to the BTO data across Arcs 2-6.
3/ Each of these paths has a probability score. It can be plotted as a function of turn time – or as a function of the 7th arc latitude that each path produces.
4/ The Google Earth figure shows what the turns would look like in practice. The chart shows the probability distribution of the path solutions, plotted against terminus latitude.
The narrow probability distribution is self-evident. 90% of area under the curve is contained within the bounds 38.28S to 40.05S. Turns that are much earlier or later than optimum produce lousy BTO fit.
If I understand your question, you are saying that a BTO fit that isn’t much worse can be obtained for a more northerly position on Arc 2. Yes, that’s right, but it involves a major loiter somewhere between Arc 1 and Arc 2. To give a sense of scale of delay:
Track 180 solution ending ~35S crosses Arc 2 at 3.5 degrees north.
Track 190 solution ending 39.4S crosses Arc 2 at 2.5 degrees south.
These two positions are separated by 6 degrees (360 nautical miles) of latitude. That’s a “delay” of >40 minutes with groundspeed of a little over 490kts (the speed needed for BTO fit from Arc 2).
In theory, you could turn south before 1840, then fly in circles killing time, and ultimately resume a course southward at cruise speeds. But does that look like a more plausible/probable solution than a route that takes you to Arc 2 without messing around? Put another way, why would we PREFER a solution that involved an entirely speculative loiter to one that doesn’t? Is obtaining a better fit to shaky BFO data really a good enough reason?
If you start your modelling at Arc 2, the severity of this dissonance isn’t immediately obvious. If you model from Arc 1, it is immediately clear that something very odd indeed is required to obtain Arc 2 latitude at 3.5N.
The second question: “is the best BTO fit to the measured data necessarily related to the best estimate for 7th arc latitudes, knowing that the data is noisy?” Let me try an alternative conceptualisation.
Imagine that the LNAV solutions have perfect rotational symmetry. The BTO couldn’t help us distinguish paths that start on Arc 2 because they all have identical probability.
Although the solution path can “rotate” freely around the closest point of approach (CPA) to satellite, there is a fixed and deterministic relationship between latitude at Arc 2 and latitude at Arc 7.
Now “extend” that course backwards in time from Arc 2 to 18:30 and again imagine the path pivoting around the CPA. At some track angle angle, you obtain an intersect where the hindcast [floating] Arc 2-6 path intersects with the outbound track. That intersection anchors the solution and now prohibits it from floating. That’s why I refer to the contiguous path model as a latitude discriminator.
So I’m not suggesting that you allow path solution to float freely up/down Arc 2 and use BTO to find the best path. I AM suggesting that if you posit a no-loiter path, it necessarily follows that there is a narrow range of path solutions. Within that narrow range, you use BTO fit to identify the optimum and the range of uncertainty – which is what my p-distribution does.
Having said this, the end zone of this “simple solution” also has the merit of producing the global best fit to the BTO data, as Victor described in his “great circle paths” many moons ago; as DSTG found (cf. primary peak), and as you are seeing in your own BTO-only analysis (peak 39S-40S).
@Paul S, Regarding your 6:31 pm post with the diagram: These examples, with 1 minute separation in turn time are all for M0.84 (?), if you extent for example to M0.82 can you then find scenarios with a later turn time?
Hi Niels.
No – from turn exit, each path solution has the speed/initial azimuth pairing that optimises BTO fit across Arcs 2-6. In the figure shared yesterday, the Mach number for 1831, 1834, 1840 were labelled.
Values for all turn exit datapoints (FL350) as follows:-
1831 M0.852 Az 192.03
1832 M0.849 Az 191.47
1833 M0.846 Az 190.91
1834 M0.842 Az 190.27
1835 M0.839 Az 189.68
1836 M0.835 Az 189.01
1837 M0.832 Az 188.41
1838 M0.829 Az 187.81
1839 M0.825 Az 187.14
1840 M0.821 Az 186.46
Should you wish to reproduce these specific paths, the data table with positions, 7th arc termini and BTO residuals for each path can be found here.
The metric of BTO fit is:-
Convert all BTO residuals to z-scores (implicit assumption of approximately gaussian distribution of BTO errors)
Using the published BTO noise standard deviation, convert the above into probabilities
Compound probabilities across datapoints Arc 2-Arc 6 to obtain a probability score for the path as a whole. This is the “p-value” in the y-axis of the probability distribution.
@Paul S
Thanks Paul; I see it now; I overlooked the labels.
And then if you continue the process and lower M even further the phone call BFO comes into play and that’s the limit of these early turn direct scenarios.
A question that remains is the calculation of probabilities from noisy data. I’ll come back to that later. I have some numerical experiments in mind to illustrate the possible issues.
Previous post should have read:-
Convert BTO residuals to z-scores using published SD; express z-scores as probabilities (two tails); compound probabilities for arc 2-6 to obtain path p-value. The numbers are in the spreadsheet linked above.
@Niels – if you continue the process (to turn later than 1840) you find yourself off the tail of the probability distribution. The 1840 bin is already only 0.6% of the area under the curve. So later than 1840 is well beyond improbable at the 99.5% level of confidence – in other words it produces a lousy fit to the BTO data, no matter what Azimuth/speed combo you choose.
In the data table referred earlier, you’ll see that this is because you can’t start from points further west without incurring a large/growing error at Arc 2.
@paul smithson
Near the peak, the deviations should be symmetric (even order) for Gaussain error distributions. This follows from a mathematical theorem. I forgot its name, but it dates back to Newton.
It is very clear that this is not the case for the flight heights you used, but it occurs just a bit lower. This also shows that BTO based solutions are very sensitive to height changes. I found the same in 2017, so you are not making mistakes. This is just what the data tell us.
If you take all the imperfections (deviations from sperical earth) into account it moves the optimum at least 1 degree lattitude north (as far as I remeber it is near 2 degrees). That is quite a lot along the seventh arc, and may solve part of your problem with fuel.
A turn near 18:31 might have been seen by Malaysian or Indonesian military radar. I would guess around 50/50 for one of these observing it.
@paul smithson: Looking at your plot of p-values, the peak value is 0.04. Typically, 0.05 is used to determine whether to reject the null hypothesis. What am I missing?
@Victor Iannello
I noticed the same, but I guess the curves are not normalized. Therefore, I only cared about their shapes.
Normalization can be tricky if one does not take the total ‘probability space’ into account.
@paul smithson
A turn at 18:31 might also have been observed by the Singaporean surveillance plane. Sadly we do not have sufficient details to determine the probability.
@Victor,
The probability should be regarded as something like a “likelihood index” rather than probability of a single observation on a bell curve.
This is because it is calculated as follows:
– convert BTOR to z-scores, assuming standard deviation 29 microseconds
– assign probability of observing that INDIVIDUAL z-score or better (2-tail test)
– now compound the probabilities: Arc 2 * Arc 3 * Arc 4 etc…
And that gives you the p-value for the path that comprises those residuals.
You only need a single BTOR near the margins to seriously drag down the entire p-value of a path. In the table, I have provided BTO residuals Arc 2-6 for each path, their z-score equivalents, the individual probability and the compound probability.
For those who are more familiar with the root mean square of BTO residuals, I have also provided that value in the table. The RMS_BTOR for the FL350 optimum single turn solution is 23.2.
This is a very good BTO fit in the grand scheme of things. Your previous analysis of great circle paths Figure 2 found a minimum BTOR_RMS of about the same value across ALL candidate great circle paths ending between 23S and 40S using constant M or LRC.
Incidentally, a path model with BTOR_RMS of 30+ generally scores very improbable indeed using my p-index because these paths generally include individual BTORs that are greater than 58 microsecs (2 standard deviations).
@paul smithson
That is roughly the way I thought you did it. It results in non-normalized curves. The problem is that there are other solutions based exclusively on BTO. Most notably one in Kazakstan (near the INMARSAT test-case), and my solution near Christmas Island. There are a few more. Group theory says 8 in total (with the basic assumptions you make), but the remaining ones have low probability and one of them is clearly outside the maximum fuel range (the one JW prefers).
Since non-Gaussian BFO errors are not expected to exceed 20Hz, we can safely exclude all northern solutions (I agree completely with that conclusion drawn by others above). However, this leads to a necessity for re-normalization of the other solutions in a ‘pure’ BTO model.
That leaves me with an important question. Did you try your model on my solution south of Christmas Island? It will of course need some loiter time, and/or that you assume a perfect U-turn (with the turn time as a free parameter).
@paul smithson: There are better ways to determine whether the BTO errors for a hypothetical path are what would be expected. One way would be to use the t-test to calculate the p-value for the mean and the chi-square test to calculate the p-value for the variance. Then you could combine the two values using Fisher’s method.
What are the BTO errors for your “best” path?
@victor – see data table previously referred. All rediduals for all paths – in microsecs, z-score and probability.
@Viking. “Did you try your model on my solution south of Christmas Island?”
For reasons already explained, I don’t favour solutions that involve loiter/detour/multiple manoeuvres without good reason. For Xmas Is. you would need a major loiter between Arc 1 and 2, then at least one further manoeuvre somewhere between Arc 2 and 6 to fit the data. I haven’t looked at the implicit BFO, but I think that would look pretty horrible too.
This solution is like starting with a pre-baked conclusion and then inventing a wiggly path to get there.
@paul smithson:
You said “This is because it is calculated as follows – convert BTOR to z-scores, assuming standard deviation 29 microseconds . . .”
You also said: “The RMS_BTOR for the FL350 optimum single turn solution is 23.2. This is a very good BTO fit in the grand scheme of things.”
On that point (that this is a very good fit) I disagree. A lower RMS value of BTOR does not necessarily imply a higher probability that the route is correct.
In fact, when a route is found to have a RMS BTOR of only 23 microseconds, that route has a statistically lower probability of being correct than a route which has a RMS BTOR of 29 microseconds, matching the expected value of BTO noise.
With only 6 samples, the bias in the estimate of the RMS BTORs is small, but the “noise” in the RMS value is quite large. It is likely (68 % of the time) that the correct route will have a calculated RMS somewhere between 19 and 39 microseconds, because the standard deviation of the RMS is quite large – about 10 microseconds at one sigma.
In simple terms, the “best” route will have an expected value of RMS BTOR = 29 microseconds. Routes with both lower and higher RMS values are generally less likely to be the correct route. However, with only one data set, it’s difficult to separate BTO noise from route errors (which can increase or decrease the BTORs).
If we had more information, such as an ensemble of BTORs for many flights, we could discriminate routes effectively. With only one set of BTORs for a single flight, we simply can’t assess whether a RMS of 23 microseconds, for example, indicates an incorrect route or is simply a result of a statistical outlier for the correct route. If we had a significant number of BTOR data sets, we could find which of these cases is more likely to be correct. However, with only one data set, we simply can’t tell the difference for the case you mentioned. In reality, RMS’s between 19 and 39 microseconds cannot be discriminated. Of course, if you allow turns or speed changes along the route, you can make all the BTORs zero. That does not mean that route is correct, just that you have removed independent degrees of freedom by allowing route changes.
The mean of the BTORs has an expected value of zero for the correct route. With 6 samples, the uncertainty in the mean is 12 microseconds for the BTO noise only at one sigma. So, one can eliminate routes as being highly unlikely which have a mean BTOR larger than 24 microseconds (at two sigma).
@paul smithson: I used to Fisher’s Method for your “best” path to calculate the p-value to determine if there is a basis for rejecting the null hypothesis, i.e., to dismiss the path has having significantly different BTO errors in comparison to what would be expected. The t-test for the mean had p=0.63 and the chi-square test for the variance had p=0.88. Combining these using Fisher’s Method yields p=0.88. Therefore, there is no basis for rejecting the path based on calculated BTO values, which we expected.
On the other hand, you didn’t consider fuel exhaustion and drift modeling, which would exclude the southern points of impact if our models are accurate.
@Victor, thank you for your feedback and for running the numbers.
I agree with you that there are essentially two things that we are looking for in BTO fit: 1/ errors on each and all arcs 2-5 that fall within 2 sigma [with due consideration for any contributors to additional uncertainty]; 2/ a BTOR variance that is not at odds with expectations. Given the case and the nature of the data, I think it is somewhat unlikely that we would obtain a variance that is “impossibly good” [but see qualifier below].
Between these two elements, I think the first is the one of overwhelming importance because it is of no help being “within variance” if your errors are too large. Ceteris paribus, smaller mean errors = better fit. On a single observation, that’s what a bell curve tells us – and that’s why I have converted individual observations into z-score/p-values because it provides a quantitative measure of how much more or less likely a value is to belong to the “true” population (error zero, sigma 29). I then combined the individual values by multiplication (because it is event A AND event B…) rather than adding them. Is that methodologically invalid for some reason? I guess it might also be intuitively informative to see the “mean error” expressed as a “p” but I feel that the product is more informative than the mean because it more properly penalises highly unlikely errors.
Now, regarding DrB’s remarks on suspiciously small RMS_BTOR and your point on variance. I do understand that it would be very unlikely (and therefore likely incorrect) to have zero error across all arcs because we expect our data to be noisy, to the extent of 1 sigma = 29 microsecs. Presumably, for this sigma and number of observations, there an “expected” BTOR variance and a tolerable range either side of that expectation. I guess that’s what your chi-square is testing?
Now, when we combine your two test metrics, would you say its better to:
1. Combine into a compound metric by multiplication?
2. If so, with or without weighting the two elements?
3. Treat the error magnitude element as the metric of interest and the second as a simply as a gate/qualifier?
Back to the variance issue. Imagine we are in 2026, the data has been recovered and we finally know precisely what route was followed. My expectation is that the “modelled” route will exhibit smaller mean errors and error variance than the “actual” route. Why? Because when we optimise paths we force errors to be smaller (for example by fine-tuning position at Arc 2 and the precise Mach/azimuth combination thereafter). This is readily evident if you take an “optimised” starting time/position, and then marginally detune any parameter (time/latitude/longitude/mach/initial azimuth), not to mention alternate altitudes.
For this reason, my expectation is that the “actual” route will turn out to be close to, but not precisely “on” the modelled optimum.
Finally – Your test said that my path passed the null hypothesis, but is that a “good” number, or “poor” or “excellent”? It is the relative probabilities that are ultimately of interest for empirical assessment of hypotheses. Accordingly, I’d be very happy to run that test routine across the 10 paths I shared and report back with those numbers so that folks can see what effect it has on the p-distribution of turn timing.
ps: “punishing” paths with even a single very large error is absolutely vital to discriminate candidate solutions. Because of the degrees of freedom issue alluded to above, it is quite easy to force 3 of the errors close to zero, particularly if you start on Arc 2. What will really sort the sheep from the goats is eliminating the paths that entail one or more glaring residual. And that’s why I am adamant that it is quite incorrect to say that routes with RMS BTOR ranging between 19 and 39 cannot be discriminated. Paths with RMS more than about 32 invariably (in my experience) have at least BTO residual that is pushing/exceeding 2 sigma error.
DrB – forgive me if I have taken your statement out of context “In reality, RMS’s between 19 and 39 microseconds cannot be discriminated.” Presumably you are talking about error variance rather than the size of the errors per se? Nonetheless, if “true” RMS mean is expected to be “29” with sigma “10”, then 19-39 obviously fall within the +/- 1 sigma range, i.e. perfectly likely/acceptable. Why then, would you regard RMS BTOR of 23 is too good to be correct? The statement isn’t compatible with the numbers you are quoting.
@Paul S – I apply a similar procedure as you do to calculate path probabilities. However, there is a big difference: I first estimate the random errors and subtract them from the measured BTOs. In a second round of fitting, with the corrected BTOs, smaller BTORs indicate a higher probability. With random errors still included in the BTOs, as DrB explained, a smaller RMS(BTOR) not necessarily indicates a higher probability.
In the link below, the probability distributions with and without error correction, for “all” flight levels M0.84, and for FL 335 – 365 and M0.78,0.80,0.82,0.84.
As you see the maximum towards S40 deg largely disappears when applying the full procedure. So if indeed my procedure is correct, it looks as if your peak around S40 deg is a “coincidence”: it could be related to the specific “frozen-in” measurement errors.
Note however, that my second round of fitting is not optimized yet: I apply the same 29 microsecond SD, while perhaps I should be using information about the distribution from the random error estimates (obtained in the first round of fitting by a path probability weighted summation of the BTORs for all 20 million paths).
https://www.dropbox.com/scl/fi/pbgdo1cykv2p3stdpp114/BTO_only_fit_2.pdf?rlkey=9bqi76prc46zi58hnmjwgd1iv&st=bdrca2g5&dl=0
Hi Niels,
Thank you very much for the replication using your methodological refinements.
In your results “without random error” a precipitous drop-off is observable south of 39S. This would coincide with paths that cross Arc 2 at about 2.0 degrees south.
For your results to include x-axis zone 40S-41S, you will need to extend the range of starting points on Arc 2 to 5.0 South. I expect that when you do so, you will observe a peak probability around 40S.
@paul smithson asked: Your test said that my path passed the null hypothesis, but is that a “good” number, or “poor” or “excellent”?
A null hypothesis can be rejected with 95% confidence for p=0.05. For your “best” path, p=0.88, so there would be no basis for rejecting the null hypothesis. Of course, this doesn’t mean that the path has a 95% chance of being correct. It just means your path’s BTO values are statistically equivalent to the actual BTO values when considering the mean and standard deviation.
Your paths are heavily constrained by the assumption of no maneuvers after the turn that occurs before 18:40z without consideration to fuel requirements or drift modeling. If your assumptions are accurate, so is your result. That’s the basic problem of assessing any path, with the exception of perhaps what Ed Anderson is doing with geolocation with acoustic signals.
@paul smithson
If you only consider BTO (for a fair comparison with your results) all you need to make a CI route possible is assuming one U-turn around 19:50 (with the exact time as a free parameter – only small variation of the time is necessary). No additional loiter is needed without implementing BFO. There is good reason for that assumption using results from Figure 2 of my paper.
During the early times of independent investigations (2014-2015) there was a group applying random walk software (similar to straggeling simulations for ion implantation or heavy ion cancer treatment) in combination with BTO. That gave enhanced probabilities around SE Kazakstan, CI and a broad distribution in the Southern Indian Ocean (including your position as well as the consensus position (including BFO)).
I am quite sure the probability south of CI would have been much larger in those simulations if they had added an enhanced direct backscatter term similar to direct nuclear collisions in straggeling. That would emulate U-turn(s).
The PIC had already made one U-turn (almost), so it would be quite likely making one more to confuse the investigators.
@DrB
I generally agree with your estimates, but I think you overlook an unusual complication. Most of the BTO noise comes from digital errors (due to limited time resolution). In that case a naive calculation yields a Gaussian sigma value near 23 us as you correctly state. However in the real world there is also some normal Gaussian noise.
Bizarrely, adding a bit of normal Gaussian noise lowers that result. Adding a lot of Gaussian noise behaves as expected (larger sigma). When I was young I did not believe this theoretical prediction, so I tested it experimentally, and it was actually correct.
That is the main reason I beleive the BTO values are slightly more accurate than naively expected. Probably sigma is around 17-22 us.
Please note that this only goes for Gaussian noise.
@paul smithson,
In your last comment, you said “ smaller mean errors = better fit.” On that point I agree. The expected value of the mean BTOR for the correct route is zero.
I do not agree with what you said in your prior post – that smaller RMS error is (always) a better fit. In fact, obtaining a smaller RMS error generally implies a better fit only when the observed RMS is larger than the expected value. When the observed RMS is smaller than the expected value, this implies three possibilities: (a) the route is overfitted with too many free parameters, or (b) the actual BTO noise is statistically normal but is being compensated, for this one data set, by route errors, or (c) the noise distribution in this one data set produces a smaller-than-expected RMS even using the correct route. I don’t think case (a) necessarily applies here. I have done those same fits myself, and the effect you see (RMS values < 29 microseconds) is quite real. However, that does not mean the route is more probable. In fact, the opposite is true – the route may contain errors which simply happen to offset the BTO noise for our one data set, or, alternatively, the noise distribution in the BTORs might be somewhat unlikely.
To be clear, when fitting routes, one should try to minimize the mean BTOR, but at the same time make the RMS BTOR MATCH 29 microseconds (NOT minimize it).
You said: “And that’s why I am adamant that it is quite incorrect to say that routes with RMS BTOR ranging between 19 and 39 cannot be discriminated.”
Because we have only one data set, and because the number of BTORs is limited, our estimate of RMS is quite noisy. The one-sigma error in the estimated RMS value is about 10 Hz. So, 68% of the time we would expect the observed RMS value to be in the range from about 19 to 39 microseconds, even with the correct route. So, any value of RMS BTOR in that range is consistent with the route being correct. Thus, the RMS BTOR is not a great discriminator of route validity. There is buried in this conclusion an assumption that, for the correct route, the BTORs have a gaussian probability distribution. One can test this assumption by looking for a non-gaussian distribution. For example, if one of the BTORs has a magnitude which is greater than 2 or 3 times the RMS of the other five BTOR values, one can legitimately exclude that route fit because of the outlier. Then you have demonstrated that the route must be incorrect because the outlier is inconsistent with the expected BTOR noise variance for the correct route.
In your last comment you said: “Nonetheless, if “true” RMS mean is expected to be “29” with sigma “10”, then 19-39 obviously fall within the +/- 1 sigma range, i.e. perfectly likely/acceptable. Why then, would you regard RMS BTOR of 23 is too good to be correct? The statement isn’t compatible with the numbers you are quoting.”
You are misstating my comment. I did not say “RMS BTOR of 23 is too good to be true”. What I actually said was this: “A lower RMS value of BTOR does not necessarily imply a higher probability that the route is correct.”
The caveat (“not necessarily”) comes into play when the observed RMS is smaller than the expected value. In general, pushing the RMS BTOR value smaller by route fitting results in a generally high probability only when the observed RMS is greater than the expected value. When it is less than the expected value, the route fitter can find routes with systematic errors which compensate (point by point) the random BTOR noise, producing an “artificially low” BTOR RMS. The problem is, you can’t tell in this case whether the route is correct and the BTO noise is statistically unusual for those data points or the BTO noise is perfectly consistent with expectations but systematic route errors produce a lower RMS value. So, you can’t discriminate the routes when the BTOR RMS is below the expected value, and minimizing the BTOR RMS in the fitting routine is biased to produce a route with systematic errors which compensate the individual samples of random BTO noise at the arcs. For the route fitting to be unbiased, and therefore to have the highest probability of being correct, one needs the route fitter to MATCH the BTOR RMS to 29 microseconds, not MINIMIZE it.
@Viking,
Figure 5.2 in Davey et al’s book shows an observed standard deviation of 29 microseconds for the observed R1200 BTORs, with known correct routes, including all analog and digital error contributions. There is no reason to attempt to decompose the error sources for the purpose of our route fitting, nor is there any benefit.
@paul smithson,
To put a fine point on the question of the best way to fit routes, the route fitter should try to match the EXPECTED VALUES of the standard deviation and the mean of the six BTORs. That maximizes the probability of finding the correct route. In fact, because of the small number of samples, there is a bias in the expected value of the standard deviation. For six samples the bias in the estimated value is about 4% LOWER than the actual standard deviation. This occurs because the estimate of the mean has an error which always tends to minimize the observed dispersion. That means the expected value of the calculated standard deviation of the BTORs is actually about 27.8 microseconds for the correct route, and the route fitter should try to match that number (not the independently known standard deviation of 29 microseconds).
So, in fact the expected range of observed standard deviation of the BTORs for the correct route is actually about 18 to 38 microseconds at one sigma.
Also note that calculating the RMS, rather than the standard deviation, is different because it does not use the estimated mean value like the standard deviation does. Instead, it assumes the expected mean is exactly zero for the correct route. Because this mean value of zero is the “maximum likelihood” value for the mean, the RMS equation is unbiased for the correct route. The expected value of the RMS is equal to the known standard deviation of 29 microseconds, even with only 6 samples.
That’s why I think using the RMS for route fitting, rather than the standard deviation, is generally effective. One can argue that using the RMS is, more or less, equivalent to using both mean and standard deviation. It is simpler and it is unbiased. So, matching the observed RMS BTORs to 29 microseconds is a good route fitting metric, better than minimizing either the RMS or the standard deviation.
DrB, thanks for elaborating. Your argumentation is clear now even if I beg to differ. IMO we route optimisation for BTO should seek to minimise error. The result should then not “offend” the expected variance at p=0.05 or p=0.10, if you like. I cannot understand how matching expected variance as closely as possible should be the guiding principle especially if in doing so you “prefer” routes with unacceptable individual errors.
@paul smithson,
The reason to match, with unbiased metrics, the expected statistical values is because that is our “truth”. It’s what we know based on many prior flights of 9M-MRO.
In practice, in my route fitting with fuel burn included, I didn’t see any outlier BTORs, so I did not experience any “preference” to include outliers. It is certainly reasonable to put in a check for outliers and to reject a route which produces an outlier, because that means the BTORs have a distinctly non-gaussian PDF, and therefore the route is probably incorrect.
Minimizing the errors does not maximize the probability the route is correct. That’s why it’s not the best method for route fitting. Minimizing the fitted residuals leads to incorrect routes because route errors can offset noise errors. It has the same effect as fitting too many route parameters, which reduces the residuals for the one data set we have by applying offsetting route errors.
@paul smithson: To be clear, the method I proposed to calculate p (t-test for the mean, chi-square for the variance, Fisher’s method to combine them) will be maximum for path BTO errors with a mean of zero and standard deviation of 29 μs, i.e., the standard deviation for the sample is the same as the population.
For all those discussing BTO noise, I recommend reviewing this document:
https://bit.ly/4l0LgxL
This document (generated at my request/suggestion to ATSB) demonstrates that the noise is almost 100% due to jitter in SDU response timing. The patterns clearly indicate that the noise is not random. There are multiple patterns in the data indicating that the SDU has a lot of parallel processes going on, and probably interrupt service routines that cause small delays that vary as demonstrated in the test. Finite recording resolution (20 us) also involved.
Thank you, ALSM – I’m mindful of that.
@DrB and Victor. I fully understand what you are saying. On the one hand, residuals on individual datapoints that fall closer to the mean/expected value of zero have higher probability. On the other hand, a path that had 5 residuals of zero is obviously “too good to be true” and the true path should exhibit variance in residuals that is not improbable. In practice, which will have a greater effect on the probability of a path model being “true”?
We have already seen that well-fitted paths have RMS within +/- 1 sigma from the expected value of 29. My expectation is that *in practice*, contrasting variances make rather small difference in path model probability because all of these values likely sit near the middle of the bell curve.
On the other hand, my expectation is that the t-test for individual and collective residuals is highly sensitive to alternate path models if some of those paths include one or more residuals that is out on the tail.
Thanks to Victor for suggesting relevant statistical tests. I will have a look and report back with the family of paths re-evaluated using those metrics.
@paul smithson said: In practice, which will have a greater effect on the probability of a path model being “true”?
You’re really not testing whether paths are “true”. Rather, you are determining which candidate paths can be classified as “false” at a high confidence level, recognizing that fuel constraints could further eliminate many of your candidate paths.
I get the “null hypothesis” test aspect.
Now, take a unitary observation and test if it belongs to the reference population or not.
Imagine one example is at 0.5 sigma from mean and another at 1 sigma. Neither should be rejected. But the 0.5 example is objectively more likely to belong to reference pop than the other. Is that incorrect??
By the way, I am not ignoring the drift model and fuel endurance compatibility issues. I’ll respond on those separately.
@paul smithson,
That is correct.
@DrB. What’s your view on the utility/appropriateness of a one-sample Kolmogorov-Smirnov test statistic in this use case?
@paul smithson,
The K-S test is useful to see if one cumulative PDF matches another cumulative PDF (which can be assumed). However, to be effective one needs a fairly large number of samples. I have not tried it with only six data points, but I don’t think it will do anything useful beyond looking for a single outlier as we have discussed (and which is basically the method of K-S). K-S simply finds the largest outlier in the cumulative distribution.
Bottom line question:
What is the best guess estimation for the final resting spot of the plane?
Thanks.
@DrB: Using @paul smithson’s methodology, n=5 (for arcs 2-6), which makes the K-S method even less reliable.
Having done some searching and inquisition of AI tools (notably formulabot.com) I have generated a plot of relative probability based on log-likelihood of (FL350 constant M LNAV) path models with track angle 175, 179..195 that reach the 7th arc between 31.3 South and 41.3 South. No upper limit imposed on Mach number and no fuel endurance limit imposed. Peak probability spans approximately 39.8S to 40.8S. The plot can be viewed here.
Comments welcome.
@PaulS
Regarding “loiter”, I thought Peter Foley made an interesting comment in his talk last year. To my ears anyways, Peter said something to the effect: the single turn made sense to ATSB (at the time ~2015) because the aircraft flies through Indonesia FIR space, which ATSB felt an active pilot might not do so. Thus this early ATSB flight path estimate was consistent with a preferred vision of an incapacitated pilot ghost flight.
I am in semi-agreement with that sentiment, except I suspect the pilot was probably active and thus did probably make some effort to not simply fly straight through Indonesian FIR. Which in turn explains why 180s remains a strong contender at around ISBIX.
@paul smithson,
Your plot of likelihood using BTOR only should not be interpreted as a plot of 7th Arc latitude probability, for many obvious reasons:
1. The flight might not have been flown with a constant Mach (which would have been unusual).
2. The BFORs are ignored, which are decidedly in favor of a more northerly latitude.
3. The fuel is inadequate for the far-south routes you have included, so fuel infeasibility invalidates latitude of 37S and below.
4. The locations of found aircraft debris strongly argue against 39-40S.
In my opinion, the peak in the BTOR-based probability at far-south latitudes is simply a consequence of the actual noise in the BTOs for these five data points in this one data set, which appears to minimize the RMS BTORs for a straight, high-speed, and constant-speed route. Because of the small number of data points, the error in the BTOR fitting statistic (which you do not show) is quite large, and one cannot rule out any latitude in your plot as being proved to be highly unlikely based on BTORs alone.
The overall route probability is actually near zero at 39-40S when these other factors are included. See Figure 15.1-2 (Probability of MH370 POI latitude near Arc 7) in the Ulich and Iannello drift paper in 2023.
@paul smithson: The fact that DrB’s factors (2), (3), and (4) all tend to push the likely point of impact (POI) north along the 7th arc in the vicinity of 33S-35S suggests that those factors collectively shouldn’t be ignored.
On the other hand, I cannot dismiss the fact that our highest priority areas have been searched without finding the debris field.
Unfortunately, the next search will be conducted over areas with a fairly diffuse probability density. This will make it challenging to succeed with the resources available.
G: In my opinion, the areas of high priority have already been searched, which means it will take systematic efforts over large areas to find the debris field.
How possible is it that fuel readings were faulty? From time to time when I check FlightRadar24 I see that the groundspeed of 777s is much closer to or exceeds 500kts. That would also favor southern latitudes.
Marijan: Groundspeed is relative to the ground, and increases with a tailwind. The effect of wind was already built into navigation and fuel calculations.
You probably already heard about horrible airplane crash in India today. May they rest in peace.
@Victor
The point which I wanted to make is that 500+ kts groundspeed required to reach the southern latitudes does not seem very high and is more commonly seen on FlightRadar for airplaines cruising at high altitude than let’s say 400-450kts. The only problem I see is the amount of fuel available, if the readings were correct.
@Marijan: That’s the point. Unless there is a very strong tailwind, the Mach numbers required to reach the southern end of the 7th arc would burn all of the available fuel before getting there.
I don’t see how it’s possible that the fuel loaded was more than recorded. We have the fuel receipts and the fuel on board at takeoff.
@Dr B and Victor. Thanks for the critique. We all come at this problem with a habitual frame of reference and it can be difficult to see another perspective. In your case, I can imagine a certain frustration with my insistence to take my thesis seriously when you are thinking: “it doesn’t matter how nice the theory looks – it’s not fuel-feasible!”. So allow me to try and set out where I am coming from, and why I regard Fuel Feasibility as a challenging puzzle to solve rather than a drop-dead exclusion.
Two distinct approaches to determining the most probable terminus, one I shall call “Current Orthodoxy” (per UGIB and First Principles Review) and the other “BTO-Purist”.
Current orthodoxy. No ex-ante assumption on manoeuvres before Arc 2, lateral navigation or thrust control mode. Path models must be compliant with BTO and BFO data. Solutions do not require manoeuvre between Arc 2-6 but this may subsequently be included to refine fit.
The RMS of BTO residuals is considered insufficient to differentiate solutions, so the problem is under-constrained. Employ independent data sources – namely BFO, drift models, fuel feasibility (+/-aerial search null result, possible satellite debris detection) – to discriminate latitude. The unweighted convergence of these gives you a most probable 7th arc latitude near 34.5°S. The approach is preferred because its validity is reinforced by multiple independent data sources. Ex-ante, we prefer a solution that simultaneously satisfies BTO, BFO, Drift, Fuel models and we exclude possibilities that are strongly counter-indicated by any one consideration. If the conclusion leaves jigsaw pieces that “don’t fit”, explanations are sought to resolve the anomaly.
BTO-Purist. The BTO fit of path models does provide meaningful discrimination between path models. Constant Mach LNAV models are preferred because they provide best BTO fit. Probable terminus latitude is highly-constrained assuming a single turn after Arc 1. Starting from Arc 2, a similar peak probability zone is predicted and the BTO-fit is no better than the “single turn” assumption. Implied terminus is 39.0°S – 40.0°S, of which only 39.5°S – 40.0°S remains unsearched. Using BTO fit to predict waypoint-flyable routes predicts a more specific endpoint, near 39.6°S.
The approach relies on the “gold-standard” dataset (BTO) and avoids potential contamination by less reliable datasets. Ex-ante, we prefer a solution that provides the best “flyable” fit to the BTO data, especially if this aligns with a simple and plausible route. If the conclusion leaves jigsaw pieces that “don’t fit”, explanations are sought to resolve the anomaly.
What’s [not] to like? The solutions proposed by either school include significant anomalies. The most serious problem for “current orthodoxy” is that the predicted zone has been searched extensively without success. The most serious problem for “BTO-purist” is the fuel feasibility of the predicted route. If a fuel deficit resolution can be found, the theory becomes highly attractive because it is parsimonious, inherently plausible and implies a specific and limited search zone.
I’d strongly welcome it if the hive applied their collective imagination and expertise to resolving the fuel deficit quandary.
@Marijan
The only thing I would say on fuel, whereas I assume we know the correct fuel loading, which was a bit extra for a diversion airport, however, what I would ask is: assuming max fuel efficiency, including unconventional possible modes such as power off until reboot, what is the “worst case scenario” for the max amount of fuel that might be available at 18:25? Boeing gives us some nice estimates, but it assumes (I assume) normal settings and some descents/ascents.
@Paul Smithson
You said: “I’d strongly welcome it if the hive applied their collective imagination and expertise to resolving the fuel deficit quandary.”
In my view, “the fuel deficit quandary” is “the radar quandary”. The alleged radar track post Pulau Perak Island is the root cause of “the fuel deficit quandary”. I do not accept the “official radar story” and never have. In my view, 9M-MRO turned left abeam PPI, and overflew Sumatra, going feet wet abeam WITC, in which case, there is no “fuel deficit quandary” at all.
@ventus45: Which BTO and BFO values do you accept, what is your rationale for the ones you reject, what is your proposed path, and what are the calculated BTO and BFO errors?
@All: Captain Steeeve, a B777 captain, presents an interesting possibility regarding the crash of Air India 171. He thinks it’s possible that the pilot monitoring mistakenly retracted the flaps at the time the landing gear should have been raised. The pilot flying, unaware of the gear-down, flaps-up configuration, raised the nose to the normal pitch up of 10-12°, and this resulted in a power-on stall all the way to the ground.
https://www.youtube.com/watch?v=z7EZkungFEE
There has been some debate about flaps configuration. I gather that RAT was definitely deployed – so what could have triggered that? From video I saw, engines appeared to be operating all the way in, though you don’t hear any thrust increase as aircraft loses altitude.
I don’t know if it’s true, but the media reported that the pilots issued a Mayday call shortly before the crash, which included the words “… no thrust, losing power, unable to lift”.
There are three condition that will cause RAT deployment:
1. Both engines failed and centre hydraulic system pressure low, or
2. Both AC transfer busses unpowered, or
3. All three hydraulic system pressures low.
Dual engine failure would account for the RAT deployment and “no thrust”, assuming those reports are correct.
I would not assume that the RAT was deployed. Some say the RAT could be seen in the video, but others say the resolution is too low. Others say the buzzing of the RAT could be heard.
Captain Steeve was fairly adamant that the crew could not have taken off with the flaps up as the warnings would be impossible to ignore.
@Andrew: What do you think of Captain Steeve’s proposed scenario of erroneously retracting the flaps instead of raising the landing gear? With both engines operating, it seems that even in this configuration, a positive rate of climb could have been maintained other than if the pilot flying raised the nose too much (i.e., AF 447).
@Victor
Inadvertent flap retraction is certainly possible and there have been a number of such incidents over the years. However, in all those cases the pilots quickly realised their error and corrected the configuration without further incident.
Did you mean to say “a positive rate of climb could have been maintained…” in your comment above? If so, I tend to agree. With both engines operating, twin-engined jet airliners have a lot of excess thrust. Never say never, but I doubt the aircraft would have fallen out of the sky like that if the flaps were retracted, even with the landing gear down.
It’s possible the landing gear was left down because the crew was distracted by something unusual that occurred shortly after take-off. I’ve seen that happen a number of times in the simulator, when the pilots became task saturated following an engine failure at a critical moment. I wouldn’t be too quick to rule out a loss of thrust due to fuel contamination either, especially in light of the reported Mayday call.
@Andrew: Yes, I meant climb, not descent. Correction made.
@all,
I think you can hear the RAT as it flies overhead and no sound of engine power. I also think it’s possible to see the Rat on some frames of the video.
The flaps are at a takeoff setting in the both videos and in the post crash pictures.
The aircraft runs out of energy as if it has had a double engine failure just after liftoff, the pilot makes a radio call to that effect.
There’s no sign of a bird strike.
Could it be a reoccurrence of a software issue that caused the shutdown of both engines after landing on ANA985 back in 2019?
Viswashkumar Ramesh (survivor of Air-India FL 171 incident): “Within a minute after takeoff, suddenly… it felt like something got stuck… I realised something had happened, and then suddenly the plane’s green and white lights turned on,” Ramesh said.
“After that, the plane seemed to speed up, heading straight towards what turned out to be a hostel of a hospital. Everything was visible in front of my eyes when the crash happened.”
https://www.khaleejtimes.com/business/aviation/ramesh-viswashkumar-survived-air-india-crash
Another statement of interest:
Vishwash Kumar Ramesh: “Thirty seconds after take-off, there was a loud noise and then the plane crashed. It all happened so quickly,” Vishwash, who received “impact injuries” on his chest, eyes and feet, told HT.
https://www.hindustantimes.com/india-news/40yearold-man-in-ahmedabad-hospital-says-he-survived-air-india-crash-101749734358509.html
All;
Re: Vishwash Kumar Ramesh’s reference to a loud noise..
Perhaps, what he heard, was the sound of the RAT deploying.
I noted on lift off following rotation, that there was ground debris blown about by the engine thrust. In another video of the aircraft passing from left to right in its final descent, the sound of what appears to be some No.2/1 eng noise, also the sound of the RAT.
I also believe that following prior issues with the Flap / Landing Gear levers in other Boeing models, the shape of the levers were changed to give a distinct tactile feel when used.
The rotation appeared to be rather late, as I believe the aircraft didn’t do an intersection take-off, but back tracked on the runway.
So, fuel contamination?? Either as loaded or by some other sinister additive.
Re: Air India
This “Flight Safety Detectives” episode (with former NTSB Greg Feith) is quite good summary. It looks like flaps were deployed (but we cannot say if it was the correct amount) and it looks like landing gear started process of going up. RAT may have been deployed. We do not know why, and there could be a lack of transparency for some period up to 3-months, is what India has set as timetable.
https://www.youtube.com/watch?v=nRJQj6pb9co
P.S. to above – It seems to these investigators, that India might be making mistake in a rush to remove debris before adequate documentation
Re: Air India
@TBill wrote: “it looks like landing gear started process of going up.”
Yes, the wheel tracks had started pitching up to their pre landing attitude. The fuel contamination theory seems to be a no-goer; no sign of exhaust discoloration. Maybe a combination of a hydraulic/FADEC problem rapidly triggering the RAT.
Re: Air India
Black Boxes- One person quoted that one of the black boxes had battery back-up on the 787-8. I wonder which one? DFDR apparently headed for USA/DC NTSB I presume.
@TBill
Very likely the CVR. The regulations require that large turbine-engined aircraft manufactured after 7 April 2010 have a CVR that can be powered for 10 minutes by an independent power source if the normal power source is lost. Discussed here previously after the Jeju Air crash.
The cause of the loss of VT-ANB may be shown to be similar in nature to that underlying the loss of G-YMMM which lost thrust during approach to London Heathrow. That is, rooted in the environmental conditions experienced on the day. G-YMMM experienced ice accretion in the fuel-oil heat exchangers after the cold soak of the flight across Russia from Beijing, fuel flow was obstructed, thrust could not be maintained.
I’m not suggesting the same issue for VT-ANB rather, something related to the extremely high ambient temperature and humidity during its stopover in Ahmedabad (VAAH) after operating its first sector from Delhi (VIDP). The VIDP to VAAH leg took approximately an hour so the full fuel load for the onward leg to London-Gatwick (EGKK) would’ve been uploaded at VAAH. Center tank capacity would be required for the onward EGKK sector and fuel is normally drawn from the center tank when it’s not empty by the override/jettison pumps.
A project is presently underway to upgrade refueling services at Ahmedabad: “the greenfield fuel farm facility is
expected to be commissioned by FY 2023-24 & Hydrant refueling system by FY 2025-26.“. Construction work at the fuel farm was evident in sat imagery acquired 2025-03-19. An off-airfield pipeline is in planning. Refueling bowsers are the routine.
@TBill, recorders: the 787 series is equipped with two ‘Enhanced
Airborne Flight Recorder’ (EADR) units. The EADR integrates functions of the FDR and CVR together with datalink comms. Two identical EADRs are installed, one forward, one aft. A Recorder Independent Power Supply (RIPS) provides a backup power source for 10 minutes to maintain audio recording in the event of power interruptions.
@Don Thompson
RE: “A project is presently underway to upgrade refueling services at Ahmedabad…”.
Indeed. It will be interesting to see if that project was a factor in this tragedy. In 2010, my former employer had an accident involving an Airbus A330 that had uplifted contaminated fuel from a hydrant refuelling system that had recently been extended and re-commissioned. The contaminants eventually caused the fuel metering units of both engines to seize, resulting in the loss of thrust control during approach. The circumstances here are somewhat different, but…
It appears there will be a lot of focus on the quality, flow, and control of the fuel to the engines leading to a loss of thrust. It might be that the crew never had a chance.
@Andrew,
Also, VAAH serves a ‘heavy’ aircraft only once or twice a day. If these are Persian Gulf area flights, Dubai/Qatar, the flight duration is 2h45/3h15. Ergo, the ‘big fuel bowser’ needed to fill a B788 for a 9h30 + reserves sector gets less use, perhaps greater likelihood for contamination.
@DonT
I saw people online talk about the dual 787 Combi CVR/FDR, but that does not seem consistent with press reports so far. Does that suggest that India does not need to send damaged DFDR unit to USA if they have a good copy already?
I’m sorry to learn that aviation writer William Langewiesche recently passed away. William’s long form story on MH370 was one of the best written. He traveled to Malaysia to get the story and he uncovered some facts that helped to us better understand the mystery.
https://www.ainonline.com/aviation-news/aerospace/2025-06-17/aviation-writer-william-langewiesche-dies-70
@DonT
If it was the “bowser” truck? presumably India sampled the fuel/heel in the tank to see the problem. It certainly seems like the India internal criticisms to date focus on domestic air safety issues more than blame of Boeing.
I am also sorry to learn of William Langewiesche’s death. His work was highly regarded.
@Victor
The surveillance flight from Singapore may be more important for constraining the route of MH370 than most investigators think. This becomes particularly clear if I assume captain Shah knew about it. I find that quite likely, since some previous air accident investigations in the area benefitted from information provided by Singapore.
It should be obvious to most of us that the first part of the MH370 flight (after making the airplane difficult to see by turning off transponders, ACARS and satellite communication) mostly aimed at getting far away from Malaysia while leaving minimum evidence about the flight route.
Considering the general route taken, there were 4 dangers of leaving evidence:
1) Malaysian airforce radar (range almost certainly known to Shah – provided valuable information to investigators)
2) Indonesian radar at Medan (roughly same range – did see MH370 – but no additional information)
3) Thai Radar (similar to Medan)
4) The Singaporean surveillance flight.
If I were Shah, the Singapore flight would have been my worst worry. The other dangers were predictable, and their coverage ended almost simultaneously shortly after passing MEKAR. On the other hand, Singapore provided no precise data for its flights – neither exact start time exact route or exact range covered. Only crude information was known officially.
I guess the range is relatively easy to guess (roughly twice that of standard ground-based radars), but the other uncertainties pose significant worries.
All we know for sure is that the surveillance flight did not see MH370. As far as I remember we also have some knowledge about its most extreme NW position.
I tried to guess its most likely route (starting in Singapore and passing roughly midway between Medan and Penang towards its NW point). This would have brought it close to coverage range of MH370 when MH370 was in the area near MEKAR. Therefore it is strange that it did not see MH370 – unless Shah had more detailed knowledge and carefully kept a safe distance.
On the other hand, it would not have been possible for MH370 to avoid detection if it turned south shortly after MEKAR. That certainly kills the highest probability BTO-only routes.
I think we may be able to push things a bit further. The highest probability BTO/BFO route for the consensus solution in SIO also seems to be detectable for the Singapore flight. If that is correct, it is potentially a total game changer for the whole investigation.
@All
Please try to guess the Singapore route yourself and repeat my crude estimate with more detailed knowledge.
More importantly, please look in your archives if there is more data from the Singapore flights. Did it sometimes (not necessesarily the same day) show up in Flightradar24?
Re: ChinaEastern Apparent Pilot Suicide 2022
China’s recent refusal to admit to apparent pilot suicide due to possible cultural “social instability” is at least a bit more honest and open than the outright denials Egypt (EgyptAir 990) Malaysia (MH370) and Indonesia (SilkAir 185). I noticed that Dan Gryder’s original 2022 YouTube expose (linked here back at the time) was apparently taken off line – perhaps reflecting public sensitivity and denial about the issue.
Regarding Air India 171, Jon Ostrower in his The Air Current is reporting that Boeing investigators are focusing on the position of the fuel cutoff switches. The implication is that the cutoff switches were deliberately used to shut down the engines.
https://theaircurrent.com/aviation-safety/ai171-investigation-fuel-control-switches/
@Victor
Holding my thoughts for Air India developments since we are at a key juncture with the initial report due.
Technical fine point for Andrew- I wonder if B777 fuel-cut offs have the detent, as in the simulator I do not seem to have to raise the cut-off lever to move it. Of course that could just be simplification in the sim.
@Victor
Your description from 2017 of possible impact of the Singaporean surveillance flight is a very nice piece of work. However, it has one important limitation. It does not take into consideration how long time the flight remained in the holding pattern, and what it did during its trips from Singapore to the holding position and back.
To estimate these things, it is essential to know the motivation for the flight. In principle I see two possibilities. Either Singapore was warned about some kind of threat that night and therefore planned an unusually long flight. More likely it was standard procedure to fly up to this position and back on nights where the Indonesian radar in Aceh was turned off. I find the warning scenario unlikely, and if it were true, it could potentially open a large can of worms. Therefore, I will only consider the standard procedure scenario.
In that case, I am convinced that Captain Shah knew about the surveillance flights and carefully planned to stay out of their radar coverage.
For such planning, it is essential to know the specifications of the surveillance equipment. I think your estimate of its range (200 nm) is realistic for standard fighter airplanes (such as F16), and for a passenger airplane with its nose or tail pointing towards the surveillance airplane. However, for a standard airplane at normal flying height with its side towards the surveillance plane the radar cross-section is much larger. This extends the range. I think to minimum 250 nm. This has significant impact.
In addition, I think it is unlikely that the surveillance airplane would have remained in the holding pattern for more than 2 hours after 19:00. That would make the flight almost obsolete unless it also had the radar on full performance for the last 1-1.5 hours of the flight up there and back to Singapore.
Those details would completely change the impact of the surveillance. As an example, most early turns to the south (e.g. between 18:20 and 18:40) would have been discovered by the Singaporean airplane. That is why I wrote that all the optimum BTO-only flights would be impossible. Most likely such a flight would also have left an unusual contrail (perpendicular to relatively many contrails along the Malacca strait). Such a contrail would have been visible in pictures from the low-flying Chinese infrared satellite, and possibly also in pictures from a Japanese weather satellite. I checked, and there is nothing like that in any of those pictures.
For a medium time turn (such as the one proposed by ATSB around 18:40) the situation is more uncertain. However, because MH370 would have had its side towards the surveillance airplane after the turn, I am practically sure it would have been discovered as well.
Turns later than 18:40 typically run into other problems. If MH370 continued in a westerly direction after MEKAR it would certainly have been discovered during the turn. If it initially turned north (or moderately NW) and later south, it might have had a chance of escaping (near Bandar Aceh), but it would run out of fuel before reaching the consensus position in SOI. That practically rules out all such solutions. Flights very close to (or over) Aceh might also run into the contrail issue.
This leaves only two possibilities:
(1) Your own solution (towards McMurdo Base). It would have needed some small modifications to escape the surveillance, but the fuel would have been sufficient.
(2) My solution with the U-turn. It would have been a bit marginal on the way back from the U-turn, and that may potentially have been the reason for descending before passing Bandar Aceh. In that way, it came under the horizon for the surveillance flight on its way back to Singapore.
The nice thing about this explanation is that there is no longer an explicit need for an attempted parachute jump over Bandar Aceh. The whole flying pattern is explained by a motive of secure avoidance of the Singaporean surveillance flight.
@Viking: When I wrote that article, I was hoping that it would stir up other evidence that could be better used to discriminate among the paths. It did not. Because of all the assumptions, I’m not sure it advanced the ball at all.
@Victor
RE: Air India 171 Fuel Control Switches
The first step in the Dual Eng Fail checklist is to cycle the fuel control switches to CUTOFF then RUN. The crew may have cycled the switches in a desperate bid to restart at least one of the engines. The sequence of the thrust loss vs movement of the fuel control switches should be obvious from the flight recorders and should indicate if the thrust loss was caused by their movement, or if their movement was a response to the thrust loss.
@TBill
The B777 fuel control switches have a detent, just like the B787. The switches must be pulled out over the detent to move from RUN to CUTOFF or vice versa. It’s a standard design feature.
@Andrew: If the article is correct, there was no fuel contamination and no mechanical failure, which I interpret as no engine failure. If there were no engine failures, why would the crew change the position of the fuel control switches?
@Victor
They wouldn’t (or shouldn’t), but I don’t necessarily trust media reports.
In any case, the preliminary report has now been released. For some reason the fuel control switches transitioned from RUN to CUTOFF about three seconds after lift-off.
The report states:
“In the cockpit voice recording, one of the pilots is heard asking the other why did he cutoff. The other pilot responded that he did not do so.”
Preliminary report available here:
https://aaib.gov.in/What's%20New%20Assets/Preliminary%20Report%20VT-ANB.pdf
@Victor
Juan Browne taking the position that it could be true that the fuel control switches were accidentally moved if the detent/mechanism was not working properly (as per a 2017 bulletin referenced in the prelim report).
The Special Airworthiness Information Bulletin referenced in the preliminary report is available here:
https://drs.faa.gov/browse/excelExternalWindow/BBE820D0419DB3358625836600761396.0001%3FmodalOpened%3Dtrue?modalOpened=true
The bulletin was issued because of reports from some 737 operators that the fuel control switches had been installed with the locking feature disengaged. The locking feature is similar on other Boeing types and the FAA recommended that operators of the affected aircraft inspect the fuel control switches at the earliest opportunity. The FAA also requested that operators report any failure of the locking feature to Boeing, presumably to determine if the problem warranted an airworthiness directive.
Frankly, I’m astonished this issue wasn’t assessed as an “unsafe condition” that would have warranted an airworthiness directive at the time, but perhaps there’s more to the story. In any event, the lack of an airworthiness directive since the bulletin was issued seems to indicate the issue wasn’t widespread.
The AAIB preliminary report states that the throttle control module containing the fuel control switches on the accident aircraft was last replaced in 2023. It beggars belief that a problem with the locking mechanism would not have been reported in the two-year period since the fuel control switches were last replaced on that aircraft. Even if there were a problem with both locking mechanisms, it seems highly unlikely that BOTH fuel control switches could have been inadvertently moved within one second of each other. Under normal circumstances, neither pilot’s hands should have been anywhere near the fuel control switches immediately after lift-off.
@Andrew: I agree with every part of your comment.
I think Juan Browne has nailed it. If the switches had both lost their locking mechanism they may have been accidentally bumped off by the Captain’s hand. He may have positioned his hand behind the throttles during the takeoff roll.
@Tim
I have an issue with some of the comments Juan made in relation to the position of the pilots’ hands during the first B787 take-off video. At 08:15 he says “Normally at that point if I’m the flying in the right seat, I’m going to transition my hand over to the thrust levers and take command of those thrust levers”. At around 08:48 he says “…the FO should either have both hands on the yoke, or get this left hand over there on the thrust levers, where they should be“.
The practice of the FO placing their hand on the thrust levers after V1 was strictly frowned upon at the airline where I worked. I know at least several other airlines that had the exactly the same position, as did a few Boeing simulator instructors that I came across during my time on the B777. I doubt that position has changed; the whole point of removing hands from the thrust levers is to remove the temptation to reduce thrust in the event of a problem after V1.
The same goes for Captains placing their hand behind the thrust levers after V1. It was reinforced that the Captain should take their hand off the thrust levers and place it on their leg, well away from the thrust levers, as the Captain clearly did in the first video. Placing a hand behind the thrust levers after V1 serves no useful purpose during take-off. The autothrottle should be in HOLD mode at that point and cannot retard the thrust. That mode should be checked earlier in the take-off roll, after the thrust is set. Unfortunately, in this case we may never know what the Captain did (if anything).
@Victor
Pending any unexpected new developments, AI171 is apparently yet another case of pilot deliberate action.
@TBill: Based on what we know, I think this is most likely, but not certain.
@all Does anyone know what the black line on the wing surface is – above the engines in this photo? Taken from a video. It seems it is there only momentarily, above each engine. Related to spoiler deployment?
https://media.assettype.com/gulfnews%2F2025-06-16%2Fu192h9uo%2Fplane-crash-AI-171-1.jpg?w=480&auto=format%2Ccompress&fit=max
@shadynuk
Not sure myself but Greg Feith I believe mentioned it here today (flaps were used because we can see spoilers, I think he said):
https://www.youtube.com/watch?v=f4cH3gRzMPA
Miles O’Brien had a good YouTube today with Les Abend, as did pilots Mover and Gonky.
@shadynuk
That certainly looks like it might be the inboard pair of the outboard (flight) spoilers.
Andrew might be able to comment on this, but I seem to recall reading that the inboard pair of the flight spoilers are what are used for roll control in the event of a failure of the normal flight control system.
@Shadynuk
The spoilers are used for roll control in conjunction with the ailerons and flaperons, in all flight control system modes (normally, secondary & direct).
It’s hard to say without seeing the video, but it’s likely the black line in the photo is related to spoiler movement, possibly in response to a large control input by the pilot flying.
Thanks for the responses.
In the video linked below you can see this feature on both wings at the 40 second mark. The aircraft is on its way down already. Perhaps both spoilers will deploy momentarily in response to turbulence? I set the YouTube playback speed to 0.25 to see it better. In any case, it is not likely related to the accident.
https://www.youtube.com/watch?v=L2Fg4tSyCFE
Is there any item in B787 take-off procedure which requires that pilot-monitoring (or pilot-flying) presses/flips two switches, e.g. radio or something else?
Earlier claims that air conditioning was not working properly were never confirmed, but there may be a different reason that would cause pilot’s fatigue.
@Shadynuk: Looks like partial spoilers on both wings at 40 seconds.
The second thing I was thinking about is the actual timing required to turn off the switches. According to preliminary report fuel switches were turned into cut-off position one second apart. Is it really possible to do that in such a short amount of time?
One would really have to be concentrated, with a hand (maybe even both hands) well positioned and without any doubts in mind to perform that action in such a short period of time, especially because they require pull, turn and release motion.
@Marijan
RE: “Is there any item in B787 take-off procedure which requires that pilot-monitoring (or pilot-flying) presses/flips two switches, e.g. radio or something else?”
The preliminary report shows the fuel control switches “transitioned” to CUTOFF about three seconds after the aircraft became airborne. At that point in the take-off the PF should have both their hands on the control column and the PM would normally have their hands on the top of their thighs or in their lap. The PF would be focused on flying the aircraft to establish the initial climb attitude, with the PM monitoring; there is little else for the PM to do at that point.
Once a positive rate of climb is established, the PM calls “positive rate”, the PF verifies the positive rate of climb on the altimeter and calls “gear up”. The PM then selects the landing gear lever up.
@Marijan
RE: “Is it really possible to do that in such a short amount of time?”
I think a person who knew what they were doing and who was determined to quickly shut down both engines, could do so in one second with one hand. The switches aren’t far apart and the “pull, turn and release motion” isn’t as complicated as it might sound. We don’t know if that’s what happened, but I think it could be done.
Thank you, Andrew.
I found the following video on PPRuNe. It has a very good description of the B787 fuel control switches and their operation.
https://youtu.be/-ur234kwnhk?t=895
https://www.thetimes.com/article/6c2c0fb9-6644-42a3-a913-c6b047e58d2d?shareToken=1e30906bb6c4cbdca53f378aba71
6f26
@Victor et al
In today’s UK Times newspaper. A report by US investigators saying the Captain turned off the fuel cut off switches.
This link may have a paywall but I’m sure you’ll read this in the American press today.
As an aside, the profile of the Captain of flight 171 has many similarities to that if Zagari
Of Zahari
@Julia. Thank you for the link.
It’s admirable that people are not quick to accuse one of the pilots before more evidence is known, but from we now know, pilot culpability is by far the most likely scenario, and has been since we first learned that the flight data recorders indicated the fuel control switches were moved to the cutoff position.
While I understand the argument below that one of the pilots was responsible for the Air India crash, to a non-expert it seems a very inefficient way to bring down an airliner.
I appreciate it was 100% successful (if the theory is correct) but it seems fraught with risks. Not least that flipping the switches back in less than 10 seconds could conceivably have allowed the plane to recover.
As I say, I’m not an expert at all, but I would say this one definitely needs the final report before the pilots get blamed.
Wouldn’t the flight recorder playback simply be showing the flight system status for the fuel supply, not the physical position of the switches? If so there is a possibility of a software malfunction, or even hacking.
@vodkaferret, @John: Admirable!
Quote: US aviation expert, Mary Schiavo, has rejected the claims of human error that a pilot downed the Ahmedabad to London flight by cutting off the fuel supply.
“There is no evidence of this,” she told FinancialExpress.com.
She based this on a similar incident from 2019, involving an All Nippon Airways (ANA) – Japan’s largest airline. She said that the switch moved from “Run” to “Cutoff” without the pilot’s intervention. This raises concerns about a potential software glitch in the Boeing 787, an issue she had flagged earlier.
https://www.financialexpress.com/india-news/air-india-crash-why-us-aviation-expert-rejects-reports-of-ai-171-pilot-sumeet-sabharwal-cutting-off-fuel-supply-fe-exclusive/3917319/
@John
I would personally trust NTSB over Mary Schiavo as she is a layer probably wanting to take Boeing to court on behalf of NoK (as per MH370). But re: MH370 the NoK USA case against Boeing was refused (final opinion by Ketanji Brown Jackson) and the case was transferred to Malaysia.
@John. I agree with @TBill. Mary Schiavo is just looking for clients so she can mount yet another lawsuit against Boeing.
She is referring to the ANA NH985 incident in which the pilots deployed thrust reversers following touchdown, causing both Rolls-Royce Trent 1000 engines to shut down simultaneously, with no injuries reported. The exact cause was not definitively determined.
Boeing had previously issued a bulletin noting that rapid selection of full reverse thrust before the aircraft fully transitioned to ground mode could trigger the Thrust Control Malfunction Accommodation (TCMA), causing an engine shutdown.
I see no mention that in the ANA incident the fuel control switches were recorded to be in the cutoff condition. Nor was AI171 landing with thrust reversers deployed at the time the fuel cutoff switches were recorded as moving to the cutoff position.
If Boeing had preserved the reputation it had 30 years ago, none of this speculative what-aboutery would have anywhere near the traction that it currently does. That people have some doubts about the integrity of their aircraft and the associated systems is a product of Boeing’s own making.
Beyond that, I’m always amused by this propensity for people to engage in hamsteresque wheel spinning on a matter that cannot be known with certainty until more factual information is released.
Whether it is driven by an interest in cerebral self-pleasuring, or some form of moralistic or intellectual virtue signalling, it strikes me as a fool’s errand to be arguing about a matter that is essentially unprovable at this juncture. People could probably spend their time more productively on a bit of introspection by checking their own biases.
The Indian Ministry of Civil Aviation has released an appeal essentially asking for people to respect the sensitivities around the personal losses involved, and to respect the investigative process which is focussed on understanding what happened. Save more leaking, most likely by the self-interested parties that have access to the investigation, the official release of more factual information is probably a way off at this juncture.
@Mick Gilbert: Readers here like to problem solve and learn. We are not part of the official investigation team, nor do I see anybody appearing in the press or creating YouTube videos stating their opinions as facts. I don’t see anything wrong with speculating, as long as it’s presented in that light.
On the other hand, I believe your point about the tarnished reputation of Boeing is well-taken.
@Victor Iannello
Victor, I could not agree more with regards to problem solving and learning.
On this site, those processes are given a healthy kick along by the experience, knowledge and expertise of some of the key contributors; experience, knowledge and expertise, I might add, that is proffered in an entirely ego-free manner (in sharp contrast to some of the self-aggrandising, talking-head waffle that forms the basis of some (at least, one) of those YouTube channels you’ve referred to).
Thanks for the insightful discussion on the Air India crash. As a reader and traveler was wondering if a future aircraft design could prevent fuel switches to move to cut off position during take-off or till flight reaches a certain altitude.
Are there any negative impacts for such a design? Is such a design feasible?
@ST
There are some circumstances (eg engine fire, severe engine damage) where the pilots may need to shut an engine down at a low altitude after take-off. Building in an interlock that prevents them doing so might do more harm than good.
A ‘rogue’ pilot has plenty of means at their disposal to do harm, if they so choose. It would be impossible to guard against them all without removing human pilots from the system altogether. That, of course, carries other risks. We might see pilotless aircraft at some point, but in the immediate future I think we’re likely to see more rigorous and frequent psychological screening of pilots.
@ST
I see many potential “design” improvements in general, but this exact situation of instant issue right at Take-Off is a hard one. Byron Bailey has a cockpit video compromise of only 30-minute loop. Longer term more automation/AI approach more like AirBus is possible.
Why are these fuel cutoff switches on the pedestal anyway ?
They should be where they cannot be moved either accidentally or inadvertently or sight unseen by other crew, which basically means, guarded, alarmed, and on the overhead.
You could say the same about the stab trim cutout switches. At least they are guarded under red covers. They should be on the overhead too.
@Andrew,
@All,
Greater automation of commercial aircraft is inevitable, in my opinion, just as is happening with automobiles and robotaxis. An intermediate step is for the automated control system to generally operate the aircraft with a human pilot monitoring and available for backup.
In the AI171 case, a simple line of code could have prevented this crash. I can’t conceive of any situation wherein cutting the fuel off 3 seconds after takeoff from that airport is prudent. Yes, one has to be careful to allow shutting off the fuel in those situations where doing so might save the aircraft by preventing additional fire damage, but in this case the shutoff was far too soon after takeoff to allow a go-around, the engines were providing normal thrust, there was no indication of fire or bird strikes, and there was no clear area to make an emergency landing without destroying the aircraft and killing more people on the ground. So, shutting the fuel off under these circumstances should never be allowed by the flight control system, whether the input was accidental or intentional.
I agree that today a determined pilot can easily commit mass murder by taking nefarious action in flight. Heaven knows there have been far too many cases of this since MH370. We certainly should be demanding cockpit video recording and frequent psychological examination and reporting to the airline and to safety organizations. Privacy concerns must not continue to obscure perhaps the greatest current threat to air travel safety – pilot suicide and mass murder. I am also in favor of preventing the flight control system from responding to certain inputs from the pilot when executing those commands puts the aircraft at great risk. In my opinion, this will save lives by both defeating and discouraging suicide attempts.
In the scenario of MH370, where depressurization seems likely to have occurred, are there any circumstances where one should one allow a pilot to depressurize a passenger aircraft at high altitude? Similarly, when should one allow a course to be flown which is not flyable without certain death? Indeed, there are a lot of methods by which a pilot can commit mass suicide/murder, and it is difficult to prevent them all. In my opinion, in today’s environment it is also quite dangerous to continue allowing the cockpit door to be sealed by the actions of one person. Perhaps coordinated action by two pilots should be mandated.
@ventus45
The fuel control/stab cutout switches are located on the centre pedestal for a very good reason: to allow the pilots quick access in time critical situations where they need to use the switches for their intended purpose. I would not like to be faffing around looking for switches on the overhead panel when the aircraft is shaking itself to bits after losing a fan blade, for example.
There might well be a case for additional protections to prevent inadvertent or malicious use of such switches, but let’s not throw the baby out with the bathwater.
In other news, the South Koreans apparently withdrew an interim report into the Jeju Air accident, following protests from the victims’ families:
Seattle Times: South Korea pulls plane crash report after victims’ families protest
(originally published in the NY Times)
The interimm report into Air India 171 says that the flight data recorder shows that the fuel cut-off switches had been moved to the OFF position.
That is actually a supposition.
The FDR cannot “see” the physical switch positions. It can only record an [electrical] “event” that would have occurred if indeed the switches had been physically moved.
Could such an “event” occur for some other reason?
Has that possibility been investigated?
@Andrew, TBill, Dr, B and others – Appreciate your inputs. It is all very insightful.
On the Air India Crash, there is an article today that is trying to determine if there were issues with the sensors on the tail of the flight
https://www.hindustantimes.com/india-news/air-india-crash-probe-investigators-examining-signs-of-electric-tail-fire-before-fatal-air-india-crash-101752985918456.html#google_vignette
@Brian Anderson
Investigators probably analyzed the CVR recording for the presence of sound of switches being flipped.
@DrB
As a ChE we never go into a confined space without monitoring the life conditions, also with manhole watch with radio call out, except I suppose we let our guard down on the flight home. Helios 522 is crazy to me: nobody even knows the cabin temperature. I believe newer aircraft in EU will now do a better job monitoring for various emergency conditions; USA did not approve yet, falling behind again. Boeings in particular need more tamper resistance for dives into ground and depressure of cabin, ELT etc.
@DrB
Yes, greater automation of commercial aircraft is inevitable, but I think we are still a long way from fully automated operations, even with a human pilot for back up. EASA recently stopped working towards a regulatory framework for single pilot operations, after concluding that an equivalent level of safety with two-pilot operations cannot be sufficiently demonstrated with current technology. That position may well change in future, but smarter cockpit technology is needed than we have available in today’s aircraft. More here: eMCO-SiPO Extended Minimum Crew Operations – Single Pilot Operations — Safety Risk Assessment Framework
I agree that shutting an engine down 3 seconds after take-off is unlikely to be prudent in an engine failure or fire scenario. Airbus and Boeing recommend that pilots should not take action until the aircraft has attained a ‘safe’ altitude of at least 400 ft AGL, with the aircraft under control and the landing gear retracted.
Current generation aircraft use electrical relays to control the closure of the fuel shutoff valve when the respective fuel control switch is selected to CUTOFF; that process is not controlled by software. Incorporating a software interlock that would prevent a pilot inadvertently or maliciously shutting down an engine at very low altitude would require a change in design philosophy. There may be a case for such a change on future generations of aircraft, but I can’t see the regulators imposing the associated costs on the operators of today’s aircraft.
You asked: “…are there any circumstances where one should allow a pilot to depressurize a passenger aircraft at high altitude?” None that spring to mind, but pilots may need to manually control the outflow valves if automatic control of the pressurisation system fails. Again, there may be a case for design changes that would prevent excessive cabin altitude under manual control on future generations of passenger aircraft, but I don’t think the regulators would force such changes on today’s aircraft.
RE: Jeju Air 2216
The media is reporting that investigators have concluded the pilots mistakenly shut down the wrong engine following a bird strike during the first approach. I haven’t yet found any official confirmation.
The Chosun Daily: Jeju Air crash linked to pilot mistake in engine shutdown, report finds
@Victor
I may have made a small breakthrough concerning the Singapore surveillance flight. After a few days of vacation, I got the idea to look for contrails from that flight outside the Malacca Strait.
In pictures from the low flying Chinese IR weather satellite (from 19:50 plus or minus 5 minutes depending on the rotation direction of the satellite orbit), I found a circular contrail feature with roughly 50 km diameter centered around your most southeastern star (extremum for observing the ATSB route).
That explains several things. First of all that ATSB did not reject their own route based on failure to see the airplane from the Singaporean surveillance plane. However, if the range of radar coverage is 250 nm from the side of MH370, this conclusion becomes invalid.
This rejects the ATSB route. In addition, due to the circular flying pattern the effective maximum range is extended by another 25 km to the east. That practically kills anything like the consensus route. Extremely early south turns by MH370 would also have been caught by the Singaporean flight independent of its detailed route if it turned its radar on some minutes (or longer) before arriving at the holding pattern. I presume it did.
The next question concerns the absolute timing. Assuming that the time for first arrival at the most NW point (19:12) is correct, the airplane should revisit this position roughly every 15 minutes, since the speed must be reduced for optimum radar performance. The Chinese satellite observed an airplane in a holding pattern 1/8 of a turn before the NW point. If we presume it is actually 1 turn plus 7/8 it fits poorly with the time (19:40) of the passage for the Chinese satellite if it moved north, but perfectly (19:55) if the satellite moved south, and the airplane took one more turn.
For those who want to revisit the Chinese pictures, please note that they are tilted roughly 20 degrees in addition to the orbit tilt (for some reason unknown to me). Please also note that the circle is slightly distorted to an ellipse (most likely for related reasons). Finally, it is obvious that the circle is flown in clockwise direction (natural for someone from Singapore, I presume).
The next big question is when the airplane arrived at the ‘holding pattern’, and when it left again. Again assuming that the first passage-time was correct it must have arrived 19:05 at the most southern end of the holding pattern. Sadly there are no useful satellite pictures to confirm this.
I used French satellite pictures (METEOSAT 7) to find out when the surveillance airplane left the holding pattern, since the position is too far west to get sufficient picture quality from the Japanese weather satellite. The French picture from 22:00 shows the last roundtrip extending much further north and then slightly south. After that, I presume the airplane returned to Singapore.
That is corroborated by the next French picture from 22:30, indicating a (very faint) contrail pointing towards Singapore, while the previous holding pattern is fading.
@viking: That’s very interesting. Can you share the images showing the curved contrails?
@Victor
The one from the Chinese satellite was previously published in this forum (and in a couple of other places). If you have trouble finding it, I will help.
I will mail you the other two (from METEOSAT 7). Please confirm that I can use your regular mail (you do not have to repeat it, just confirm).
@Viking: It would be helpful if you provided the link to the Chinese satellite image as well as to circle the curved contrail.
Whatever email address you used in the past to reach me should still work.
@Victor
I emailed you the two METEOSAT7 pictures a few seconds ago.
Unfortunately I did not keep the original link to the Chinese picture (I am sorry for that mistake). That is the reason it may take some time to find the original, and we may need to help each other. There is a modified version in CAPTIO’s homepage, but I fear reduced resolution.
Before mailing the two METEOSAT7 pictures I checked again that the elliptic feature is at the same position in those (faint at 22:30) as in the Chinese picture. That proves the surveillance plane kept circling in the same ellipse for a long time (almost 3 hours).
The last round was taken further east and north. Perhaps the Singapore airplane was warned about MH370 at that time and therefore moved east for the last round? Sadly an hour too late, and too far north.
@ventus45
Probably worth noting that Boeing have built something like 15,000 commercial aircraft (later B737, all B747, B757, B767, B777, and B787) with that style of Fuel Cutoff Switch located at the base of the thrust lever quadrant, and neither the switch design nor placement has presented much of an issue previously.
@Andrew
Re: JeJu- If I understand, apparently unfortunately the wrong engine was turned off, and the HALON fire bottle(s) were released into it. Why does the HALON permanently extinguish the engine?
@TBill
The halon shouldn’t “permanently extinguish the engine”. The extinguishant is released into the nacelle area around the outside of the engine core, not into the engine itself. It dissipates quickly and shouldn’t cause any damage that would prevent a restart.
However, the pilot would have pulled the engine fire switch before the bottle(s) were fired, isolating fuel, hydraulics and electrics from the engine. That action is reversible and the engine could then have been restarted, but if the pilots had mistakenly shut down the wrong engine they were likely in a state of confusion and under immense time pressure. I can only speculate, but it seems unlikely they would have had the presence of mind to reset the fire switch and then attempt a restart.
@TBill
A couple of additional points:
First, if the pilots did shut down an engine as widely reported, it’s likely they used the following memory items from the Engine Fire/Engine Severe Damage or Separation checklist:
1. Autothrottle (if engaged) —– Disengage
2. Thrust lever (affected engine) —– Confirm —– Close
3. Engine start lever (affected engine) —– Confirm —– CUTOFF
4. Engine fire switch (affected engine) —– Confirm —– Pull
5. If the engine fire switch or ENG OVERHEAT light stays illuminated:
Engine fire switch —– Rotate to the stop and hold for 1 second
If after 30 seconds the engine fire switch or ENG OVERHEAT light stays illuminated:
Engine fire switch —– Rotate to the other stop and hold for 1 second
Several reports said the pilots activated the fire extinguisher(s), but if there were no fire or overheat warnings, the pilots may have pulled the engine fire switch without activating the extinguisher(s). That action would have isolated the engine and is reversible, as I said previously.
Second, Reuters reported that the investigators have clear evidence the crew shut down the left engine, including a “physical engine switch found in the wreckage”. The report does not say if that switch was the engine start lever or the fire switch, but the use of the word “switch” may indicate it was the fire switch. If that’s the case and the switch had been pulled and not reset, then a restart would not have been possible.
@Andrew
…many thanks
The NTSB has released an animation of the crash between the PSA CRJ700 and the military Sikorsky helicopter at DCA. Other than the painfully slow narration, it’s very good at explaining what occurred.
https://www.youtube.com/watch?v=SQm-fRrNMjM
My name is Ryan Hall. I’ve posted a public-data MH370 analysis that forces every domain to agree or the candidate is out. My working hypothesis has been:
“MH370 could be found if all data (publicly available) could be constrained into isolating a proper location by tying all fields that link satellite data, drift data, crash physics, fuel exhaustion, previous failed searches, and descent models to determine a POI (Point of Impact) where—when reverse-drift analysis is applied from Réunion, South Africa, Mozambique, Tanzania, and Madagascar’s debris recoveries—all data should agree with each other to determine the coordinates of where MH370 crashed. All physical domains have to agree in order to properly locate the aircraft.”
How I enforced that in practice (short version): BTO/BFO fit first; then terminal right-bank spiral dynamics; then crash physics + buoyancy classes; then UWA-style forward/reverse drift timing; then prior-search mask + bathymetry (no “missed by sonar” unless there’s a defensible shadow). Any miss at any step → that candidate is out. LLMs (ChatGPT-4/5 and Copilot GPT-5) were used only as calculators/orchestrators under supervision—fixed seeds, fixed tolerances, no hypothesis generation.
Result: independent domains converge at 31.4°S, 90.4°E with an operational 20-km ring. The real BFO final segment (182 → 0 to −2 Hz in ~8 s) is reproduced by a tightening right-bank spiral; drift timing to Réunion → Mozambique/South Africa → Tanzania/Madagascar lands in-window without widening initial spreads. Sites inside prior lanes are rejected unless a terrain-shadow rationale exists.
All artifacts are here (BFO CSV/PNG; 20-km ring GeoJSON/CSV; PDF/MD), plus two small tools: a Replicator (rebuilds artifacts locally) and a Pass/Fail Validator that applies the same gates to any alternate POI.
Folder link:
https://drive.google.com/drive/folders/1ko-et1Vv123lQEqk7g0BpsY85-j48Z4A?usp=drive_link
Constructive scrutiny invited and needed. I’m not married to this POI, but my hypothesis I am.
—Ryan H.
@Andrew
Re: AI#171- How do the AirBus throttle controls compare to Boeing re:accidental or intentional cutting of fuel or fire cut-offs? I am thinking the fuel cut-off switches are functionally similar, but some sources say the throttles must be cut back before fuel switches can be set to cut-off. Presumably still there would be vulnerability during early climb phase. Not sure about engine fire cut-offs (which were used in the 2023 Alaskan Air (Embraer) attempted case).
Re AI 171,
Where is the evidence that the cut-off switched were manually moved?
The FDR records the result of some electrical transition, a signal that might have originated elsewhere, without the switches having been moved at all.
Brian Anderson,
From what I have gleaned, the fuel control switch has 4 poles, and at least one of those poles is the sensor that directly controls the signal sent to the EAFR (i.e., FDR).
@Brian Anderson,
Re: VT-ANB,
I agree with your sentiments regarding the so-called transition of the cut-off switches. The report provides no evidence that the CVR had recorded the ‘clicks’ associated with the movement. Neither does it give any indication of the EICAS alarm sounding, which could have masked those clicks. Possibily that Alarm was switched off before the aforesaid actions and comments.
Though a limited description of a few words exchanged between the FD crew – without identifing which pilot said what, is even stranger. We know the Co-Pilot was flying the aircraft, and at that stage of the flight is something that he wouldn’t have done. The implication is that the Pilot was responsible as the #1 Eng cut-off was the first recorded.
I’m well aware of the “good-morning to you” & “good-afternoon to you” YouTube videos coming out of Perth, WA., along with a few disjointed videos by another Perth resident, who has focused on the 24VDC supply relays etc..
The AAIB of India has not helped in this ongoing saga. The securing of the crash evidence was less than ideal, evidenced by numerous videos available online.
@TBill
The fuel control switches on Airbus aircraft (known as engine master levers) operate similarly to those on the B787 and other Boeings. The engine master levers may be selected to OFF with the thrust lever in any position. However, the normal shutdown procedure is to retard the thrust lever to idle and then select the master lever to OFF.
The engine fire switches are located on the overhead panel on Airbus aircraft. They have a guard that must be opened before the switch can be activated.
It appears Richard Godfrey is now aqn expert on the Air India Crash.
https://42kft.com/air-india-171-a-plausible-hypothesis/
@Pilatus
I will just be philosophical and observe the analogy to MH370. On MH370, Malaysia along with FAA, NTSB, Boeing, AAIB announced apparent deliberate diversion to the SIO as early as 15- and 25-March-2014. I personally trusted that announcement as the preliminary truth, but there was widespread outrage, denial and distrust. Many aviation “doubting Thomases” insisted on complete disclosure of all secret military and other data. Some MH370 data is still secret, but after about 5-years we had most of the key data, and deliberate diversion was reluctantly accepted by most of the doubters.
Deja Vu all over again for me, although it is still a little early, and we need to open to any new (AAIB/NTSB) developments.
@Pilatus: If Godfrey wants to challenge the accuracy or the interpretation of the FDR and CVR data based on less reliable and less accurate data sets, it won’t be easy. Until I see reasons to doubt the official investigation, my money is on the NTSB and Boeing properly validating and interpreting the data.
On the other hand, their approach is likely to get more clicks.
@Victor
I have a concern that the U.S. legal system tends to encourage frivolous law suits against Boeing and/or Boeing suppliers for hypothetical aircraft system failures, in cases that are actually apparently deliberate pilot action (eg; SilkAir 185). What can happen is NTSB testimony about the apparent truth is disallowed, leaving the U.S. juries vulnerable to being convinced of highly technical phantom glitches. Part of the issue is human nature to deny a pilot could do this. I wonder if there is some way to allow NTSB input to such cases. Not sure, just asking if (Boeing) sees a need for some fairness protection like this.
TBill: Boeing has a team of lawyers and lobbyists to protect themselves against the Mary Schiavos of the world. There isn’t much we can do to make things “more fair”.
@TBill
Bill, I suspect that you would garner very little general support if you are promoting the notion that Boeing has been treated unfairly by the US Courts. Boeing are hardly “the little guy” in these matters.
Regarding evidence gathered in the course of Annex 13 investigations being generally excluded from criminal and civil proceedings, there should be no restriction on witnesses being re-interviewed under the relevant criminal or civil standards such that their evidence is admissible. That was certainly how the criminal case against Boeing was built regarding MCAS. The same process was open to Parker-Hannifin in the SilkAir civil trial.
@Mick
Not a shareholder, I feel Boeing/industry should be taking action to reduce potential for deliberate pilot action. But intense public denial of that problem as an accident cause, seems to cause misguided legal action citing non-existent design faults. The jury can be swayed to the wrong. I am protecting U.S. juries from being faked out by aggressive lawyers and focus issue on real problem. I am not opposed to valid legal action. As you say, the accident report(s) can be hidden from jury, so that is my issue.
It is really scaring there is no way to find or understand what was the fate of this airplane and those on board.
Hopefully, one day truth will come out.
@TBill,
… potential for [inappropriate] deliberate pilot action …
A Human Factors problem for harm minimization through training, selection and continuing support for the human.
Have you any other suggestions ?
@TBill
Bill, it is entirely misleading to say that accident reports can be hidden from a jury.
There is nothing that appears publicly in an Annex 13 investigation report that cannot be presented to a jury in a criminal or civilian trial so long as it is entered into evidence in the appropriate fashion.
That’s how the PCU failure data from the United Airlines Flight 585 and USAir Flight 427 accident investigations was entered into evidence in the SilkAir civil trial.
In fact, there are provisions in the relevant code, 49 U.S. Code § 1154, for parts of an accident investigation that have not been made public to be put before a jury.
And just by the bye, it is the presiding judge’s role to protect juries from “being faked out by aggressive lawyers”. And then there’s an appeals process open to the losing side in the event that they feel that the law has not been applied properly.
@Mick
It is really NTSB/FBI/FAA etc. opinion that I care about for USA. Annex 13 report can be wrong in cases of suspect human action due to intense denial overseas.
Re . . Air India 171
While searching for relevant schematics I came across this . .
https://www.reddit.com/r/aviation/comments/1mo7vph/how_air_india_ai171_b787_crashed_lithium_battery/
I wonder if others have seen it.
It is pretty tedious listening to the rubbish commentary, but the logic seems plausible.
He blames a faulty diode in the Lithium battery control module.
More likely in my view, than the pilots doing something dumb.
@Brian Anderson
What a horrible presentation!
I don’t know enough about the B787 to comment on the technical side, but the presentation links the battery short circuit to the selection of gear up. The timeline implies the gear was selected up about one second after the aircraft became airborne at 08:08:39 (air/ground sensors transition to air mode). That timeframe seems impossibly short, given the coordination between PM and PF that occurs before the gear is selected up.
The PF first establishes a positive rate of climb, which obviously occurs after the wheels leave the ground. The PM then verifies a positive rate of climb on the altimeter, and calls “Positive Rate”. The PF verifies the positive rate of climb on the altimeter and calls “Gear Up”. That whole process takes 4-5 seconds.
@Brian Anderson
Another point, he also seems to assume the main battery relay remains energised, allowing power to flow in the reverse direction from the Captain’s Instrument Bus back through the failed diode to the battery. I don’t think that’s correct. On the B777 at least, the main battery relay is only energised under certain conditions, none of which were applicable in this case. If the B787 system operates similarly, the main battery relay would not have been energised and power could not flow in the reverse direction.
@Brian Anderson
That guy has three videos on the subject.
His account is: https://www.youtube.com/@jeremyjohnthompson5187
25 minutes – 1 month ago: https://www.youtube.com/watch?v=fl1kUudq7Tw
31 minutes – 3 weeks ago: https://www.youtube.com/watch?v=OLxLN3osS1Y
12 minutes – 11 days ago: https://www.youtube.com/watch?v=Od5FFOipRBU
@Brian Anderson,
I obliquely mentioned the other Perth resident in my post of
August 13, 2025 at 11:56 pm
I haven’t bothered delving too deeply into his ramblings, as his reference to a ‘soft fail’ of the battery output diode didn’t resonate with me. In practical terms, diodes can pass current or block current, i.e. a binary function. However, I don’t know the type of power diode used on the B789, and it maywell be a MOSFET type which possibly could initially have had a ‘soft fail’.
Gathering the Boeing schematics and comparing them with the claimed failure sequence and what can be deduced from the known or deduced take-off timings, may reveal something.
Concerning the aircraft electrical system and Air India 171/VT-ANB
A number of commenters have landed upon the 787’s Electical System, claiming it to be the root cause of the thrust failure leading to the crash of VT-ANB.
First of all, two things are required for the engines to deliver power: 1) fuel, and in the case where the main tanks contain fuel only the spar valve and EEC controlled valves need to be considered for flow (gravity can do the rest); and 2) electrical power, typically supplied by a bus with a high level of supply redundancy (e.g. capt’s instrument bus that can derive 28V supply from main AC (4x generators) via TRUs (2x via bus ties), RAT or main battery). Note also that the EECs, once the engines are started and N2 remains above ~8%, are powered by 2x PMAs discrete the EEC’s engine and that supply only the EEC. In effect, the engines and EECs are self-powered.
In order to cease the supply of fuel it is not a matter of simply removing power. Valve positions must be changed from open to closed: this requires electrical power: the spar valve is an electrically motorised valve, motive power to open, motive power to close. Similarly, the shut-off valve within the engine domain requires power to a solenoid to open it, and to close an opposing solenoid must be powered to close it (plus, obviously, the opening solenoid must be de-energised)
Perhaps reflect on the case of QF A380 operating QF32 out of Singapore. When the aircraft landed it became apparent that control of the No.1 engine was not possible. It could not be shutdown due to the damage incurred within the wing and cabling to No.1 engine from the disintegration within No.2 engine. The spar valve could not be shut, the EEC remained powered and fuel flowed to the burners and the engine was finally shutdown by flooding the intake with fire suppressant foam by AFRS.
JJT’s YouTube presentation, referred by Brian above, exploits a non-Boeing training aid as its source of information (together with the presenters own knowledge of Starlink ground station designs that he claims to make “for the world’s richest man”). That training aid comprises 398 pages whereas the Boeing 777 Training Manual I have to hand comprises 6656 pages and, even at that, does not describe all systems explicitly to a depth that is definitive requiring other sources to complete our understandings.
The ‘Dviation’ produced 398 page training aid makes simplifications that, in this case, are entirely misleading. For example, a diagram is presented in the video that is derived from ‘Engine Control System’ at p369 (of the ‘Dviation’ source): the Engine Fuel Control Relay is depicted as a simple single pole, double contact device with a single solenoid control from the local RDC (remote data concentrator). In reality, as I set out above, two solenoids must be controlled so that power loss does not, inadvertently, cause a relay to relax and shut fuel flow.
To discuss every assertion and claim made by JJT would require the presentation to be re-edited, or a long descriptive compilation cross-referencing times in the YouTube video. I don’t intend to expend that effort, I’ll wait for AAIB-IN to publish further investigation reports.
Disclaimer: I am not related to the YouTube presenter, Jeremy John Thompson.
I think the attempts to find alternative explanations for the dual-engine shutdown of AI171 is interesting and perhaps helpful in better understanding how complicated systems on modern airliners interact.
However, Boeing understands these systems much better than the YouTube video posters who are proposing alternative explanations. At this point, I see little reason to question the official investigators’ assertion that the engine shutdown was caused by the fuel control switches moved to the off position, which is what the FDR data taken at face value would suggest.
There are also reports that the captain had a history of mental health challenges, and that “he had taken medical leave multiple times in the past three to four years, allegedly due to struggles with depression and other mental health challenges.” I don’t know if these reports have been verified. (Has anybody seen anything that supports or refutes this assertion?)
https://www.aeronewsjournal.com/2025/07/air-india-flight-ai171-crash.html
I think we also have to acknowledge that the YouTube posters cited in comments here get compensated on clicks and views, and not on accuracy.
Re: AI#171 Engines
Courtesy Ventus45, there was an early June 14 PPrune post by a former Boeing (TCMA/controls) engineer.
The former Boeing engr said Quote- “…you wouldn’t want a loss of electrical power to shut down the engine…” Paraphrasing, therefore the 787 engines are designed to keep running (come hell or high water, my words). The former Boeing engr said (on June 14) that there are only two probable explanations:(1) TCMA activation shutdown the engines, or (2) both fuel switches set to CUTOFF
He said we should hope it is not TCMA (as a version of that that is on all aircraft) (this is the thrust control management system that Mary Schiavo was citing). But at this point TCMA activation is off most radar screens (does not seem likely).
R.E. electrical failure shutting down the engines, excluding cases involving fuel exhaustion, I have found 4 instances of an aircraft that suffered an electrical failure sufficiently severe that the plane was reduced to using the battery and/or RAT. In no case did the engines fail. In one case the plane continued flying on battery alone until that was exhausted. The plane was still able to land, but then the pilots were unable to turn off the engines using the fuel cutoff switches. Once running, engines do not like to turn off.
R.E. the TCMA, I have found three instances where it was triggered, two in a 787 and one in an Airbus A220. In all cases they happened as the plane was landing and was firmly on the ground. They were all caused by the pilots doing something very non-standard with the throttles giving the illusion of a runaway engine. Technically, the TCMA had functioned properly.
The newest attorney in town is Mike Andrews of the law firm of Beasley Allen, and he, not Mary Schiavo, is scooping up the families of the passengers as clients. He has given several interviews now, and, frankly, he makes Schiavo look knowledgeable.
Re: AI 171
Thanks for all the comments. I can now go back to my normal boring routine without having to sweat over tedious Utube videos.
@sk999, interested others.
Thank you for putting some numbers against incidents where aircraft electrical power failure and EEC-TCMA were confirmed to be root causes of incidents.
Further on EEC (aka FADEC) and TCMA: EEC design must comply with Design Assurance Level (DAL) A, similar to flight control computers, and thus expected to demonstrate less than 1 in 10⁹ chance of failure/flight-hr. The EEC employed on the B787’s GEnx engines is not unique to that aircraft, the B747-8 also. The EEC design involves two identical controller channels, one actively controlling the engine and the other in standby, which alternate control with each engine start. TCMA exists to manage an uncommanded high thrust event (UHT) when an aircraft is on the ground as was described by @sk999. Unlike the pilot demanded reverse thrust on landing, take-off thrust is managed by the aircraft thrust managment function by progressively advancing the thrust levers to a pre-flight defined setting. Even though TCMA is active in both EEC channels, unlike the greater EEC function, its operation (by DAL A principles) should be bounded by other state conditions such as WOW.
Statistically an EEC-TCMA failure and, at that, simultaneous EEC-TCMA failure in both engines is not impossible but such an event is way out on the extremity of probability.
To recap, the AAIB-IN Preliminary Report states: ‘[…] Engine 1 and Engine 2 fuel cutoff switches transitioned from RUN to CUTOFF position one after another with a time gap of 01 sec. The Engine N1 and N2 began to decrease from their take-off values as the fuel supply to the engines was cut off.‘
I concur with Victor when he writes there is ‘little reason to question the official investigators’ assertion that the engine shutdown was caused by the fuel control switches moved to the off position, which is what the FDR data taken at face value would suggest.‘ I offer my comments as explanation of how some systems operate, not as a focus of root cause.
The AAIB-IN team will attract continued public scrutiny as their investigation progresses, @TBill’s oft repeated concern over human factors and influences outside the flight compartment on pilot actions is important. However, Annex 13 investigators cannot attempt to gloss over any investigative thread. AAIB-IN does not appear to be ignorant of physcological factors (ref 1 & 2).
It appears that Ocean Infinity will commence the new search for MH370 later this year. The targeted search area is still unknown.
Has anybody seen any news related to Simon Maskell’s investigation of WSPR tracking of MH370? As everybody here knows, I believe the underlying physics clearly demonstrate that it is absolute madness to believe that low power HF signals can track aircraft over distances of thousands of kilometers using the WSPR database, but Maskell disagrees. We’ve been waiting for years now to see his statistical analysis proving that somehow the established physics don’t apply. He’s received a lot of media attention for what he claims is possible, but I have seen no tangible results.
For the search conducted earlier this year, Maskell failed to produce anything usable by the search team. With the new search looming, we again are waiting to see anything of significance.
There was a previous comment that I failed to see while it was sitting in my moderator bin:
https://mh370.radiantphysics.com/2025/03/31/update-on-the-search-for-mh370/#comment-39006
It’s a new analysis of the location MH370’s debris field. I believe there are some serious shortcomings in the analysis, but I will let others form their own opinions.
@all,
Among the possible shortcomings of the work by Ryan H. there is no mention in his comment of air temperatures and winds, aircraft speed control, LNAV method, and fuel availability.
@all, re Ryan Hall’s dissertation.
Three points.
1/ ‘Bathymetry & search coverage. Global bathymetric grids (e.g., GEBCO, ETOPO) converted to slope/relief layers for sonar‑shadow analysis.‘
ETOPO 2022 uses GEBCO as one of its sources. Geoscience Aus certainly contributed Phase I MBES data to GEBCO. However, both GEBCO and ETOPO 2022 process data to 30 arc seconds resolution. At the latitudes of interest 30 arc seconds is circa 400m X-Y resolution in a DEM of the seafloor. That is poor for assessing seafloor slope and next to useless for assessing shadows in SSS results.
2/ ‘SATCOM solver (BTO/BFO). Propagation effects: Dawn‑side ionospheric and path geometry effects are applied as bounded corrections.
The scale of these effects are not quantified in the dissertation’s narrative, a reader may infer mischaracterised uncertainty in BTO, Perth-satellite-aircraft range and, hence, line of positions for 7th arc. In reality, such effects are insignificant in the gamut of all error contributions to BTO derived aircraft range.
3/ The University of Western Australia (UWA) Indian Ocean drift framework was the single most discriminating constraint in this investigation
A bold statement. Perhaps the UWA team privately shared greater detail of their work and the methods than is available in the public domain? Previous questions to the UWA team concerning detailed aspects of their work have been unanswered leading to uncertainty in the UWA’s results.
The stated result converged at S31.4° E90.4° is some 450km to the northwest off its closest point on the BTO derived 7th arc and further again from the arc of seed points used by UWA in its drift modelling. I’d suggest the dissertation should explain, more emphatically, its result.
The most recent sighting that I have found of Simon Maskell discussing MH370 is during this “inaugural lecture” from May of this year:
https://www.youtube.com/watch?v=w-G5wrqBJOY
The relevant section is 23:16 to 25:20. Simon describes his contributions to help define the search area. No mention of WSPR data.
@sk999
And a 1:00:26 through 1:01:32 provides a reality check.
1:00:26 through 1:01:32 provides a reality check.
@sk999, @George G: Thanks for the video and comments. As you say, no mention of WSPR tracking and the revolutionary new capabilities it offers.
@Don Thompson
My understanding is similar to Ryan Hall’s that the UWA drift model is relatively insensitive to longitude, so 31S/90E is valid as is 33S/100E. But you cannot go much further south than 33S and still meet the debris timings.
However, I am not endorsing Ryan’s solution as we need a logical flight path explanation, consistent with the Arcs, fuel supply, etc. In other words, how did it get over there? so far from the Arcs?
Ryan’s solution is somewhat philosophically similar to Vincent Lyne’s except Lyne flies far east of Arc7 to approx 33s/100e (and further outside the Arcs), and Hall goes further West (and inside the Arcs) to 31s/90e. If MH370 did fly well beyond Arc7, which I believe it probably did, we do not have data to know which heading it actually flew. My current betting would be south/easterly outside the Arcs not turn back inside the Arcs.
UK gov, Sep 2024: Projects from 4 UK organisations will share £2m in the inaugural AUKUS Pillar 2 Electronic Warfare (EW) Innovation Challenge. One being Univ of Liverpool whose ‘project aims to improve the ability to detect multiple individual faint signals in close geometric proximity to one another. This will be achieved using a combination of machine learning and statistics.‘
Univ of Liverpool’s Signal Processing Group brochure.
During the inaugral lecture, referenced by @sk999 above, Prof Maskell mentions ‘Neil’ in his anecdotes, that is Dr. Neil Gordon of DTSG.
An update on the Armada fleet (at Reddit), not wishing to inflict another long post from me here.
@TBill
My point 3 concerns the level of detail that is available in the public domain about the UWA work, or whether Ryan Hall obtained additional unpublished information, not how various commenters might feel about the work.
Hi,
I occasionally check in to see the thought trends.
As for such new studies like Ryan Hall’s, they are welcome as we seem to be stuck in a rut.
My most serious concern is that many of them treat each of the nominally independent factors (e/g. satellite, ocean drift) as having equal weight. As TBill points out, any proposed solution needs to account for the majority of the path, which is at least constrained to position the plane somewhere on the seventh arc at a precise time.
If a ocean drift study suggests an particular location for the debris location, then there would need be a path from the 7th arc to that location.
Sid
Re: Ryan Hall
I had a number of email exchanges with Ryan Hall back in June. He is a 32yo musician living in New Orleans, LA. According to Ryan, his father did the “simulator” work (all the math). He promised to send details to me, but never delivered. Below is the last message received from him on 6/14/2025.
Hey Mike! Thank you. I’m working out a java bug at the moment. Flight path animation specifically. Then I will put it to Google Drive. Thank you again. And just so you know, my multi-domain physics calculation is what I have the confidence, not necessarily those. If there are any variables, steps, or other hard evidence we can insert to sure up probable POIs then please feel free to edit a copy of what you wish or send the recommendation back to me. I think I covered it all, but no one can ever really know that. Thank you again. I hope this will be of help.
@airlandseaman: If he makes some interesting graphics, describes his results in a way that sounds very impressive and maximizes the buzz around AI, creates a personal narrative around the results, and releases his results and narrative to the press, certainly it will get picked up as a breakthrough for finding MH370.
@Sid Bennett.Your,”If a ocean drift study suggests an particular location for the debris location, then there would need be a path from the 7th arc to that location.”
You have a higher confidence in drift anlysis’s exclusivity and accuracy than do I. The CSIRO found that the flaperon’s arrival at La Reunion was consistent with its then work but that was before some trials indicated it would drift through the sea some 20 deg left of the wind. After that estimate, it was even more consistent.
But firstly those trials and results were themselves woolly. Secondly, the confidence in debris beaching improbability on the WA coast, a key finding as to the crash site’s southern limit, supposes that the probabilty of lasting beachings is well enough known to predict what the Leeuwin current will do to those.
To me, many findings might still be ‘consistent with’ the such studies even if of a low probability from a drift perspective.
David F. Well said.
I’d like to hear Ryan Hall explain how any pilot could fly a 777 (or any other aircraft) at 500 kts/40,000 ft (near the coffin corner) by hand for 6 hours.
@Sid Bennett
I agree completely with your comments from September 1. Because of the complexity of the case, and because of the extremely different nature of the various types of data, a carefully designed method for statistically weighing different contributions is needed.
@David F
You have a particularly strong point concerning ocean drift. However, if the thermal history derived from stable isotope ratios in the biofouling is included, and we take the average of all simulation results (weighing is certainly also needed here), the drift data point mostly to two different solutions. One lies in (or very near) the consensus search area. The other one is close to the NW corner of Australia, slightly east of the 7th arc.
Referring partly to Sid, there is no way getting to that position, and crash there in any acceptable agreement with the Inmarsat data. The closest is the most northern ending of the curved routes proposed very early in the investigation. However, getting so far north along such a route would demand flying beyond the stability point (with respect to stalling). Any straight route to there simply agrees poorly with Inmarsat data. Even BTO alone can practically rule it out.
These points have caused most people to ignore that northern drifting solution entirely.
@All
However, 3 bloggers pointed out (using different arguments and from different proposed crash positions) that the northern area might be reached by some of the debris after it was dragged there by a hurricane passing through the area from the north roughly two weeks after MH370 disappeared.
Such dragging over a long distance depends on a delicate balance between impact parameter, constant wind speed, constant distance to the eye of the hurricane, and constant macroscopic movement (both direction and speed) of the hurricane. These conditions were fulfilled on the eastern side of that particular hurricane from slightly north of Christmas Island to the relevant area. Later the hurricane quickly disintegrated and moved in westerly direction. That would mean all trapped debris would have been released in/near the relevant area roughly 3 weeks after crashing further north.
The big question is not if trapping is possible, but how big a fraction of the debris would be trapped and dragged to the new starting position for drifting. I guess around 10% from my proposed position (the shortest and most stable path), something like 1% from the position proposed by the group of French pilots, and below 0.1% from the position proposed by Ed.
The hardest nut to crack is to account for debris (the majority in all cases) which was not trapped in the hurricane. Looking qualitatively at simulation results, it seems like the majority of non-trapped debris would have beached in Somalia or Yemen from all 3 proposed crash positions, and remaining parts would never have beached anywhere, but kept drifting around the Indian Ocean until it got powderized. This could in principle explain everything, since any searching in Somalia and/or Yemen was (and still is) totally impossible.
@DavidF
You miss the drift (sorry ’bout that….) of my remarks. I have been involved of the IG from its inception and have worked with Geoff and Barry in the early days to refine a deterministic flight path model and contributed a number of papers to this group which are still probably available in the archives. With a fixed set of flight parameters it is possible to choose a Mach speed and an altitude where the flight path has a clear and distinct minimum error with respect to the satellite BTO/BFO using a constant true track scenario after the FMT,taking fuel consumption into account. It is close to the earliest IG suggested position on the 7th arc.
Originally, it was considered “bad form” to suggest that a person did it intentionally. So, the only alternative after fuel exhaustion was a spiral descent as suggested by simulator results.
Other options such as a controlled glide were not considered in the definition of the original search area (which was defined by statistical simulation techniques).
Once the initial search failed, most people (including the official searchers) lost confidence in the initial approach and the debris started to come ashore. The drift studies, however detailed they are are essentially an unproven approach over such a large time and distance. But the results have been accepted as at least as valid as the BTO/BFO method.
The highest confidence should be placed in the position of the aircraft at the time of the ping defining the 7th arc. Where along the 7th arc that occurs may be a subject of controversy. But there is no evidence whatever that the 7th arc is invalid as to either time or position.
I personally believe that there is a maximum distance of the impact point from the 7th arc that is possible using the glide end game scenario. It is certainly not 450 km. But if one was truly trying to use Bayesian reasoning (and I am hardly qualified to do so), the location predicted for the impact point should lie between somewhere on the 7th arc and the higher probability area of the drift study, properly weighted for the confidence in each result. I am not prepared to do so as my resources are limited and I and retired.
I continue to hope that this blog and reasoned analysis of informed speculations (NO not WSPR) may someday lead to a positive result.
@Sid
If MH370 hit Arc7 at say 30s there is potentially quite a bit of extra fuel because you have to have a big descent after Arc5 (without much power) to do that and also it is a shorter flight to Arc7. This is what I believe the data suggests actually happened, if we allow a “bad form” scenario. But 400nm over to 31/90 is still a bit too far from Arc7.
@Victor Iannello: Is the sonar imagery still available somewhere? You kindly provided links but they are down and unfortunately have not been archived on archive.org.
Peter Norton: You will have to search yourself for the raw sonar data provided by Geoscience Australia, as the links I have may not be current. However, I did update the blog comment from 2018 with a working link for the merged contact file from GO Pheonix that I created. The link is here:
https://www.dropbox.com/scl/fi/g4rrpia0l2mmvv6728i8y/2018-12-01-Contact-Investigation-Report-GP-merged.pdf?rlkey=2w7jsjxmgmbw6k8a3wn4mjxal&dl=0
@All
In the past months I’ve been working on testing my “Monte Carlo” approach, in which BTO/BFO data is compared to values belonging to nearly 20 Million generated flight paths. Before, I introduced an error correction scheme based on compensating for the mean residuals after a first run of fits. While this makes sense intuitively (at least to me), it needs to be validated. I’m trying how well I can reconstruct an artificial path with known settings and specific calculated BTO/BFO values (different from the measured MH370 values). I added random errors (10 times) to the initially generated “perfect” BTO and BFO values to see how well the error correction scheme works. I hope to report the results in a couple of weeks. It’s more work than I expected.
One point I confirmed over the past period is that it looks impossible to extract an estimate for the BFO frequency bias by fitting. Changing the 150 Hz leads to equally good fits at other latitude ranges (so the “bell” shaped latitude histogram shifts north or south). Suggestions are welcome. For example, one can check (for paths near the peak probability) how close to zero the residual is for the second phone call BFO, assuming 150 Hz for the frequency bias, but we know the measured values fluctuate quite a bit.
10 Reasons for 40°S
1) BTO-optimised LNAV model from LKP predicts ~39°S +/- 1° and BTO is the best data we have.
2) Matching waypoint-defined route obtainable at plausible altitude and speed (ends 39.6°S).
3) This is close to “global peak BTO probability” (40.1°) from Arc 2 models
4) The zone has not been searched and is of manageable size (4-10 days).
5) Conclusion supported by DSTG’s “BTO-only” results. Their final pdf was biased northward by solving simultaneously for BTO and BFO
6) The BTO data strongly supports “no turns” after FMT.
7) BFO data is demonstrably “unsafe” and the bias might have shifted or drifted
8) Ocean drift models support a 4° more southerly origin with plausible parameter changes: leeway/stokes @3-4% of wind and directional offset of 10°-20°. Inherently fuzzy anyway.
9) Fuel deficit (vs normal/expected) of ~6% can be bridged by a combination of: usable residual fuel (1.5%), reduced electrical loading (0.5%), savings attributable to ECS off (3-5%).
10) The rest of the 7th arc has been searched and significant data holidays eliminated. Wider swathe requires highly dubious end-of-flight scenarios. Further north (than 27/28°S) is vanishingly improbable.
Recommendation: Search 39.5°-40.0° South (reference 7th arc at 35,000ft) out to 20NM either side. Box size ~50NMx40NM. Top priority about half of that.
@paul smithson
I agree that BTO is our best data set, but I think you may have overlooked my conclusions based on the positions of the Singaporean surveillance flight, and its radar specifications. A position so far south is simply impossible to reach without radar detection, particularly if you also take fuel limitations into account.
@Viking.Thanks for the comments.
1) The Singapore surveillance aircraft presence/position/detection capabilities are conjecture. I don’t see how this line of speculation can rule anything in or out.
2) My point #9 addresses fuel feasibility.
@Oleksandr asked me to submit this comment. (He has had problems getting the site to load.)
I would like to share my MH370 path modeling software on GitHub. The software allows for numerical path integration based on a user-defined sequence of maneuvers and flight parameters, with the possibility of automated optimization of those against BTO and BFO data for user-selected pings. A fuel consumption model is included. All the essential setup parameters are to be specified by a user in ASCII files – there is no need to recompile the source code. Auxiliaries include an extractor of meteorology, and a BTO & BFO calculator. A stand-alone executable compiled for Win64 is included in the first release pack, along with the meteorological and magnetic declination datasets needed to run the model, examples of model setup, and a brief user’s manual. Path modelers and software developers are welcome to contribute.
https://github.com/TheLastDinosaur2025/MH370Solver/tags
@All – Not to change focus from above discussion but interesting article here on Air India flight from yesterday.
https://www.nytimes.com/2025/10/06/world/asia/air-india-boeing-planes.html
@ST
Of course, Boeing needs to help determine why the RAT deployed in that recent case and determine if any fixes are needed.
To some extent this is exaggeration. Its really all about the Air India #171 tragic crash in June and India’s pilot union defense of the pilots in that case. As we know, the preliminary AAIB report indicated the prelim 171 accident cause was that the pilot(s) manually switched off the engines. In both AI171 and MH370, we see a defensive argument that we cannot put blame on pilot hijacking when it could be autonomous aircraft behavior, the so-called remarkable failure.
It is a bit upsetting to me to see NYTimes dumping on Boeing when the 171 accident was probably India’s fault. On the other hand, Americans have forgotten about AI171, as we heard it was not Boeing’s fault. We (USA) need to hear that some in India blame USA/western media/Boeing/poss GE for the 171 accident and to them it is unfair to blame the pilot(s). Which to me is classic case of denial in India, much like MH370 in Malaysia.
@Paul Smithson
I don’t agree with your statement 7. For the far majority of measurement instances DSTG gave a (pragmatic) error budget based on fig. 5.5. The number of excursions like shown in fig. 5.4 seems limited (see fig. 5.5). So, yes, one should use the BFOs with prudence. No, IMO we should not totally discard the BFOs in the analysis. We don’t have this luxury with the sparse data set at hand.
Also, we don’t know for sure if the large excursion seen fig. 5.4 is caused by oscillator drift. The regional dependency suggests something else.
I was thinking in the direction of perhaps a steep gradient in the constant pressure altitude in cruise due to a local weather system, leading to a (geometrical) vertical movement. As DSTG only considered in-flight measurements, excluding climb and descent phase, the question would be if they properly included momentary vertical displacement during cruise or not.
@Niels,
If you’re looking for a reason to explain any vertical flight path changes, please consider how an autopilot off, ghost flight, flying phugoids could explain the scenario.
Hi Niels – I appreciate the comment. Here is some further elaboration on my reasons for suspicion on BFO. I honestly do think that misplaced confidence in it is probably the single biggest reason for an unsuccessful search to date.
1. As discussed previously, I think we agree that the documented BFO aberrations appear to be rare. However, if/when they do occur, they are large enough to completely distort the inferred path solution. That makes them UNLIKELY + HIGH RISK/IMPACT – so worth paying attention to.
2. A close reading of DTSG reveals that they EXPECT a shift of unknown magnitude/direction (oscillator retrace error) from the re-powering, even if their was dramatic change in ambient temp. So if BFO is “well behaved” we could anticipate non-trended residuals, but not necessarily zero mean. That alone makes it difficult to model – and DSTGs rather crude solution was to “guesstimate” a more cautious sigma of 7Hz instead of 3Hz. You have also pointed out above the difficulty with discerning an implicit bias shift.
3. I think it is pretty plain that the best (BTO) path solutions DO have trended residuals.
4. Even at 34S you see a larger than expected residual at 1840. UGIB resorts to suggesting a coincidental manoeuvre to resolve this error – despite the fact that the multiple BFOs around 1840 are extraordinarily consistent, possibly exhibiting patterned “trig error”. sk999 has documented that BFOs as steady as these are unlikely if there was a non-zero rate of climb. It seems to me more plausible that the “BFO error” is real.
5. Another difficulty (and related to #3) is the difference between BTO-optimised paths (40S) and BFO-optimised (34-35S). These are fundamentally divergent – as ATSB/DSTG showed early on (ref slow/curved/northern data-optimised paths vs auto-pilot paths).
6. Plotting BTO fit as RMS residuals obscures the difference in relative probability of LNAV (or TT) paths from Arc 2 that end around 34S. The BTO fit is lousy. BTO paths with track angle ~190 to 40S are about TEN TIMES more likely than track angle ~180 to 34S using relative probabilities computed from log-likelihood.
7. In view of #5 and #6 above, you basically have a choice of betting on BTO fidelity or BFO fidelity. Optimising for both simply falls between two stools. Compared to the “BTO-only” discrete peaks, their final pdf shifted north and was diluted/blurred by optimising for BTO+BFO combined. As a result, 39.5S – which was close to peak probability by BTO, is right out at the 5% tail of the final pdf. By the bye, you’ll also notice that the UGIB priority zone is also right out at the [northern] 5% tail of DSTG’s pdf, while the areas most recently searched are out way beyond even that (1%?).
8. Finally – reliance on BFO fidelity is beating a dead horse. We know where that line of analysis takes us (+/-34S) and the plane very obviously isn’t there. The zone has been searched out to the north, the south, equivalent of 70NM beyond the arc, and data holidays have been filled. Persisting with further search around that area with such poor prospect of success doesn’t look wise.
9. Pre-search, I think there was a more reasonable case placing greater weight on BFO. In the light of null search results – as well as for all of the other reasons, I think we should be a) more suspicious of it b) consider the growing likelihood that BFO is in error due to bias shift, drift or both.
@Paul Smithson
Thank you, Paul, for your quick follow-up.
Yes, indeed a possible “jump” in the frequency bias after cold restart generates a major challenge. One can possibly estimate it by looking at mean BFOR for different scenarios (assuming a fixed value near 150 Hz), I believe that’s for example what @sk999 has done in his March 19 2023 paper (“On the future search zone for MH370”). By the way ,he worked there with a SD of 2.8 Hz for the BFOR, including a random and a drift component. DSTG used 4.3 Hz (in flight). This latter value is what I use in my analysis, in combination with an arbitrary fixed value for the unknown possible “jump”, which I refer to as the “frequency bias offset”. I would also like to remind you of the validation that Inmarsat did early on based on the Amsterdam flight, resulting in an estimated +-7 Hz error margin for the BFORs. Looking at the number of measurements this estimate would probably account for more than 1 sigma, perhaps in the direction of 2 sigma. In UGIB (appendix F) there is a more detailed analysis including an attempt to characterize the drift component better based on MH371 flight data, including an Alan Variance plot (fig. F-5). For longer time segments the drift can be considerable, see for example fig. F-1, but for 4.5 hours segment you would not expect more than around 5 Hz drift. Actually, this is consistent with the conclusions from studying technical documentation (datasheets) of typical OXCOs that @alsm found and that we discussed in the past. So the 4.3 Hz sigma proposed by DSTG looks pretty OK to me and the overall picture consistent, except perhaps that the 2.8 Hz used by @sk999 looks a bit tight. And indeed, prudence is needed because of the rare unexplained larger BFO swings with unknown origin.
Regarding your point 4: I also tend to read the data in the direction of an early turn. Although I don’t see why that would imply 40S rather than 34S. We don’t know what happened between 18:40 and 19:41.
Could you please share the BFO residuals at 18:40 for your preferred path?
Regarding your point 6: If I remember well the “lousy” BTO fit mainly is caused by the arc 5 value being rather far off (between 2 and 3 sigma). It would perhaps help to estimate how likely or unlikely one such value would be given that we consider 5 measured values without and 10 measured values including the BFOs.
Your point 8: I don’t agree that the S33-36 zone has been searched enough and that all data holidays have been filled in. For a long time I have been interested in the area around S35.5 close to the arc, as it relates to the “Pleiades” hypothesis, and to my best knowledge this area has not been re-searched, at least not by OI.
Regarding point 3: could you perhaps share the “trended BFO residuals” for you best estimate path? I would like to take a closer look.
Your point 7: I don’t think it’s sensible to choose between either what is indicated by the BTO or the BFO data. Given the small number of samples and the estimated error characteristics there can be “normal” statistical effects that cause some deviations between the two. Actually, you can simulate this: try to reconstruct employing an artificial set of BTO values based on a certain assumed flight path and add random errors several times. Then do your BTO-only analysis for this several instances.
@Niels and Paul Smithson
I am happy to see that you are beginning to question the quality of some of the BFO values. However, I agree with Niels that we cannot disregard BFO entirely since we do not have sufficient independent data. My initial reaction to potential problems with BFO was to use a very large random error of plus/minus 15 Hz. That covers most of the simple technical reasons for systematic BFO errors such as a Helios-like incident. On the other hand, it would not cover things like hacking or other deliberate interference with the data.
Using a large error like 15 Hz is sufficient to exclude solutions far south, and it only shifts the consensus solution a few tens of kilometers. On the other hand, it opens a new type of solution near Christmas Island.
In a situation like this, the best option is to use independent types of information to decide which solution is best. That means any information besides BTO and BFO from handshakes. There is plenty of independent information, so the task is far from impossible. Here is an (incomplete) list of options:
1) Contrails
2) Seismic signals
3) Received S/N ratios
4) Statistical distribution of BFO values recorded during attempted phone calls
5) Beaching pattern and isotope ratios in biofouling of debris
6) Potential atmospheric signs of an impact near the seventh arc
7) Eye-witness reports
Recently I have come up with more options. I am going to present these later.
For the moment the seven options above should be sufficient.
They all point to a solution much further north. The only problem is that very few people believe in results derived from them.
@viking – without assuming any ROD throughout a BFO 15hz error budget easily accommodates path models ending at 39-40S.
@Niels, paul smithson, all
As I understand it, the BFO is not ROD compensated in the design of the SDU, therefore, I assume that the BFO should be very sensitive to ‘instantaneous ROD’. In the past, it seems to me, that all of the BFO analysis (with few exceptions) has been effectively based on the implicit assumption that the aircraft was in stable level flight “at the instant of the BFO records” (except at the 7th arc). If incorrect, disregard the following.
I find it hard to accept that assumption. Anyone who has ever flown in anything knows that there is always ‘some turbulence’ (however slight, perhaps un-noticeable to most pax). Anyone who has ever sat at a window in cattle class (as I always have done) looking at the wingtip for a while, will notice that every now and again, even in what seems like a ‘smooth as silk ride’, the wingtip will ‘wiggle’ up and down a few inches, perhaps a foot or more, perhaps with a frequency of 2hz, even when the cup of coffee sitting on your tray table has no ripples in it at all. In other words, the atmosphere is never entirely ‘static’, some turbulence is always there, no matter how slight, and the aircraft structure always responds to it, no matter how slight the atmospherically imposed impulse is.
So, over time in any given flight, (particularly when travelling north to south over the Indian Ocean, crossing not just latitudinal airmass boundaries, but even jet streams), there must be a significant amount of turbulence that the aircraft flies through. Now if you were to look at the FDR trace for vertical accelerations for a whole flight (of many hours) you would definitely see that. But, the issue will be not only be the actual intensity but also the time duration of ‘each instance’ of a turbulence encounter that was flown through. Now, I assume that the actual ‘sensitivity’ of the accelerometer itself must be very good (which it logically must be, since ‘high sensitivity’ with a very high data rate is necessary within the ADIRU for it to function as it does) but the actual FDR sampling of that data may have a very course threshold level (say point one ‘g’) which obviously will not record the possible hundreds of point zero five ‘g’ hits, and, the ‘rate’ of sampling of that acceleration data may be perhaps only at one second at best. So, what would look like a ‘smooth as silk ride’ in the FDR data, is unlikely to bear any resemblance to ‘the real truth’.
So, I re-emphasise “instantaneous ROD”, because the ‘rate of change’ of ROD (ie the instantaneous change) is the issue, because the BFO was recorded in a Tx burst of a few microseconds.
So, what I propose is this.
Since we obviously don’t have any useable vertical acceleration data, (even if we had FDR data from other 777’s that data is highly unlikely to be of sufficient sensitivity or at a high enough sampling rate) we have to model some hypothetical ROD conditions.
Therefore, would it be worth re-running the ‘existing flight path models’ (which I am assuming are zero ROD based) with the deliberate insertion of increments of say plus and minus 0.5m/sec to say plus and minus 10.0m/sec (approx. ROD of +/- 2,000 feet per minute), and see if that leads to any ‘new insights’ ?
For instance, might it be possible to detect that (hypothetically) that 9M-MRO was (say) very likely at ROD +8m/sec at 19:41, and at ROD -2m/sec at 20:41, (and similarly at the other arcs) and then, (by cross checking with the metrological model data we have) determine a ‘likely area on the relevant arc’ where that sort of ROD might have occurred, and thus, for all arcs, see if that develops a different plausible flight path ?
I know that it’s a long shot (and then some), but at this stage off the game, I think we have to throw the computers at everything that is even remotely plausible.
Thoughts ?
@ventus45 wrote ‘ because the BFO was recorded in a Tx burst of a few microseconds’
The 1200bps R-channel bursts from the aircraft that are the response to the log-on interrogation from the GES have a duration 460ms (milliseconds, 500ms slot less guard time). If the BFO is derived from the log-on confirm response when the aircraft transmits on a 600bps R-channel then the burst duration is 960ms (1 second slot, same 40ms guard time).
Not sure it has ever been nailed down from which segment of the R-ch burst, or across the entire burst, that BFO is derived in the GES channel unit.
@Paul Smithson
If you do the combined statistics carefully, anything south of 38S is impossible with 15 Hz BFO uncertainty (and normal BTO uncertainty). This conclusion is further corroborated by the combination of fuel limitations and that Singapore did not observe MH370 on their airborne radar.
Actually, the early arrival of one piece of debris in South Africa is itself a proof, completely independent of the rest.
@ventus45
You have some good general points on noise, but if your cup does not show any surface waves you can be relatively sure BFO is not affected. This is corroborated by Don Thompson’s remarks on sampling duration.
@All
The probability of turbulence varies a lot across the globe and with weather patterns. Generally it is worst near jet streams and when passing thunder storms or abrupt mountain ridges.
If MH370 flew so far south as 40S, I would have expected increased turbulence during the last 1.5 hours of the flight. That is in disagreement with less BFO noise during the last attempted phone call.
@Ventus45,
I think that’s a good idea. I would also like to know how random vertical movements could affect the arc positions. Autopilot failed, ghost flight scenario, resulting in phugoid motion may explain any vertical motion.
@Tim: In the history of aviation, has there ever been a case in which a plane with no pilot inputs and no augmented flight control had experienced phugoid motion and yet continued to fly for hours?
@Ventus45
I was thinking to explore possible influence of geometrical gradient in pressure altitude a bit more. I think this can be extracted from the recorded weather data file.
I don’t expect to get much out of it, as the typical BFORs I observe can well be explained by the known error sources.
Away from laptop for now. But if I remember correctly, you need ROD of a few hundred fpm to shift BFO by a few Hz. I can’t see that either of turbulence or delta height to maintain pressure altitude could explain major BFO errors. I also recall looking at the difference in height using spherical vs ellipsoid earth and that was only a fraction of BFO sigma.
It is incorrect to say that 38-40S involves errors >15Hz. The BFO 1941 to 0011 range from roughly -8 to +6Hz. So around 2 sigma, using narrow 3Hz number, or around 1 sigma using DSTG’S more conservative 7Hz.
@Viking
“less BFO noise during the last attempted phone call.”
I would have said the 2314 has more noise, compared to 1841.
@Victor @all
Here’s an IGARI technical tidbit I think I can say from public info (FlightAware), that I do not recall anyone mentioning.
If we look at MH370 flights day before and day after accident, those flights were at FL370 before IGARI. That perhaps gives a little different meaning to maintaining FL350 calling twice on that detail. Additionally we could look at FlightAware every day but the FA ADSB data cuts off approx 20-25 mins after takeoff. The only obvious for me so far is some days are FL370. I have no idea why for sure, but the FL350 could be related to PAX/loading.
@TBill,
Yes the optimum initial cruise level for its weight is FL370. Due to stronger head winds at higher levels that night, the flight plan had the first cruising level as low as FL330. With a step to FL350 around the HCM boundary if I remember correctly. Fariq requested FL350 in his preflight call to ATC.
It is this slight change from the normal level that I believe prompted Zaharie to make the two 350 calls. He had expected climb to the more usual 370…Perhaps he hadn’t heard(on the pa or something), or remembered Fariq requested 350.
Suspect the Take Off Weight of the aircraft of each of the previous flights had more to do with the selected initial cruise altitude than anything else.
Malaysia has always rejected claims that Zaharie loaded extra fuel for nefarious purposes. The pre-flight top-up of cockpit oxygen is IMO ‘borderline’.
@Barry Carlson
Barry, no one needs to rely on “Malaysia” having ‘rejected claims that Zaharie loaded extra fuel for nefarious purposes’. Any independent, objective assessment of the fuel load, with due regard to destination weather forecasts and MAS operational policies, would come to the same conclusion.
What should be noted with regards to the fuel ordered, had the Captain wanted to carry more fuel, he could have quite legitimately nominated either of Shanghai (ZSSS) or Shanghai-Pudong (ZSPD) as his alternate airport; both are listed in the MAS operations manual as authorised alternates for Beijing (ZBAA). Nominating either of those alternates would have allowed him to quite legitimately upload about an extra hour of fuel above the quantity that he ordered.
The very simple fact of the matter with regards to fuel is that MH370 carried the MINIMUM allowable fuel for its route under MAS operational policies, given the forecast weather for Beijing and the nearest suitable alternate airports.
@Barry Carlson
RE: “The pre-flight top-up of cockpit oxygen is IMO ‘borderline’.”
What do you mean by that statement?
@Mick Gilbert, @Barry Carlson: That the captain took on extra fuel is one of the many false beliefs about MH370 that refuse to die.
That the path of MH370 can be reconstructed using historical WSPR data is another false belief, although it has been many months since hearing that craziness out of Liverpool. Maybe Simon Maskell has quietly come to his senses.
Does anyone know if OI is actually going to restart the search (in November – as we have been led to expect) ?
Concerning a resumption of the MH370 search.
The closing paragraph of the ‘SUMMARY OF MH370 SEARCH OPERATIONS REPORT’ disseminated in April 2025 stated that “Search operations will recommence when favorable weather conditions return in November 2025“.
That, of course, will be dependent on favourable weather conditions early in the austral summer, the availability of a vessel and deployable resources. The Armada fleet is presently well utilised in European, US, and Guyanan continental shelf waters (all A78s plus A86-01 and -03). A78-06 was recently pictured with the seafixing bases for the three AUV shelter/garages still in place, however, it’s very unlikely that it would reposition from offshore Netherlands to the sIO.
More likely is that one of the four new A86 vessels still positioned in SE Asia will be mobilised for the task. A86-02 is presently underway for Taiwan where deep seafloor might be exploited for a general ‘shakedown’ and calibration trials of vessel-AUV subsea positioning.
@MH370 Fuel
I would certainly take a more nuanced approach on MH370 fuel loading. The pilot did add some extra fuel, the question is- was that necessary?
FlightAware shows that MH370 was running 8 minutes early with a posted arrival time as early 22:15. The day before flight was on-time departure and arrived 22:34. Some MH370 flights depart a little early, so that is not particularly unusual.
In any case, there was enough extra fuel to fly 2-hours more to 00:19 not including descent/glide allowance say +20-30 minutes potential.
@TBill: Airline captains such as @Andrew have looked at the flight plan and weather and concluded that the fuel loading was nothing out of the ordinary. Why do you doubt the professionals?
@TBill
RE: “The pilot did add some extra fuel…”
The final load sheet sent to the aircraft before departure shows the take-off fuel was 49,100 kg. That figure is consistent with the operational flight plan (OFP) issued to the crew at dispatch, which shows the minimum sector fuel was 49,100 kg.
The fuel analysis section of the OFP shows the following:
Trip fuel: 37,200
Contingency fuel (3% ZSPD): 1,200
Alternate (ZSJN): 4,800
Final reserve: 2,900
Company fuel: 3,000
Total: 49,100
That figure is calculated by the airline’s dispatch staff to meet both the regulatory and company requirements. It is the MINIMUM fuel that must be carried to satisfy those requirements.
If the aircraft commander chooses to load extra fuel for any reason, it is annotated on the OFP as “SURPL” (ie surplus) fuel and the take-off fuel adjusted accordingly. There is no such annotation on the OFP and the take-off fuel is shown as 49,100 kg, the same as the minimum sector fuel. A taxi allowance of 500 kg is added to that figure to provide the requested block fuel of 49,600 kg.
In summary, the OFP shows the commander did not change the fuel load and elected to depart with the minimum sector fuel required by the company.
Some commentators have queried the alternate selection and the company fuel, claiming the fuel requirements were excessive. I don’t believe that’s the case, for the following reasons:
1. The destination alternate aerodrome shown on the OFP is Jinan (ZSJN). Tianjin (ZBTJ) is often used as the destination alternate for Beijing, because it is closer and requires less fuel. However, on the night in question, the weather forecast for Tianjin shows the aerodrome was not suitable for use as a planned destination alternate because of reduced visibility in light snow and rain. Consequently, it seems Jinan was selected as the next closest destination alternate that had suitable weather; the Shanghai aerodromes are more distant and would have required more fuel. In my view, the alternate selection was entirely reasonable given the forecast weather conditions at the closer alternate.
2. The company fuel requirement on the OFP was 3,000 kg, which provides for about 30 minutes of ‘extra’ fuel. Company fuel is described in the MAS Operations Manual as “fuel to cover specific operational requirements.” The operational requirements are not specified in this case, but Beijing is notorious for extended ATC delays. In my experience, airlines operating into China normally add extra fuel to cater for unexpected delays and to mitigate the risk of a diversion. I don’t believe the company requirement for 30 minutes of extra fuel is unreasonable. Furthermore it was a COMPANY requirement, not fuel that was added at the whim of the commander.
Another myth that keeps rearing its head is the notion that ‘extra’ crew oxygen was loaded. The crew oxygen system was replenished to 1800 psi, after the pressure was noted to have been 1120 psi during the night stop check completed before departure. Some commentators believe that replenishment was unnecessary because the MEL only requires 310 psi for dispatch under the relevant conditions.
That notion is wrong. The MEL figures are the minimum pressure required to meet the regulatory requirements for the carriage of crew oxygen; however, they do not necessarily reflect normal airline practice. Indeed, the MH370 SIR notes: “It has been the practice of the airline to service the oxygen system whenever time permits, even if the pressure is above the minimum required for despatch.”
The limit used by MAS engineering to trigger crew oxygen system replenishment isn’t specified. However, the leaked police documents show that replenishment was conducted on three previous occasions, after the pressure had dropped to 1150 psi, 1180 psi and 1200 psi respectively. It seems that MAS engineering practice was to replenish the system after it dropped to or below 1200 psi, when time permitted.
I don’t believe there is anything unusual about that practice. At the airline where I worked, our standard dispatch pressure was 1600 psi and the systems were normally kept at or above that pressure. However, 1200 psi was sufficient for us to depart our home base for any destination on the network.
@Andrew
Thank you I appreciate it. But I am trying to use investigative logic, and as you say there was a decision to have a bit further alt airport, so perhaps MH370 had more fuel than the average MH370 flight, which is defensible from an aviation viewpoint, but at the same it happens to be enough fuel to get all the way to Arc7, in my view, to approximately where Arc7 intersects the early sim path to NZPG.
@TBill
There was no “early sim path to NZPG”. That conjectural end point only arises if you extrapolate a track that joins 10N and 45S directly. It only arises if you take the ultra-simplistic “string around the map pins” view of the sim data. We know from a thorough examination of all of the sim data that there was not a direct track from 10N to 45S, rather that, from 10N, the sim aircraft was flown south-eastward toward Banda Aceh.
You would be hard pressed to explain the “investigative” bona fides of eschewing an objective technical expert explanation in order to pursue a theory largely based on a very basic examination of the sim data. That said, you are most certainly not alone in that sort of endeavour.
A week or so ago, some of you requested further information regarding my interpretation of path model probabilities from Arc 2, their implied BFO errors, and their consistency (or not) with a single-turn solution.
I have written this up, showing:
1) relative probability distribution (by BTO, BFO) of path solutions ending between 31.3S and 41.2S, corresponding with track angles ranging 175-195.
2) an examination and discussion of the BFO error pattern
3) a demonstration that Arc 2 path models with track angle close to 190 (ending 39.x South) correspond closely with a single-turn solution modelled from aircraft’s last known position.
Data tables are provided at the end for anyone who wants to look into the detail or replicate the results.
The write-up is here https://tinyurl.com/y7ccj78y
Critical feedback and commentary warmly welcome!
@Paul Smithson,
I have four comments on your most recent work:
1. The standard deviation of the BTORs is well established to be close to 29 microseconds. You seem to ignore this because you assign the highest probability to the lowest RMS values (20 microseconds in your examples). Your formula predicts the highest probability when the BTO residuals are all zero, having both a zero mean and a zero RMS. This is nonsensical. We know the expected value of the mean of the BFORs is zero. We also know the expected value of the standard deviation of the BTORs is 29 microseconds. The farther the computed mean BTOR is from zero, the lower is its probability. Similarly, the farther the computed RMS BTOR is from its expected value of 29 microseconds, the lower is its probability. We need to consider both BTOR statistical probabilities (mean and standard deviation), which your formula does not do. Matching both the expected value of 0 microseconds for the BTOR mean and 29 microseconds for the RMS BTOR should produce the highest route probability. RMS values of 20 microseconds have lower probability, being roughly the same probability as at 38 microseconds. The mis-match of the RMS BTOR with 29 microseconds (being either higher or lower) reduces the probability of the route being correct.
2. Using a standard deviation of 7 Hz for the BFORs is beyond “conservative”. It is inconsistent with the data from nearly 100% of the comparisons done with previous flights. Inmarsat chose a peak error of 7 Hz based on their analyses, and I agree with them.
3. Figure 43 in UGIB (2020) clearly shows that the 19:41:03 BFOR appears to be an outlier, as you have also concluded. That means we should not count on it being an effective route discriminator if we assume there were no maneuvers after 19:41 until fuel exhaustion.
4. There is an assumption for these BFOR statistics that there is no ongoing climb or descent. We don’t yet know if this assumption is valid at 19:41, but I don’t believe it is. In UGIB (2020) I postulated that a climb to cruise altitude was almost completed then, with a (decreasing) ROC of about +400 fpm occurring at that handshake at 19:41:03 and the climb being completed circa 19:41:21. The ROC is automatically decreased by the FMC so that the aircraft reaches the desired cruise altitude with zero ROC and no overshoot in altitude. Therefore, depending on the relative timing, any ROC from ~3,000 to zero is possible. So, in UGIB we postulated the larger BFOR at 19:41:03 was due to a small ROC shortly before the end of the climb. There are other possible causes, but they involve turns or ground speed changes which seem less likely to me.
Your proposed route to 39S has three strikes against it:
1. The BTORs are too small.
2. The BFORs are too large.
3. There is not enough fuel.
@Dr B.
Re fuel consumption, worth a look?
https://docs.google.com/document/d/1RFhbHJIP3stMlc0E813750TPdGEHYYk7/edit?usp=sharing&ouid=105452605762640315384&rtpof=true&sd=true
@David F. Intriguing! But I think you will find that the effect size of small changes to assumed aircraft gross weight on fuel consumption is too small to be of consequence.
@DrB
I agree completely with your comments 2 and 3, and mostly with 4. However, I disagree with number 1. The reason is that if you use some kind of systematic optimization to find the solution (e.g. least square fitting), and the experimental error distribution is Gaussian, you will end up with a smaller RMS value than expected. For problems similar to MH370 the RMS value becomes roughly 50% smaller than expected. This does not mean that the solution is less probable. It just means that the routine is doing a good job – perhaps a bit better than naively expected. On the contrary, an experimental RMS below the expected value means a red alarm.
Concerning your arguments against 39S, I agree with your points 2 and 3, but not with number 1 (for the same reason as above). Instead I would include the non-observation by the Singapore radar airplane. That combined with the fuel shortage and poor BFO fit (and a few more details) totally excludes 39S.
@Viking,
Your assumption about the figure of merit I use in route fitting is incorrect. I am not trying to minimize the RMS BTOR. Instead, I am trying to maximize the overall probability of matching the included statistics. One of these statistics is the standard deviation of the BTORs. So, my fits try to make the observed standard deviation match the expected value (not to make it as small as possible). My method avoids the potential bias you are describing. In addition, I correct for the inherent bias due to having a small number of samples. This results in underestimating the standard deviation if uncorrected.
@David F,
The engine fuel mass flow tables do not depend on latitude or track bearing. I don’t think this is simply because those effects are small and complex to incorporate, but because they are nonexistent.
Using an aircraft-centric reference point, I don’t think the aircraft fuel mass flow will depend on its flight direction or latitude. The weight of the aircraft, the lift produced by its airspeed, and the thrust to overcome the drag each depend in the same way (linearly) on gravity, so small changes in gravity should not affect the required fuel mass flow of the engines. In addition, the fuel mass flow tables are calibrated at each cruise altitude. So, I don’t think the small variation in gravity with altitude produces any change in fuel mass flow that is not already built into the fuel flow tables.
@DrB. You state “The BTORs are too small”. But how small is “too small”?
Could you provide a quantitative response to:-
1) what is the “expected” RMS BTOR across 5 values?
2) what is the probability of a path having RMS BTOR of 20.8 [the minimum RMS in the data I presented, at track 192]?
Thanks in advance.
@Paul Smithson,
The impact of a small number of BTOR samples is illustrated on page 153 in the Appendix in UGIB(2020):
“The expected sample means and sample standard deviations for each statistic were determined by injecting random BTORs and BFORs (instead of using the actual route-fitting residuals) into the fitting program and recording the values of the output statistics. We used 100,000 trials with random samples derived from the population statistics, which are a mean of zero and a standard deviation of 29 μs for the BTORs . . .”
“For example, the sample BTOR standard deviation is less than the population standard deviation of 29 μs, because of the limited number of samples. 100,000 trial sets had an average value of 27.30 Hz. The observed standard deviation of the sample BTOR standard deviation was expected to be 29 ∙ SQRT(1 – 9 ∙ π/32) = 9.90 μs, and the trials gave an average of 9.90 μs. The expected value of the mean BTOR is 0 μs, and the average of the random trials was -0.07 μs. The standard deviation of the mean BTOR is expected to be 29/SQRT(5) = 12.97 μs, and the observed value was 12.98 μs.”
So, the observed standard deviation of the five BTORs should be 27.3 microseconds in order to best match the known BTOR statistics (and the uncertainty in this calculated value is 9.9 microseconds at one sigma). Your “best-fit” (i.e., smallest) value is 20.8 microseconds, which is 6.5 microseconds lower. The standard deviation of the computed standard deviation is 9.90 microseconds for one data set, so your lowest value is 6.5 / 9.9 = 0.66 sigma too low. How likely is that? The relative probability that an observed statistic will be at -0.66 sigma (Z = -0.66) is 80% of the probability it will be at zero sigma (27.3). So, the probability that your 20.8 microseconds result for one data set is solely due to BTO noise while using the correct route, and not due to systematic route errors, is 80% of the probability that a route with an observed standard deviation of 27.3 is the correct route.
In an absolute sense, the uncertainty in this relative probability estimate prevents us from saying that there are definitely NO systematic errors with 27.3 or that there ARE definitely route errors with 20.8. With only one data set we cannot separate systematic route errors from statistical noise in the approximate range of 27.3 +/- 9.9 = 17.4 to 37.2 microseconds. Outside that range it is 63% likely that systematic route errors are affecting the BTORs. Within that range there is only a 37% chance that systematic route errors exist. Thus, one cannot prove that either result is the correct route, but one can say that the 20.8 route is only 80% as likely to be correct as the 27.3 route.
Thanks for your response, Dr B. I have emailed you with my workings that arrived at the same conclusion. In short, the BTOR at 190 are not “too small” in the sense that the SD is too small. On the contrary, it is pretty close to expectation.
On the other hand, the SD of BTORs of path 180 to 34.3 is somewhat larger than expectation.
By my calculation, the two tail probability of SD at 190 is roughly 2.8 times greater than that at 180.
I do understand your point about overfitting – that the SD [RMS] of BTOR should not be suspiciously small. As we have seen, this is not the case.
Meanwhile, subject to the proviso on overfitting, a model that has smaller errors is better than one that produces larger errors. The log likelihood method correctly ascribes maximum probability to an error of zero, because that’s at the centre of the bell curve. Unlike comparing RMS_BTOR, it offers the benefit of being able to make an empirical comparison of the probability of alternative models.
I thank you for pointing out that goodness of fit should not be used on its own – at least not without checking for overfitting. That does not make log likelihood – and the relative probability derived from it – an invalid metric for solution analysis.
@DrB
Thanks for clarifying your procedure. We are now much closer to agreement. There is just one (important) detail missing. I do not think your proof for the BTO standard deviation can be inverted. There are 3 reasons for that:
1) Thew data sample is too small
2) Not certain the error-distribution is Gaussian (I suspect it is not)
3) An inversion of the argument is totally compromised if the FMT took place within the relevant time interval.
Concerning point 3, that was exactly my reason for making Figure 2 in my paper. That figure made me conclude that a FMT in the form of a U-turn was highly likely.
A Chinese infrared satellite picture seems to confirm it.
@Paul Smithson
If the FMT was close to a U-turn, it was most likely made deliberately to make everybody think the airplane flew to around 39S. If the BTO values before 19:50 are interpreted without a FMT they do indeed fit such a solution. However, all the previous arguments speak against it.
This is further corroborated by the exact coincidence between the (presumed) U-turn and the extremum of the Eigen-movement of the Inmarsat satellite. That coincidence creates a nasty mathematical entanglement, making the problem extremely difficult to solve.
In other words, I think Shah got help to design the flight from someone with knowledge of satellite communication and advanced physics.
@Viking. There is no evidence that Shah was aware of the “handshakes”. Nor that he was aware of the potential to extract range data from Inmarsat proprietary satellite metadata. The idea that he was a) laying a deliberate bread trail b) using advanced physics to frustrate subsequent analysis of the metadata is fanciful.
Paul Smithson,
I did some spot checking of your Annex 1 data tables (specifically initial track angles 175 and 189) and more or less reproduce your results. The mach values are significantly different between the two track angles, which means that the track 189 route will require more fuel than the 175 route. Assuming fuel exhaustion at 00:17:30, I estimate the extra fuel required starting at 19:41:05 is about 2 tons, and if one extrapolates back to the 1st arc, the extra fuel required is closer to 3 tons. Given the estimated fuel remaining at the time of the 1st arc, the 175 route is feasible but the 189 route is not. The way to compensate is to fly at a higher altitude – in this way I estimate that the 189 route could be flown if one increases the altitude from FL350 to FL400. However, even though the mach number stays nearly the same, the lower temperature means that the air speed and thus ground speed are lower. Tweaking the starting latitude and mach to get the best fit, the BTO rms increases from about 20 microsec at FL350 to 32 microsec at FL400. Thus, when accounting for fuel burn, the BTO rms is relatively insensitive to final latitude.
@Paul Smithson
You are of cause allowed to have the opinion that a solution with a U-turn is fanciful, but I would rather call it necessary thinking out of the box. The facts on the seabed are that we have never found any debris.
To be more scientific, your solution using exclusively BTO is roughly 600 km from the consensus solution using both BTO and BFO. The statistical uncertainty on the center of the consensus solution is roughly plus minus 100 km. That means that your solution is 6 sigma from the consensus solution. I used only crude values here so one can discuss the decimals, but it is still obvious that there is something wrong when you are roughly 6 sigma away.
I am not saying that you are doing anything wrong mathematically, but clearly the information from the two subgroups of the complete data set disagree significantly with each other.
My conclusion was (and still is) that only data from after 19:50 should be included. That way I get the southern solution less than 0.5 sigma from the consensus solution even when I assume 15 Hz uncertainty on the BFO. Clearly that is an insignificant deviation.
Therefore I conclude that something significant happened near 19:50. Most likely it was the FMT.
@Viking.
1) I didn’t say that “a solution with a U-turn” was fanciful. Rather that it was fanciful to suggest/imagine that Shah was choosing tracks/manoeuvres to obfuscate analysis of the satellite metadata.
2) It is incorrect to suppose that 34S sits at the peak of a probability distribution based on BTO+BFO. According to DSTG, it sits roughly on the 5% northern tail of the BTO+BFO pdf. I don’t know where you came up with a “sigma” of 100km. If you look at the analysis referred yesterday, it is plain where peak probability zones are based on BTO (~40S) and BFO (~36.5, depending on which/how many values you use).
3) I don’t know why you prefer to ignore where the aircraft was at 1828, and therefore where it MOST PROBABLY was at 1941 (barring exceptional reasons for theorising otherwise).
4) I am baffled by the method by which you have arrived at a probable terminus location.
@Paul Smithson
Please read my paper. It contains answers to most of your questions.
Concerning connections between a U-turn and attempts to promote entanglement and confuse the investigators, it is simply the only logic reason I can come up with for turning the satellite communication back on, receiving (but not answering) an attempted phone call, and slightly more than one hour later (just after the first automatic handshake) making a U-turn.
It simply could not be made better. The timing is precise to better than one minute, matching the top of the Eigen-movement curve of the satellite. I do not believe that is luck or a random coincidence.
@Paul Smithson,
1. You said: “By my calculation, the two tail probability of SD at 190 is roughly 2.8 times greater than that at 180.”
The two-tailed integrated PDF is not applicable here. That tells you the probability that one result would have a value of SD greater than or equal to the threshold value (on both sides). That’s not the problem I addressed, which is what you requested: “what is the probability of a path having RMS BTOR of 20.8 [the minimum RMS in the data I presented, at track 192]”. The answer to your question is the relative probability of getting a value of 0.66 sigma compared to getting a value of zero sigma. As I said, that is about 80%, not a factor of 2.8 (i.e., 280%).
2. Now, you have raised a second, new question, which is what is the relative probability of those two of your fits? The 180 degree route has a BTOR standard deviation (SD) of 38.0 microseconds (the UGIB route is 34.3), and your 190 degree route has a BTOR SD of 23.8. Only considering your BTOR SDs, that relative probability is 60%. Of course, there are many other statistics which must be considered in addition to the BTOR SD, and the overall probability for all statistics is what really counts.
3. You also said: “Meanwhile, subject to the proviso on overfitting, a model that has smaller errors is better than one that produces larger errors.”
Smaller standard deviations of BTORs and BFORs are not “better” when they are smaller than their expected values. Those have a smaller probability of being consistent with the correct route, and they are therefore “worse” not better.
4. You said: “It is incorrect to suppose that 34S sits at the peak of a probability distribution based on BTO+BFO. According to DSTG, it sits roughly on the 5% northern tail of the BTO+BFO pdf.”
The DSTG made an assumption that is unlikely to be true, in my opinion, and they ignored one critical factor (fuel). They assumed the aircraft was always, after 18:02, at cruise altitude and speed. I think MH370 was flown low and slow for a period of time before 19:41 (as I illustrated in UGIB). I expect that if their analysis was limited to 19:41 to 00:11 they would have gotten a PDF which was shifted NE along Arc 6. The problem was that DSTG did not allow for any loiter, such as a holding pattern or a descent and ascent. Such a delaying maneuver is needed to connect the 18:02 position to the 19:41 position for southbound routes near 180 degrees. By not allowing that to occur between 18:02 and 19:41, the DSTG method excludes the superior fits that addition provides to BTOs and BFOs. So, I don’t consider the DSTG PDF to be definitive. It is only as good as its assumptions, which may not all be correct. That’s why fitting just the data starting at 19:41 is critical because it avoids this assumption which may be wrong. Still, one has to (1) find a flyable path to connect the 18:02 position to the 19:41 position and (2) demonstrate the fuel available for the whole flight is adequate to match the MEFE time. So far, no one has proposed a route ending SW of 36.5S which has adequate fuel to be flyable. Of course, the DSTG analysis did not factor in fuel. That is its biggest shortcoming, in my opinion. One can, of course, simply throw out all the DSTG PDF south of 36.5S because those routes are unflyable, but even their PDF north of 36.5S is flawed because they did not allow for holding patterns at cruise altitude (which then create a fuel insufficiency) or descending to low altitude for a period of time, which provides the needed delay and SAVES fuel compared to a holding pattern.
@Viking,
1. You said: “2) Not certain the error-distribution is Gaussian (I suspect it is not)“.
DSTG’s Figure 5.2 (Histogram of BTO residual measurement errors) clearly demonstrates that the BTOR PDF is close to being a gaussian. Of course, we cannot construct a comparison PDF using only five BTORs from MH370. We instead assume it is the same gaussian at Figure 5.2 for the case when we are modeling the correct route. Route errors will change the shape of the MH370 BTOR PDF and broaden it by adding systematic errors to the random (gaussian) PDF. So, the best we can do when route fitting is to maximize the probability that the five BTORs match the known gaussian PDF.
2. You also said: “3) An inversion of the argument is totally compromised if the FMT took place within the relevant time interval.”
Of course, all results are compromised if any of the assumptions are wrong. However, I don’t see any issue that is resolved or any probability that is significantly improved by inserting a turn after 19:41. A late turn is unnecessary to match the BTORs and BFORs from 19:41 to 00:11 within the expected noise.
@DrB
Concerning your first contribution, I agree with everything you say.
Concerning your second contribution, I have some reservations. I am not ready to comment on all of them immediately, but for the error distribution of the BTO values, I do not agree.
Based on data from Inmarsat it is obvious that the values have a digital resolution of 20us. If the error (based on a Gaussian analysis) is 29us, this is not much higher than the digital resolution. This issue alone leads to a measurable non-Gaussian contribution.
In addition, I suspect the error without the digital contribution is somewhat smaller than 29us. I do not know how much smaller, but I would guess around or slightly smaller than 20us. Putting this together with a 20us digital resolution will look very close to 29us, if it is analyzed with standard Gaussian tools.
@Viking,
The RMS error caused by quantization is only LSB/SQRT(12) = 20 / SQRT(12) = 5.8 microseconds. Its contribution to the observed 29 microsecond standard deviation is thus rather small. A random noise component of 28.4 Hz combined with the quantization error of 5.8 yields a 29 microsecond standard deviation. So, in fact the quantization noise is trivial compared to the random noise sources.
Viking:
Re the assumption: “…the only logic reason I can come up with for turning the satellite communication back on…”.
There is no evidence that the PF intended to turn the satcom on or off at any time. What we know is that the PF cut power to the left main ac bus, which powers many things, including the satcom. There is no separate On/Off control on the flight deck for the satcom. It is far more likely that the PF had no idea that the satcom was depowered for an hour. He disabled the ACARS and thought that was all he needed to do. He had no idea Inmarsat would later figure out how to use the BTO and BFO data to estimate the path and POI.
@DrB
You would be right if the quantization could be treated as a Gaussian distribution. My point is exactly that this is not the case.
@airlandsman
I think you partly forget the attempted phone call. The pilot would know immediately that the satcom was on when the phone started ringing. You can argue that he did not care, but I think you are not right about that. Why would he (in your case) have chosen to turn the power back on exactly as he got outside Malaysian military radar coverage?
It all points to that he was doing everything possible to disappear completely, and that the escape was carefully planned.
@Viking,
You said: “You would be right if the quantization could be treated as a Gaussian distribution. My point is exactly that this is not the case.”
Of course it is not exactly a gaussian. The shape of its PDF does not matter when it is much narrower (by about 5X) than the dominant gaussian noise. In that case their convolution is indiscernible from a gaussian with the numbers of samples we have to work with, no matter the exact shape of the quantization PDF (which is generally a “rounded-off” rectangle).
@Viking: I haven’t forgotten about the phone calls. The PF had no idea that the BTO and BFO data existed, much less that they could be used to estimate the track and arcs and ROD. He just ignored the incoming calls. The Transponder and ACARS were OFF, so he assumed he was not being tracked by any of the systems. He probably returned power to the Left Main AC Bus circa 18:23 UTC at the same time he repressurized the cabin so he could go back (without a mask on) and check the main cabin.
@DrB
I think we now agree on the general principles for the statistics of the BTO. For the moment that is fine for me. I will be back later concerning the details.
@airlandsman
I think most people were aware in 2014 that advanced telephones could often be traced. The pilot did not have to know the technical details on how this could be done to realize that a ringing telephone was bad news for his secret mission.
I simply do not understand how he could ignore the phone ringing. It basically indicated that the satellite system knew where the airplane was.
I would have panicked and cut the cables to the SDU box or pulled the power plugs to the big amplifier, unless I knew that the trail it left was advantageous to my mission.
@All
Before I explained I generated 20 million flight paths Monte Carlo style, with flight levels (FLs) in the range of 310 – 410, thrust modes ECON, LRC, MRC, M078, M080, M082 and M084, and for navigation modes CTT and LNAV. These were fitted to the measured BTO and BFO data and it was determined that the dominant thrust mode is LRC and the dominant FL from 390 to 399. To reduce “noise” in the probability distributions obtained after fitting, a specific set of 4 million flight paths was generated for LRC and FL 390-399, equally distributed between CTT and LNAV.
Fitting these new set of flight paths to the data showed that the (slightly) dominant navigation mode is CTT. Further analysis has therefore focused on CTT mainly.
A major concern is the choice for the value of the BFO bias. The value of 150 Hz (before restart of the SDU) was taken as nominal value, but as often discussed, this value may be offset after the cold restart, or may have drifted over the course of the several hours that the flight remained airborne over the Indian Ocean. In principle, the drift component is taken into account in the BFO error budget, following the approach of the DSTG, but a possible offset has to be anticipated separately.
I therefore repeated the fit for different possible offsets to the 150 Hz BFO bias: +4, +2, 0, -2 and -4 Hz. These resulted in a shift of the latitude probability distribution, with the distribution maxima ranging from S38.4 deg to S32.6 deg. The area under the curve (the total probability) hardly changed, except for a significant lower value for -4 Hz offset related to the drop in fuel probability south of S37 deg.
The fit procedure also produced an estimate for the BTO and BFO residuals (BTOR, BFOR), through a path probability weighted mean of all BTORs and BFORs obtained for all fitted paths. To test this procedure, these values were compared to single path fits. The tracks of these paths were chosen such that the best fit 7th arc latitude would typically match the latitude belonging to the max of the latitude probability distribution for the specific bias frequency offset (177, 179.5, 182, 184.5 and 187 deg). The linked tables show the results (N.B. “D” is the compensated Doppler). As can be seen, the BFORs generally show a similar pattern between the mean values and the single path values, while the correspondence is less for the BTORs. The question is why. Additionally, I plan to look at the distributions of the residuals, in stead of looking just at the means. It would enable a comparison between single path residuals and the residuals indicated by the maximum of the probability distributions.
What also becomes clear form the tables is that there are only small differences in for example the SD_N-1(BFOR) and SD_N-1(BTOR) for the different offset values. The values suggest a slightly better fit in terms of BTO for the southern latitudes, but the differences are small. Although a more detailed statistical analysis is still ongoing, I tend to conclude that based on this analysis it is not possible to determine a value for the BFO bias offset for the later hours of the flight. The sobering consequence is that an indicated 7th arc latitude range as wide as S32 to S37 deg would not be excessive.
https://www.dropbox.com/scl/fi/x9jc29on0y84bomm3pj9f/Results_v4.pdf?rlkey=9kh5tgn1z0p1jb5cskd566ekl&st=rknui4s6&dl=0
@Niels,
In UGIB (2020) I did not use the mean BFOR as a route fitting statistic for exactly the reason you have illustrated in your most recent results. We have no way to separate OCXO drift of the BFO bias from a systematic route error. In fact our best-fit route has all negative BFORs. I would guess there was actually about a 3 Hz drift averaged over the southbound leg. I used the 1-hour standard deviation of the BFORs as a fitting metric, but it is calculated about an assumed mean value of zero, which is equivalent to assuming the 150 Hz bias value did not change (I.e., there was no drift during the flight). That is the same assumption made by DSTG in their analysis of BFOR noise.
@All
“There is a rebel in each and every one of us. Let it out!”
One generally held belief (I think) is that Z tried to wipe clean his home flight sim. Really? I highly doubt he truly believed that incriminating evidence would not be found by any of the investigative parties.
The flight sim is suspicious because he wanted It to be suspicious.
If he really wanted to get rid of the quasi incriminating flight sim evidence, surely other ways existed.
There is the unlikely possibility that the Anwar verdict (for sodomy no less…trumped up charges) forced/compelled Z to take action before he was completely ready, thereby failing to give the home simulator the “proper” attention it deserved…but far more likely is that he left it knowing full well exactly what evidentiary data it was the investigators would recover.
It’s my belief that this pijacking was in the works for at least a year prior to it taking place. Zaharie posted a video entitled “Window Seal” about a year before MH370. In that video there APPEARS to be a purposeful placement of newspapers that contain various’ inflammatory’, cryptic headlines.
The two most prominent headlines are as follows: THE END IS NEAR FOR TWILIGHT and BOND DROPS BY AFGHANISTAN.
There is no way the newspaper layout was something random like one would expect given the nature of the job at hand (sealing windows).
There are also articles/headlines about Nepotism and shadow banking, among others.
Zaharies Facebook account has hundreds of posts from April of 2013 to May of 2013, almost all of which are political in nature.
Given the meticulous planning Z apparently undertook, I would be cautious when asserting he almost certainly didn’t know about the SATCOM/BFO/BTO. Would MAS flight engineers have any knowledge about the SATCOM?
I’m sure Zaharie was very, very diligent in regards to how the hijacked airliner could be tracked/not tracked and surely knew there were ways other than ACARS and RADAR, first and foremost being SATCOM.
Anyways, I digress. A big thanks to ALSM, Victor, TBill, Dr.B, Viking, SK999, Neils, Mick, Andrew, Paul and everyone else trying their best to find the airplane.
I don’t possess the technical wherewithall to meaningfully contribute here, but I am an anthropologist who has taken a really close look at Malaysian culture and culture bound violence, including being possessed by an evil tiger spirit…which traditionally has been used too justify Amok and the violence associated with it.
Go to 5:11 to see the newspaper arrangement.
Cheers
https://www.youtube.com/watch?v=WAxgsL_evec&t=13s
Go to 5:11 of the video to see the newspaper headlines
Cheers
I have updated my BTO and BFO analysis here – charts on p2: https://tinyurl.com/4swzdb3c
I’d like to recap and summarise where I believe we have landed with the discussion on the comparative fit to the satellite data of LNAV paths starting from 1941.
1. The gaussian distribution – to which BTOR distribution approximates – implies that maximum probability of an individual datapoint is attained when the path error is zero. The probability of non-zero errors increases as the magnitude of errors increases: 68% probability of falling within +/- 1 sigma, 95% probability within 2 sigma and so forth.
2. By aggregating the probability of individual errors, the log likelihood method allows us to arrive at an empirical relative probability of path models according to the goodness of fit to BTO data values observed. We call the relative probability thus obtained p(fit).
3. While smaller individual errors are more probable than larger ones – we need to check that our models do not produce residuals that are not improbably small (over-fitted) or large (under-fitted). Using chi-squared we specify the probability of any model BTOR SD value vs reference SD and we call this value p(SD).
4. Dr B recommends that p(SD) itself has informational value and should be included in model discrimination – not simply as a “qualifier” to check for over-fitting. I repeated the analysis using p(fit) together with p(SD) by expressing the compound metric as a relative probability.
The compound statistic shows greater discrimination / strength of preference for southerly solutions circa ~40S, while termini north of 34.1S fall within the 5% tail.
5. This analysis re-confirms the earlier assertion that path fit to BTO data on its own provides marked latitude discrimination, with a demonstrable preference for more southerly solutions. This conclusion is at odds with the interpretation BTO in UGIB (figure 38, p44) that finds a multitude of probability peaks across the range 29S to 40S, implying that BTO is of negligible utility for latitude discrimination.
6. Turning now to BFO fit, the first pair of charts show the results with Arc2 included/excluded. In either case, and with either specification of BFO sigma, the peak probability by BFO is found at around 37S. Tracks 180 and 190 (to 34.3 and 39.3S respectively) are of roughly comparable probability.
7. The latitude of the BFO probability peak is affected dramatically by assumption of bias shift (last chart). A difference of just +/- 3Hz in fixed BFO bias after power cycling is sufficient to shift the mode from 37.0S to either 34.3S or 39.3S. In light of this, I find it difficult to see how BFO fit can be considered to have informational value – which I think agrees with Niels and DrB.
8. I conclude that the BTO is strongly suggestive of a more southerly terminus and that the BFO data does little to qualify this conclusion one way or the other. Compared to the BTO optimum, a terminus around 34S is extremely sub-optimal and there need to be incontrovertible (non-satellite-data) reasons to prefer it.
@Viking
RE: “I think most people were aware in 2014 that advanced telephones could often be traced. The pilot did not have to know the technical details on how this could be done to realize that a ringing telephone was bad news for his secret mission.”
The SATCOM set up in the cockpit is not like a telephone and I doubt that many pilots think of it as such. They are not taught much about the system, apart from how to make and receive calls. In that sense, it is simply another means of communication, much like VHF or HF.
If ATC calls the aircraft via HF SELCAL, for example, there is a single hi-lo audio chime, a CALL light appears on the pilots’ audio control panels (ACP) and a visual alert appears on the upper EICAS screen where the engine indications and other alerts are displayed. A pilot answers the call by pressing the associated MIC button on their ACP and speaking via a headset or hand microphone. Exactly the same happens if a ground station calls the aircraft by SATCOM – it’s not a ringing telephone.
People might well have been aware that mobile phones could be traced in 2014, but I very much doubt that pilots thought the same about SATCOM systems before MH370.
My phrasing of point #1 above was clumsy. 68% of values should fall *within* +/- 1 sigma, and 95% *within” 2 sigma, of the mean. So probability of an error *diminishes* with distance from the mean. There’s a 0.32 probability of an error further away than +/- 1 sigma, 0.05 further than +/-2 sigma, etc.
@Andrew
Do you know maybe does that also mean that the pilot will only get CALL notification, without specifying (in the notification) if the call is received via VHF, HF or SATCOM?
@Marijan
The CALL light is in the transmitter select switch for the respective transmitter. If a SATCOM voice call is received on SATCOM Channel 1, for example, the CALL light in the SAT 1 transmitter select switch illuminates.
@Angola
I agree qualitatively with most of what you say. Some details are a bit overdone, and a few technical details are not precise to the ‘last decimal’, but that is ok, since you are obviously trying to provoke us to do more to think out of the box.
I like that, and I can assure you that I have more ‘ammunition’ of that kind left for the next couple of months.
@Niels
You have done a great systematic investigation. I am extremely impressed.
I am also happy to see that there is good agreement with figure 3 in my old paper concerning all southerly routes. Sorry for the strange units I was using.
@ Angola
Looks to me like successive pages from one issue of a newspaper to me. How much influence upstream do you think Z had with the publisher?
@Andrew
HF and VHF can also be traced. Even a relatively simple antenna can locate the source to better than one degree (horizontally). Distance location is a bit more tricky and inaccurate, but certainly possible.
It improves a lot if you can measure things like BTO, or if you can do triangulation. Sadly MH370 did not shift satellite during its last flight.
@Andrew
Thanks. Basically, that means that pilot is aware if a phone call comes via SATCOM or other means, so for MH370 person in cockpit would be fully aware that the SATCOM is operational (by looking at the transmitter select switch during call).
@angola
On the sim data, we now understand from ATSB verbal guidance, that the sim data files are a type of Microsoft Flight Sim temporary file. Thus it would be my view that the sim data can be taken as candid, actual, not subterfuge. I do see the sim data as planning.
On the SATCOMs, and in general, I take a “savvy active pilot” view that the pilot was smart enough to know that the SATCOM was a vulnerability, which is probably why it was turned off for the end of the flight after Arc 6/7.
Tbill. The fuel simulations predict fuel exhaustion at/before arc7 – which makes that far and away the probable cause of fresh logon at the end. Or are you saying he turned satcom off then went for a glide?
@DrB
Your post of Oct 23, para 4, that asserts there are two criteria (1) and(2) for a solution that no one has demonstrated. May I remind you that some years ago I presented a study to the group which showed that there was a single simple path that went from 18:02 to the 7th arc and about 186T from the FMT which had sufficient fuel. It is by far, the simplest path with the fewest assumptions.
I recognize that there is one weakness in the study. I adopted a different approach to understanding the BFO drift. Specifically, I used the data for the satellite translation oscillator as the basis for analysis and found that it was highly correlated with the satellite temperature. The diurnal behavior of this offset is not repeatable as the solar illumination is seasonally dependent. In particular the time when the satellite is in eclipse results in a major offset.
I have been chastised for not knowing how the overall Satcom signal chain works, but I have never found other explanations satisfying. Indeed, if the sole BFO offset effect was due to the aircraft TCXO, how did the DSTG/Inmarsat teams identify the eclipse BFO effect?
In addition, I believe that several years ago, the CTT track was only one of several navigation regimes being considered. So far as I can determine, CTT can only be entered by a panel switch (manually). The only question is whether that event coincided with the FMT or with the next waypoint, but the path is the same. No loiter. This leads to a result close to the original IG conclusion.
@DrB
First I hoped that analyzing the second phone call BFO would help, but apparently the effects of the different bias offset values (on top op 150Hz) and the different track angles for best fit exactly cancel.
Also I was planning to look again at the BFO values just after the restart (18:25 – 18:28). However, most likely the warm up transient is a complicating factor and bias frequency may not have fully settled.
@Viking
Thank you! Actually, I hope to find time for a complete write up including the methodology and some of the intermediate result. The Monte Carlo approach offers a lot of interesting options for analysis, once you have this large set of flight paths.
@all,
Hi everyone, we have just published a report on the decoding of Malaysian ATC radar data with an accompaining video. We have obtained access to the data capture and will soon publish it on our website. Some might argue that this is already available data.
In fact, this is new data, although we believe the data published in 2018 and 2009 come from the same authentic dataset.
A careful reading of the detailed report will clearly illustrate the new features of what we call the “ATC-Caption-2025” dataset.
In summary:
1- Mr. Trise extracted all relevant information down to the last bit, including all information about flight MH370, as well as the surrounding traffic. This was not the case with the previously published data.
2- All information sources were carefully examined: PSR, SSR (the various modes), ADS-B, and ADS-C, making the extracted dataset the most comprehensive ever created. All these sources were correlated to improve their quality and reliability.
3- The accuracy level is far superior to that of previously published data. This time, the decoding exploited all available bits, thus providing optimal accuracy without compromise, and used the authentic data without artificially converting it to another coordinate system, for example (from Cartesian to polar). The authentic data are provided first in ATC-Caption-2025, along with other coordinates, when available in the recordings, without any additional processing on our part.
4- A clear understanding of the system’s flaws is demonstrated. This is particularly important for BW angular data, where the mismatch between service messages and target report messages is now correctly accounted for, thus replacing the previously published data.
5- Accurate reading of the “conversations” between the network servers also made it possible to interpret and understand all the data, down to the messages encoded in a proprietary format we called “undefined binary format” (UBF), which has nothing to do with the ASTERIX format. Among all the data, a single field value, proportional to the power of the received echoes, has an obscure function… for gain control? For filtering? Or for RCS tracking?
6- Very fine details were extracted, allowing, for example, 100% proof that the transponder had been manually put into standby mode. There is now no doubt. This is due to the reading of the NUCp, a monitoring parameter, included in the messages.
7- In addition, weather messages from surrounding traffic made it possible to validate the best quality of the COPERNICUS ERA5 weather dataset compared to others. We now systematically use it for our analyses.
For these reasons, this dataset should be considered “new” due to its completeness, accuracy, and level of detail. It renders previous studies of less use and requires, at the very least, a redo and new studies. That’s an open door, isn’t it? … Even if it wouldn’t have a significant impact on the flight’s End Point.
We’re making the ATC-Caption-2025 dataset available to the community for everyone to use it as they see fit… 🙂 on this page: https://www.mh370-caption.net/index.php/malaysian-civil-atc-radar-data-atc-caption-2025/
We recommend the accompanying video: https://youtu.be/-hOy5lAKK2A
PS: It will be posted online in a day or two… final checks…
@Sid asked: “Indeed, if the sole BFO offset effect was due to the aircraft TCXO, how did the DSTG/Inmarsat teams identify the eclipse BFO effect?”
We know how this worked. The pilot signal from Burum was very precise, and was received at Perth. Because the orbit of the satellite was also precisely known, so were the Doppler shifts for the uplink and downlink signals. The measured BFO at Perth for the pilot signal was the sum of uplink Doppler, downlink Doppler, satellite offset, and EAFC contribution. That means the sum of satellite offset and EAFC could be directly determined, and applied to the MH370 reconstructed paths.
@J-LM
I’m glad that you have been granted access to low level records from the Malaysian radar data network for the night of 7th March. However, attribution of source appears absent.
As has been repeatedly stated, we undertook not to share more widely the original records from which our conclusions were derived. I trust that you have been accorded with permission to share, publicly, the records you have to hand.
Briefly, to respond directly to your points above.
1/ My extraction process did “extract all relevant information down to the last bit“. Each ASTERIX message and their data fields were extracted, completely.
2/ We certainly “carefully examined” all the extracted data from the available sources.
3/ Our “decoding exploited all available bits“. Nothing synthesised or artificial.
4/ My processing of the WMKB radar correctly accounted for what you term as the “mismatch between service messages and target report messages. For the benefit of other readers, the service msg included a north passing time while the target report msg included information which, together with the north passing time, enabled calculation of the target report time (within the accuracy of this remote radar head’s clock).
5/ It appears the utility of the “obscure function” remains a mystery to your team.
6/ The final WMKC mode-S EHS messages and Kuala Terengganu WAM/ADS-B messages received from 9M-MRO were definitive in demonstrating the transponder mode had been transitioned to standby. This is not breaking news, your claim that the NUCp parameter sets this conclusion in stone is risible.
7/ Concerning weather data, the most informative environmental information to hand was the contemporary radiosonde data reports. These strongly suggested the presence of ducting effects in the atmosphere which was confirmed by the range of aircraft from the Kuala Terengganu WAM receivers where the ADS-B messages had been received.
I presume that now you understand how DCA-MY used the ASTERIX message SAC and SIC fields. Or, rather, they didn’t use these fields?
@Jean-Luc,
Thanks for your hard work compiling this radar data,
Please expand on how you know the transponder was 100% manually turned off. This is so important. Have you conducted any airborne trials of what exactly happens to the ADS-B data when a transponder switch is rotated?
I can see it’s more likely to be a breakdown in the data to a failing transponder causing the final outputs, rather than switch rotation.
@VictorI
Wait a minute. The BFO error is an observed effect. It needs to be accommodated in the system model. It is not for the eclipse effect alone. Thus the model is incomplete.
If the eclipse effect is a result of a temperature-dependent translation oscillator, ANY temperature dependent effect must be considered.
All you have said is that the auxiliary measurements that you have cited were used to confirm the eclipse effect. But the underlying cause was the temperature-dependence of the translation oscillator. As such the temperature dependent effects at other times must also be considered.
Read the Inmarsat paper and my modeling again.
In any event, the result of my modeling is that a minimum of the errors in BTO and BFO occur at the same parameters.
.
@Tim,
How can you see that the data flow to a ‘failing transponder‘ causes the altitude value to be zero?
Why would only the ADIRS originated altitude word passing over the databus to the transponder be affected and not also the MMR originated GNSS lat-long words?
@Don,
What causes an altitude of zero?….i don’t know, but I’m pretty sure there’s not enough transit time for the switch movement to do it..
That’s why a proper trial needs to be done, to see what happens, with varying speed of switch movement.
This needs to be done before people claim to know the results. It seems premature that this idea is now being quoted as proof as to why mh370 was nefarious.
@Sid: The BFO for the pilot signal is composed of the uplink Doppler, downlink Doppler, satellite oscillator drift, and EAFC terms. Having the recorded BFO and knowing the uplink and downlink Doppler shifts for the pilot signal from the orbital dynamics, the EAFC plus satellite drift terms are known at all times and can be directly applied to MH370 reconstructed paths, not just during the eclipse. The MH370 reconstructed paths must account for uplink Doppler from the SATCOM terminal, the pre-compensation, and bias drift.
@Tim said: “It seems premature that this idea is now being quoted as proof as to why mh370 was nefarious.”
Who said a postulated switch movement is proof that mh370 was nefarious?
I understood that the way the switch operates was a possible cause of an independently observed event.
@VictorI
Its been a very long time since I investigated the issue. Using AI I located relevant portions of the ATSB report and in Appendix C where they disclose their multiple efforts to understand the BFO data and they did use the the pilot frequency data to estimate the eclipse effect.
However, the report says “Satellite translation frequency variation outside of the eclipse effects were not modeled by this approach, but were assumed to be small.”
That is, the translation frequency effect was NOT compensated for the other ping times in the ATSB analysis. By appropriately modeling the data presented in the Inmarsat paper I found that translation frequency errors of up to about 5Hz did actually occur. Moreover, since the temperature effects are different from day-to-day due to, inter alia, aspect angle changes with respect to the sun; it is not clear whether any attempt was made to carefully analyze the error data for the later test flights.
Having said all of that, the path that I proposed also has a smooth singular minimum in the BTO data at 186T.
I believe that all of my studies have been published to your blog, and before that to Duncan’s.
All that I recall doing after developing the BFO correction model was to re-run the path analysis with minor adjustments to the
FMT time as a function of CTT azimuth, select 186T as the best CTT azimuth and then iteratively explore that path to find the best altitude and Mach speed.
DrB agreed that the path is feasible.
Due to the known data limitations after the 6th ping, I do not have a strong opinion on the end game path.
@Sid: I won’t speak for what the ATSB did or did not do relative to BFO modeling. However, we’ve known the correct way to model the BFO, including correcting for oscillator drift of the satellite, since Inmarsat published the JON paper in October 2014. It’s the method I’ve used in path reconstructions.
@Don,
You said—‘Why would only the ADIRS originated altitude word passing over the databus to the transponder be affected and not also the MMR originated GNSS lat-long words?’
Surely there could be a failure of the altitude input to the databus half a second before the transponder suffered power failure, or before the transponder was physically destroyed in the avionics bay.
Sid Bennett,
The statement from the ATSB report that you quote regarding “Satellite translation frequency variation outside of the eclipse effects …” refers to one of the early attempts to model the BFO (OAMS model of 22 March 2014). Subsequent attempts at modeling (the “Eclipse model” and finally the “Unified model”) eventually eliminated any need for assumptions about behavior outside of the eclipse effect. Note particularly the section on “BFO Analysis UNIFIED” (19 June 2014): “The Unified model was developed when Inmarsat found out more about how the Miteq tracking receiver, used in the Perth GES, worked.” The Miteq receiver uses the EAFC algorithm, which you can find documented here:
https://nardamiteq.com/docs/20TEC.PDF
@Jean-Luc
It is great that you got access to the original radar data. I feared it would for ever remain secret.
I hope you will also publish data for the other flights in the area, including flights up until 21:30. My reason to ask for this is that there are still some contrails I have not managed to assign to specific flights. Original radar data may thereby help give contrail studies more credibility.
@Sid Bennett,
I stand by my statement.
The limit for adequate fuel is close to 36.5S, which occurs for a bearing of 184 degrees true. That route has acceptable BTOs and BFOs after 19:00, although it is not the highest in route probability. I have not been able to match all the BTOs and BFOs after 18:25 with a single FMT to that bearing.
A bearing of 186 degrees true ends at 37.5S on Arc 7. That route is not flyable because of fuel insufficiency, regardless of how you connect it to the 18:02 position. See
Figure 14.2-1 in Ulich and Iannello (2023).
Viking wrote ‘It is great that you got access to the original radar data. I feared it would for ever remain secret.‘
What is NOT great is that J-LM and his colleagues’ exposure of the origin data detailed form points to its source and the likely path of its ‘leakage’. I have a firm lead for how J-LM and colleagues came by the data in its origin form, that in itself is revealing of the motives of certain individuals.
The Standards and Practices published by the Bellingcat investigative collective provide a touchstone for ethical behaviour in any open source investigation.
@Don Thompson
I agree completely with you that we should always do our best to adhere to good ethics standards. Bellingcat has been a fantastic example during the last couple of years, particularly with respect to ‘bad things’ coming out of Russia. Of particular relevance to discussions in this Blog is their successful attempts to solve mysteries related to MH17.
I have absolutely no knowledge about how the original data came out, so my previous comment should not be understood as a motivation to violate these standards.
However, when the discussion is limited to MH370, I fear that secrecy from a few Government agencies across the world blocks for finding the correct solution. This can only be solved if we accept limited input from whistleblowers. Everything coming out this way must always be checked against independent evidence like Bellingcat did for MH17.
@sk999. Thank you for sharing the documentation on the EAFC.
By Victor’s earlier description of how it operates, I understood that doppler on the pilot signal up/down links can be computed because location of Burum/Perth and satellite telemetry are known.
Reading the documentation, I understand the distinction being made between the AFC and the EAFC version. The former monitors the pilot signal, makes no allowance for doppler, and makes correction accordingly. The latter makes allowance for doppler. By stripping that out it is able to compensate more precisely for the satellite frequency translation error component.
The doc also includes a description of an approximate method for measuring combined doppler on the Burum-satellite-Perth comms link by monitoring the 24 hourly sinusoidal “periodic error” that is [mostly] attributable to doppler.
Unless I’m missing something, it doesn’t seem to say that the EAFC system calculates the “actual” dopplers based on satellite telemetry. And in paragraph 5.0 where it specifies what info is “known” to the secondary earth station [Perth], it mentions satellite longitude but NOT current/real-time satellite position.
So how can we be sure that the system uses real-time satellite position, as I understood Victor to be saying? If doppler component of pilot is being approximated in some way, that would have obvious implications for the mis-specification of correction applied in the GES receive chain.
@sk999. Having re-read both the MH370 – Definition of underwater search areas”, 26 June 2014 (updated 18 August 2014) and the 8 October “MH370 – Flight path analysis update”, I see that the latter basically says:-
The June report assumed that EAFC employed a sinusoidal approximation, whereas it actually uses a 24hr moving average to estimate “periodic error”. Now we have the actual, known EAFC output, we know what satellite frequency translation error + EAFC correction is – so we don’t even need to know precisely how EAFC performed its calculations.
pdf p8: “This combined factor was used to determine the ‘Satellite and EAFC Effect’ in the BFO calculations (Table 1). This improved approach had the advantage of removing any dependencies on assumptions about how either system behaves.”
If that’s correct, speculation on precisely how EAFC works, including any systematic errors in its doppler corrections, is a red herring?
@paul smithson said: “So how can we be sure that the system uses real-time satellite position, as I understood Victor to be saying?”
No, that’s not what I meant. The EAFC measures the sinusoidal error due to L-band (uplink) and C-band (downlink) Doppler shift (as well as the long-term drift) and attempts to apply an error correction to compensate only for the C-band (downlink) Doppler shift. However, this was incorrectly done in software, so the downlink Doppler was only partially compensated. If correctly done, the satellite orbit doesn’t need to be accurately known (other than the longitude) to eliminate most of the Doppler shift in any downlink signal.
But none of the matters, because as you say “This combined factor was used to determine the ‘Satellite and EAFC Effect’ in the BFO calculations (Table 1). This improved approach had the advantage of removing any dependencies on assumptions about how either system behaves.”
This was my point. We don’t need to know exactly how the EAFC works (intended or actual) nor do we need to know the shift caused by drift of the satellite oscillator. We know the combined effect and can apply this directly in path reconstructions.
Got it [finally], Victor. Thanks.
@Andrew, Victor, Viking, Anybody LOL.
How likely would you say it is was that an MAS 777 engineer would be familiar/
know about the SATCOM, ‘handshakes’ (BTO) and BFO?
Was this information accessible to someone “looking for it?”
Thanks in advance
@Angola: I’d say few people if any outside of Inmarsat would know that BTO and BFO values were recorded and could be used to provide information about the location of a terminal.
Concerning s/c master oscillator stability…According to my records, we knew the by late June 2014 that the 19:41 BFO value was about +8 Hz high due to the s/c eclipse. More important to the current discussion, the data indicates there was no material drift due to s/c temperature outside of the eclipse period. The error was a smooth change as predictable for the EAFC. Put another way, if the change due to the major s/c surface temperature change during eclipse caused only an 8 Hz change in the BFO, then it can be certain the BFO did not drift due to s/c master oscillator tempco during the rest of the MH370 flight. See chart here:
https://bit.ly/4oiMOFw
@angola
I’d say it was unlikely. Line maintenance engineers are trained in troubleshooting and fixing problems with aircraft equipment. In the avionics world that normally involves swapping out LRUs (aka ‘boxes’) in response to fault messages or other problems – they don’t get involved in actually fixing the LRUs and don’t need an intricate knowledge of their inner workings. They may have some very broad knowledge about the existence of a handshake process, but I very much doubt they would have been familiar with the intricacies of BTO/BFOs. As Victor mentioned, few people outside of Inmarsat knew that information could be used for tracking purposes before MH370.
@ALSM. I’m glad you raised this. Before commenting further, where does the +8Hz figure come from? Ashton (2014) in figures 13, 12 shows peak to peak “satellite oscillator temperature” variation of 19 degrees and a corresponding 40Hz change in “satellite translation frequency”.
@Paul Smithson: If you look at Table 9, the EAFC plus satellite drift values are presented for times between 16:30z and 19:29z. Those values can be fit to a sinusoid of amplitude 40.2 Hz, with corrections at 0.7, 10.1, and 3.2 Hz at times 19:41z, 20:41z, and 21:41z, respectively. That’s not too far off from Mike’s estimate of 9 Hz. Or just use the values in Table 9 for the combined EAFC and satellite oscillator drift terms.
Paul/Victor: Correction to previous post:
I meant to write, ~8-9Hz at 20:41, not 19:41. My June 2014 estimate was based on the simple graphical solution in the link. 10.1 Hz is probably more accurate. Regardless of the exact “eclipse BFO bias” at 20:41, the simple graphic shows that the error due to s/c temperature changes only affected BFO values at 19:41z, 20:41z, and 21:41z (around the time of the eclipse). At other times, the s/c surface temperature (and therefore the internal temperature) remained nearly constant. Thus, I don’t think Sid’s concern about possible BFO errors due to s/c temperature changes at other times is a problem.
@Victor, thanks – I appreciate that. I had a similar estimate (11.7Hz) of min-max pilot frequency error. So let’s agree that its somewhere in the 10-12 range.
The main comment that I wanted address to @ALSM.
You have drawn an analogy between the frequency aberration due to eclipse and the likelihood of significant disturbance in SDU frequency in the aircraft. I think we need to be careful of drawing a precise parallel if the conditions were possibly not analogous.
1. Equipment. Was the oscillator and it’s temperature stabilised oven the same on the satellite as the aircraft? Different equipment will have different “ppb” stability rating.
2. Temperature. Was the scale of temperature variation comparable? I don’t know where the “external oscillator temperature” in Ashton Fig.13 was measured, but shows a max-min range of (33.5-14) = 19.5 degrees Celsius. A depressurised aircraft would lose temperature from its original ~24C to as much as -40, so up to 64C delta. That alone should lead us to expect a (proportionately?) larger frequency disturbance.
3. Power cycling. The oscillator on the aircraft has been power-cycled. The one on the satellite has not. We know that a bias shift (I am uncertain of the magnitude) occurs each time the SDU is power-cycled, and this adds to uncertainty of BFO mean error which is no longer directly comparable to calibration values pre-diversion.
4. Settling curve. From what I gather, the settling behaviour of an oscillator that has been cold-stored is demonstrably different to one maintained at steady ambient temperature. From my limited understanding: a) this is partly because the crystal itself can behave a little differently; b) because the temperature-stabilisation loop has to make larger correction with possibility of overshoot (vs one with continuous temp regulation during the external temp shock); and c) because temperature gradients in the structure of the enclosure affect frequency.
If these points are essentially correct, then it seems to me we have good reason to expect a larger perturbation of frequency on the SDU after a cold-soak restart than on a satellite that was continuously powered during eclipse. Ergo, a bias drift larger than 8Hz is entirely conceivable.
@Paul Smithson
Re: Fuel/glide after Arc7
In some areas of Arc7 such as 30s, due to upswing curvature of the Arcs, the distance to Arc7 is shorter distance/poss MRC (eg; from ISBIX). Further, to fly over Arc7 in that area requires quite some significant slowdown/descent before Arc6. Therefore it is possible to trade off altitude for fuel by descending before Arc 6/7 and saving fuel. The flip side of that approach is lousy fuel efficiency at the lower altitude, keeping in mind we see yet another descent at Arc7. One possible minor fuel savings is that only one engine is needed upon descent, and/or Right IDG could be off as far back as IGARI I suppose. This overall scenario I see as what probably happened, based on my studies, but right now I have a hard time exceeding much over 150-nm from Arc7 (at the low altitude approx FL050 I assume after Arc7 descent).
@Paul Smithson
I have no detailed knowledge about temperature stabilization in the satellite. However, the temperature coefficient seems to be near 0.5 Hz/K, while the coefficient for the SDU is known to be 0.3 Hz/K. This difference indicates that the two systems were different.
Concerning the temperature, I am relatively sure the cabin remained closed during the off time for the SDU. The pressure was most likely reduced, but we do not know it the heating remained on. That makes a large difference. If the heating remained on, the temperature drop was probably minimal. If it was off, I think your estimate of the drop is reasonable but slightly too large.
If I were the captain, I would have left the heating on. The security cameras at the airport did not show him with an arctic coat, and he came from a hot climate. I doubt he would have been able to control the airplane at temperatures below minus 20 C for more than a few minutes. I come from a cold climate, and I think I could maximum make it for around 20 minutes at minus 20 C with a light coat. People from Greenland can make it for somewhat longer. They have special DNA mutations, particularly to keep their fingers warm without loosing too much heat to allow that.
In other words, I am quite sure the heating was on. In that case the SDU offset after turn-on would have been minimal.
Based on these arguments (and the what I said before), I concluded that the offset at 19:41 was due to a U-turn a few minutes later.
@Viking. Good points. But doesn’t cabin heating come from the pack air. If packs are off, no cabin heating. I am unsure if there is some separate way of heating the cockpit. Or packs on but only heating the cockpit zone?
@Paul Smithson
I do not know the detailed wiring of the airplane, but I think there are a couple of ways to keep heating on and the SDU off simultaneously. I do not know which specific method the pilot used.
@Paul:
1. “Was the oscillator and it’s temperature stabilised oven the same on the satellite as the aircraft?” Definitely not. Completely different assemblies. But I was not making any such suggestion/comparison. I was only referring to the TCXO drift on the spacecraft, not the OCXO in the AES. My point was that the spacecraft master oscillator bias error is ~zero except around the eclipse period. After the eclipse, the BTO bias returned to what it was before the eclipse (assuming no AES OCXO changes).
2. Again, apples’ and oranges. BFO drift due to master oscillator (a TCXO) drift on the spacecraft and BFO drift due to OCXO drift in the AES are independent issues.
3. “…this adds to uncertainty of BFO mean error which is no longer directly comparable to calibration values pre-diversion…”. Not true. Power cycling results in <0.1 Hz BFO Bias change (after warmup…3-4 minutes typ).
4. "…settling behaviour of an oscillator that has been cold-stored is demonstrably different to one maintained at steady ambient temperature….". I'm not sure what you mean. After the nominal 3-5 minute warmup period (1823-1828), the OCXO is at the exact same temperature it was at 1641 and 1723 and 00:10. Thus, no change.
As discussed by email, the only issue that needs some further consideration is what would happen to the BFO Bias if the cabin temperature dropped to, say -30C when depressurized circa 1721, and the cabin REMAINED at that temp? That could result in a 10-20Hz error. However, I believe the cabin was repressurized circa 1823 and thus the cabin temp was probably back to normal for the rest of the flight.
@ALSM. Thanks for your input here and the benefit of your subject-matter expertise via email. In my earlier post, I mistakenly thought you were drawing an analogy between the temperature sensitivity of the oscillators on the spacecraft and the aircraft.
Of the points I subsequently raised, you were able to quantify the plausible temperature sensitivity of the OXCO on the aircraft by reference to technical compliance documentation of analogous equipment under test conditions. This yields a coefficient estimate of 0.43Hz per 1°C across the temperature range of interest. Assuming initial cabin temperature of ~25°C , cooling to 0°C or -20°C would produce a temperature delta of around 11Hz and 19Hz respectively. So a bias drift of 10Hz-20Hz looks plausible if the aircraft was depressurised with packs/heating off.
Victor clarified that any frequency disturbance within the satellite is a non-issue, as is any aberration in the EAFC compensation because the combined effect has been stripped out of the calculations. You clarified that the retrace error associated with power cycling and the asymptotic settling beyond initial warm-up period is too small to be of relevance (<<1Hz).
Meanwhile, based on Bobby's characterisation of Allan Variance as distinct from BFO measurement noise, UGIB(2020) states that "small drifts of several Hz during a single flight are seen on prior flights" (p155) and that "changes of as much as 5 Hz can occur in 3 hours" (p48).
@airlandsman
I agree with most of your points with the exception of one very tiny detail and a somewhat larger issue.
The tiny detail concerns your point 3. If the temperature of the equipment is within the factory specified range (-40C to 40C is typical for modern equipment, the older SDU might be a bit worse), things are as you say. Outside this range things can go very wrong leading to large offsets and/or drifting. Since the temperature during the relevant period before SDU restart almost certainly did not get outside the relevant range for the SDU it is irrelevant here. Sorry to be a nerd in this case.
However, during the later parts of the flight (after 21:00) strange things happened gradually. The biggest problem is that the S/N ratio for the received signal (at the satellite) increased systematically in a situation where it should decrease systematically. The worst discrepancy is something like 5 sigma, so it is basically against the laws of physics – unless the temperature in the cabin fell to below minus 40C.
Since the big amplifier was rated to something like minus 60C, the issue most likely originated in the SDU box.
@Viking
How would the cabin temperature fall to below minus 40°C?
@Andrew
Heating turned off would do the job at normal flying height towards the south. However, the deviation develops too fast to fit this alone.
Heating off plus an open door overshoots towards south.
Heating off plus an open door while flying towards Christmas Island fits nicely to all points. In addition, the expected shift in BFO from the same temperature curve makes all BFO values fit within the consensus standard deviation of 4 Hz. There is no longer any need for assuming 15 Hz standard deviation for the BFO values.
Thanks to all of you for the discussion of the derivation of the BFO correction. It shows that there are several ways of estimating the BFO offset and they yield similar but not identical results.
My statements were made to clarify the corrections that I applied to the BFO data when doing the path computations.
Since the path I proposed happens to result in a joint minimum of the BFO and BTO errors and encompasses the path from before the FMT to at least the 6th arc it seems reasonable to occasionally remind the group of the result.
Once this path satisfied the fuel criteria I did not spend any time exploring other altitude regimes. Once set, all of the parameters relating to the aircraft were left unchanged.
As many current discussions accept that there may have been a pilot present for some or all of the time after the FMT, one could speculate that there might have been a more efficient altitude profile (step climb?).
The only argument against the IG position is that it was thoroughly searched (with older technology) and the wreckage was not found. There does not seem to be definitive evidence of whether or not there was controlled flight after fuel exhaustion and this is our biggest impediment regardless of which solution you favor.
@VictorI
It is sometimes useful to revisit the ancient portions of this blog. Yet I find it impossible for me to recall and integrate all of the discussions.
I located some interesting discussion in a series of posts starting at About April 10, 2020 of a dialog between DrB and I (with interpolations and distractions introduced by a person who shall not be named).
One of them included fuel probability study at 186T that presented a computed result where the probability was 100% between FL395 and FL415.
I can understand that many are reluctant to consider locations such as the 186T result as the location is not supported by the drift analysis. Being an “electrical” type and not an oceanographer I am loathe to proceed further on the subject.
@Viking
You said: ” Heating turned off would do the job at normal flying height towards the south.”
I don’t believe it would. The ram temperature rise at M0.84 is about 30°C. If the static air temperature is -55°C, for example, the total temperature is about -25°C. The cabin temperature would certainly fall if the air conditioning were turned off, but it would not fall to below -40°C.
@Viking
1. Your assumption about the AES operating temperature spec is incorrect. Here is the correct operating temperature range from the manual:
Operating temperature –67 °F (–55 °C) to + 158 °F (70 °C)
2. Your comments about the C/N0 values are also wrong. As I have explained on this blog and elsewhere many times, S/C to A/C distance is only one of several factors that impact the C/N0 at the Perth GES IF. (Path Loss = -37-20*logF-20*log D, (F in MHz and D in st miles). AFor the specific flight path of MH370, the variation in distance was quite small, resulting in a path loss variation of <1db over the whole path. Factors such as the AES antenna pattern, S/C L band antenna pattern, transponder loading, etc. have a greater impact on the combined C/N0. In summary, no laws of physics were broken. The Receive Power and C/NO (at the GES IF) were completely normal for the entire flight. Even the small anomalies circa 1825 have all been explained. Co-channel interference in the random access channel caused the low C/N0 values then.
@alsm
On Oct 30 you wrote: “Power cycling results in <0.1 Hz BFO Bias change (after warmup…3-4 minutes typ)."
Do you have experimental data/ literature / datasheets that show this?
I also don't expect big jumps, but I don't know how to put a number there (except for running some experiments). I would expect large temperature sweeps to potentially have some impact.
@Andrew
You misunderstand my first remark. It is my fault because of the way I phrased it. What I mean is that if you keep flying south (beyond the maximum fuel range) you will pass a cabin temperature of -40C.
That is exactly the reason I rejected that option. The cooling happens too slowly with just the heating turned off.
Long post alert!
@Sid, thank you for reminding us of the inherent attractiveness of the first generation models. Back when fuel was infinite and BFO was inviolable, 38°S looked like the best bet. It was the solution that simultaneously minimised errors on BTO and BFO without resorting to multiple manoeuvres. Even more compelling, it only required a single turn around 18:37 – matching the BFO-predicted time window for a turn south. At the time, estimates of maximum range cruise distance thought it was reachable and when the flaperon was found, the CSIRO deemed the priority zone “compatible” with ocean drift models.
The trouble is that the plane isn’t there. Unless you elect to reject search efficacy, or cleave to highly improbable end-of-flight notions – the entire high probability zone around 38°S can be confidently excluded.
When “First Principles” reviewed the quandary, they took DSTG’s pdf to denote the universe of reasonable possibilities. By removing everything searched so far, they identified two residual probability zones along the 7th arc, one to the north and one to the south. The north looked like a better bet on the basis of (refined) drift models. Also in its favour: fuel feasibility, possible debris detected by Pleiades, and gaps in the aerial search. Great, let’s go there!
In 2018 OI did go there. They searched wider out at the northern end of the zone. Then, proving the power of their search capability, they went on to cover 112km^2 in a little over 100 days with a swathe width of 55-60NM all the way through to 27°S. Still no plane.
Meanwhile, Bobby’s fuel modelling had uncovered something very interesting. In a footnote to the Boeing fuel flow tables there is a 3% penalty per 10° elevation in temperature. In light of this, he calculated, south of ~36°S should be impossible. Now that we have identified the fatal flaw in the first (infinite fuel) search strategy, we can focus on areas north.
Attempting to squeeze more informational value out of the sparse satellite data, UGIB came up with a statistical approach that required simultaneous satisfaction of multiple statistical tests rather than just minimising measurement errors. It turns out this didn’t help much with latitude discrimination, but by triangulating with independent datasets (refined fuel and drift models plus aerial search null), the solution emerged from the murk: a distinctive and exclusive probability peak at 34.2°S. What’s more, this coincides precisely with a track of 180° to the South Pole.
Thus UGIB doubled down on the area recommended by First Principles. The trouble was: this had just been searched unsuccessfully in 2018 by OI. Worry not. This time we are so sure of our solution that we need to look at the “residual of the residual probability” – namely any gaps in the search in the high probability area, then work out to the “saddle” on either side, corresponding with less likely end-of-flight scenarios.
We don’t know exactly which seabed areas were ticked off by OI in early 2025. It looks like they covered the “data holidays” within the UGIB high probability area, as well as the “saddle” beyond the arc to a distance corresponding with a 70NM glide at 180°. There is a lower probability segment of “outer saddle” to the north, plus a saddle along the inside of the 7th arc beyond the 25NM already covered. However, the sad truth is that the best prospects have been searched without success. What remains now is the residual of the residual of the residual, and a rapidly-diminishing prospect of success.
Or otherwise Something New. Part II follows…
@airlandsman
I think you are talking about the main amplifier. In that case I presumed -60C to 60C. Your numbers are exact, but it makes no difference to the conclusion, since the cabin never got below -55C.
For the rest, you are right that there are other noise contributions, but I do not think they dominate for the S/N levels observed.
Two details are particularly important. One is the effective area of the local antenna at the airplane (the area projection perpendicular to the direction of the satellite). That gives a much larger contribution than the simple diffraction contribution you mention (which is indeed almost negligible).
The other one is the angular momentum projection (due to the discrimination between the two circular polarizations L and R). The dominant factor of that one is the same as the one for the effective area, so that contribution is similar, and also much larger than the diffraction contribution.
All 3 contributions have the same sign (negative for the S/N ratio). That is the reason an increasing S/N ratio is very strange.
For any straight route (along a great circle of the earth) that is effectively impossible (beyond 3 sigma standard deviation for the S/N) unless the airplane is flying ‘backwards’ (toward the satellite).
For curved routes it is more complicated. However, the best curved route in agreement with both BTO and BFO is the most northern of the 3 proposed (curved) ones by Inmarsat at the very beginning of the searching. It is a good guess, but still off by around 3.5 sigma for the received S/N. In addition, it is hard to fly (close to instability against stalling).
@Paul Smithson
Indeed a long contribution, but clever words.
You missed 3 zeros in one place (typo), but otherwise I agree with most of it.
@Paul Smithson
OI did not search the most important area indicated by “Pleiades” images, which is S35.4-S35.5 near the 7th arc. This area was searched by Go Phoenix using towed synthetic aperture sonar (SAS) if I remember well. OI did not rescan most of the “Go Phoenix” area. One of the questions I have is how they provided platform stability needed for SAS, in a towed configuration, and how that relates to coverage %.
@Viking. The temperature range was for all components of the AES.
As for the C/N0…You obviously have no idea what you are talking about. The path loss varied only 0.25dB over the whole flight (obviously not dominant), while the other factors in the link budget can easily vary on the order of 0-10 dB for any individual transmission.
“…effective area of the local antenna …”? The antenna was a Ball phased array. The gain varies depending on the pointing angles (az and el). The system as a whole attempts to keep the EIRP ~constant, but that can vary several dB depending on feedback from the GES. It seems apparent you don’t understand how any of this works.
“…angular momentum projection…”? What are you talking about? Nonsense.
“…simple diffraction contribution you mention…”? I have not written anything about a “diffraction contribution”.
Time to shelve this. There was nothing unusual about the C/N0 values.
@Niels askes “Do you have experimental data/ literature / datasheets that show this?”
Yes. There are many papers online about the subject. Here are a few OCXO files. https://bit.ly/435HNHN
See the file: OCTCXO.png for typical retrace stat’s.
MH370 C/N0 vs time: https://bit.ly/48TAPJF
@alsm
Many thanks for sharing!
The figure “OCTCXO.png”, is it taken from a datasheet / what is the source?
What I notice is the really small 24h aging number indicated as well in the figure.
@alsm
It looks like I found the source (through Google):
Electronics, Inc. Frequency Electronics, Inc.
Tutorial Precision Frequency Generation Utilizing OCXO and Rubidium Atomic Standards with Applications for Commercial, Space, Military, and Challenging Environments
IEEE Long Island Chapter March 18, 2004
Olie Mancini
In my understanding, the figure you shared is related to a special product of “FE”, in which the seem to use a “DOCXO” (double oven). So how does that compare with the OCXO we think was on 9M-MRO?
@airlandsman
At least we now agree on the size of the simple diffraction loss. For the rest I have to give you the same answer as you gave me.
A standard phased array antenna can only correct for the phase errors (optimizing the pointing), not for the polarization terms and/or for the effective area. For the area part that is simple to prove. Turn the antenna so it points perpendicular to the direction to the satellite. Independent of which phase corrections you apply there is zero signal.
The polarization part is more difficult. If you know the difference between Fraunhofer and Fresnel diffraction it is not too difficult to explain. If not, I need to send you a link to a textbook on the subject.
More modern antennas (this does not apply to the relatively old version at MH370), can partially correct for area/polarization, but only around 10% of the effect, so not enough to explain more than a tiny fraction of the problem. This technology was invented later using methods similar to AI.
@Niels Yes, that Frequency Electronics, Inc. figure referred to a DOCXO. The Honeywell 81771-MBE is described as a OCXO. For many years, I have tried to get more details on the 81771-MBE assembly used in the MCS-4200/7200 series, but no luck. Don also tried. The original manufacturer has changed hands a couple of times making it difficult to find any hard data.
The question of “retrace error/offset” for the 81771-MBE specifically is difficult to answer precisely. I’ve reviewed dozens of papers and commercial spec’s for OCXOs, and the best I can offer is an “informed speculation” range. For high performance OCXOs (I’m confident the 81771-MBE fits in that class), the 24hr retrace offset is probably in the range of 1X10^-9 to 5X10^-10. And here is one of the complications. No one spec’s the retrace offset for short times, like 1 hour (18:25) or 1 minute (00:19). The off/on time in most spec’s is 24 or 48 hrs. But the literature does consistently indicate that the retrace offset tends to increase with off time. Thus, the retrace offset after only 1 hour off at 18:25 is probably <1X10^-9.
I would remind everyone that the BFO data only provides 2 solid critical pieces of information, neither of which are materially affected by any type of OCXO error:
1. MH370 went south following the FMT.
2. MH370 was descending very rapidly at 00:19.
Of course, path models that use a combination of BTO and BFO data will be affected at the margins if the retrace offset at 18:25 was a few Hz, but I assume the generally accepted +/-7Hz error budget already accounts for some retrace offset.
My earlier long-post alluded to a Part II. So what’s the Something New?
TL/DR. Analysis of solutions with no manoeuvres after the “FMT” yields a tight spectrum of BTO-optimised solutions ending at 39°S-40°S. This corresponds with the “global” optimum from Arc 2. An LNAV solution is only feasible using waypoints – and this route corresponds very closely with a 2-stage turn (MEKAR-SANOB-IGEBO), then a trajectory south using IGEBO-RUNUT and possibly 4085S. The predicted terminus (39.6°S) lies just beyond the southern limit of the search to date. The predicted speed is M0.84 at FL360 and the route is demonstrably fuel-feasible if certain assumptions are revisited.
Now the detail.
The first step was to revisit path models with a single turn from N571 – this time optimising for BTO rather than BTO+BFO combined. This predicts a terminus just south of 39°S (as DSTG also found).
Next, I determined the impact of model initiation assumptions: lateral offset from N571, timing error at Arc 1, speed and track angle. This demonstrated that a “general theory” of no-loiter paths can be characterised because all variants share the same BTO after FMT at 1840 after turn completion. In practice, this means that BTO optimisation constrains the longitude of the turn (sitting along the same virtual 1840 arc), while Arc 1 crossing constrains the latitude. The result is a tightly-focused generic BTO-optimum (39.0°S – 39.6°S) for no-loiter paths that isn’t contingent on a specific “prior”.
Now, we know that LNAV must use waypoints. Examining the predicted trajectory southwards, a unique pair of “enroute high” waypoints (IGEBO-RUNUT) offers a viable match. At the FMT, no way point exists to match a single-turn from N571. The next best solution (least-bad BFO errors) is a turn from N571 in two stages: MEKAR-SANOB-IGEBO. From RUNUT, the 4085S suggestion is conjecture, although the path terminus by this stage is insensitive to waypoint assumption.
Now stand back. Our analysis – derived from empirical BTO optimisation – has predicted a path that turns left along the first available airway (P627) at the end of the Malacca Straits. Predicted speed and altitude (M0.84, FL360) are highly plausible. The BTO fit is close to the global optimum – whether modelled from Arc 1 or Arc 2. The predicted terminus zone at 39.6°S has not been searched. By all accounts it looks like an attractive and parsimonious solution – except for fuel.
Using conventional fuel models, packs-on, inclusive PDA and TAT adjustment, the aircraft should have run out of fuel at 23:58:30, a deficit of 19 minutes against assumed flame-out at 00:17:30. This deficit can be bridged by a combination of the following:
1) 17 minutes for packs-off from diversion onwards, if true bleed air fuel penalty is ~4%
2) 2 minutes for trapped usable fuel downstream of the tanks
3) 3 minutes gained by reduced electrical load
4) 2 minutes if last 14 minutes were under one engine
Additionally, but less likely:-
5) ≤4 minutes if actual fuel exhaustion occurred a little earlier than 00:17:30
6) ≤2 minutes accessible fuel below FQIS zero calibration point
7) ~4 minutes error budget (FQIS and fuel model combined)
Some folks raise queries regarding BFO and drift model compatibility. Both of these are vulnerable to error. The first because of plausible bias drift. The second because of sensitivity to parameter assumptions for wind forcing magnitude and directional offset.
Conclusion: the Smithson solution to 39.6°S sits on the only (promising) section of 7th arc that has yet to be searched. It is derived from empirical path optimisation that prioritises BTO over other lines of evidence. It is demonstrably reachable if fuel assumptions are re-examined. Objections on grounds of either BFO or drift models are not compelling. The corresponding search area is relatively small.
@Viking I’m sorry to be blunt, but 100% of your post at November 2, 2025 at 7:52 am is nonsense. Nothing in that post makes any sense.
This thread started with your claim that the C/N0 values systematically increased with distance, inconsistent with the laws of physics. I have explained why that that claim is false, and provided Inmarsat data to prove that the C/N0 did not increase as you claimed: (https://bit.ly/48TAPJF).
@Paul Smithson: Do you find it odd that the best estimate BFO, drift, and fuel models all point to a POI further north?
@Paul Smithson
Thank you for both Part 1 and Part 2. I whole heartedly agree with your decision to concentrate on BTO (the most robust data we have) analysis. Both BFO and drift analysis are far less robust, indeed both have so many ‘ifs and buts’, that they have only ‘muddied the waters’ (so to speak) all along.
Your track IGEBO RUNUT passes approximately 20/nm west abeam MABIX (N3°15.97′ E94°50.92)
I would be interested in knowing your time altitude speed mode and fuel remaining at that point, and on arrival at RUNUT.
@Victor. In an ideal world, the various lines of evidence would converge. If they did, the aircraft would have been found. That in itself is ample signal that there is at least one major assumption error.
BFO. We have known from the outset that “autopilot” routes and [BFO] data optimised routes are incompatible. The “optimal” BFO terminus shifts +/- 3 degrees with a 3Hz drift that Bobby thinks is likely. Much larger drifts should be expected if the aircraft was cold – and we have good reason to believe it probably was. The 1941 datapoint is anomalous by any path – and this alone should ring alarm bells. If BFO was “true” it would be marginally useful. With everything we now know, there are good reasons not to trust BFO fidelity.
Drift modelling. Various proponents have claimed it to support completely different origins. The CSIRO model – which I consider the most credible – is not as deterministic as you would like to suggest. Changes to wind coefficient (+2% shifts latitude south 2 degrees) and directional offset (10 degrees left shifts latitude south 2 degrees) show us how sensitive the model is to parameter assumption. For most (flat, buoyant) debris items, both parameter changes are relevant. Don’t forget that when the flaperon was found in Reunion, CSIRO promptly declared this to be aligned with the (then) priority search at 38S.
Trust the BTO. In my view, it is well and good to resort to secondary, less precise latitude clues if your primary data has inadequate solution discrimination. In my analysis, I have demonstrated that BTO is a much better discriminator of latitude solution than most believe (DSTG found the same). From the sat data, there is no more compelling a reason for a manoeuvre between FMT and Arc 2, than there is anywhere else between Arc 2 and 6.
Fuel. Some believe that fuel feasibility is that compelling reason, even if requires a path that flies in the face of the most parsimonious solution. In my post, I have set out how I believe that southern termini may be reconciled with fuel endurance, rendering the simplest and most plausible solution feasible.
From my perspective, the initial search area definition made sensible use of available data. The null search result tells us that the plane is most likely in one of the residual areas. At First Principles the weight of consensus favoured the 34-35S area but the plane has not [yet] been found there after 3 searches and counting. I have set out why I think the southern one is a better bet, and always was. With adequate fuel, the Smithson solution has a great deal to recommend it.
@Ventus.
Altitude FL360, M0.84 from Arc 1 onwards (and probably earlier).
IGEBO abeam 18:34:20, assuming zero BTO/timing error at Arc 1. Fuel remaining 40,760kg – per “normal” fuel model, before applying any efficiency savings.
RUNUT eta 21:01:50. Fuel on board 17,750kg before applying efficiency savings.
MABIX abeam 18:55:50. Fuel on board 31,455kg before applying efficiency savings.
@Paul Smithson
Thankyou.
@Paul Smithson
Not to steal your thunder, just curious what is your take on “The Waypoint Hypothesis”?
https://globusmax.wordpress.com
@John. I’m aware that globusmax has favoured 40S since way back when. Other than positing a waypoint on the other side of the world I dont think they either demonstrated HOW, provided a detailed empirical case, or provided counterarguments to the naysayers. “3 out of 4 pilots recommend it” doesnt cut the mustard. Still, I’m glad that others believe in a southrtn solution and am hopeful that our numbers wkill grow!
@Paul
Thank you for your reply. At a human behavioural level I find found Glubusmax’s argument compelling. Curiously it fits known technical observations to an extent. After more than a decade of position fitting analyses of known (and unknown) data and searches being unable to find the flight endpoint, I have been wondering if or when a technical explanation might make the region of the 7th arc at 40S worth searching. Your posts make interesting reading (at least for me).
@Paul Smithson
@John
As far as I can tell, Globusmax’s waypoint theory track closely aligns with the best estimate that I was able to make (from various graphics way back when – the VERY early days) of the more western (450 knots GS) of the original NTSB-NYT “two tracks”. (So far as I am aware, the actual coordinates of the “two tracks” were never published).
My (estimated positions) are as follows.
ORIGIN of NTSB-NYT TWO TRACKS approximately (0.6°N 94.1°E).
NTSB-NYT Fast Track (450 knots GS) End approximately 48°S 082°E
NTSB-NYT Slow Track (400 knots GS) End approximately 46°S 086°E
The Globusmax track is about 10Nm west of the NTSB fast track, and if your 39.6°S on the 7thArc is at 84.744°E, then you are about 15Nm west of the NTSB fast track.
@Ventus. Apologies, corrigenda for yesterday’s question/response:-
IGEBO: 18:34:20, fuel remaining 33,850kg
MABIX abeam: 18:55:50, fuel remaining 31,455kg
RUNUT: 21:01:50, fuel remaining 17,750kg
All fuel figures before applying efficiency savings.
@Victor. There’s a briefer “scientific method” response to your question yesterday regarding divergent evidence.
My science problem has 2 lines of relevant evidence. Line A is gold-standard evidence that leads to one answer but with wide confidence limits. Line B is a bit fuzzier and uncertain, that leads to different conclusion. Which evidence should I use?
1. Hypothesis 1: believe both, go to solution predicted by f(A,B)
2. Hypothesis 2: discard B, prioritise f(A)
I would start with (1), then move to (2) if the first answer is proven wrong.
@Paul Smithson
Thanks for that Paul. The 40,760kg abeam IGEBO threw me, but what I really wanted was time & fuel abeam MABIX and at RUNUT, and they looked OK (and have not changed).
@alsm
Having thought a bit more about it, I would say that DOCXO vs. OCXO perhaps does not matter too much regarding retrace error. As it is possibly related to thermal stresses in the resonator assembly and mounting. The residual stresses would depend on temperature in an unknown way, so following this hypothesis it is hard to say if keeping the oscillator at lower temperature for shorter time makes the retrace error smaller or bigger.
From what I’ve seen so far, to me the 1e-09 estimate seems reasonable, which would mean less than 2 Hz in the MH370 case. I’ll continue the literature search and let you know about the findings.
@Niels
I tend to agree that the retrace offset should be about the same for a OCXO and DOCXO. But some of the papers and spec sheets I’ve read say the DOCXO has less retrace. I can’t see any theoretical reason for that. Maybe it’s more of a marketing claim to get folks to trade up.
What ever the value, it is well known that the effect is not cumulative. IOW, the error does not always increase of decrease. It is random, thus approximately zero on average. Of course, there is the long term drift component, but in the case of the MCS 4200/7200 OCXO, that error is continuously monitored via GES and adjusted to zero (16 Hz step resolution).
Changing the topic slightly to the conduct of the resumption of the sIO search that is expected.
Ocean Infinity is to operate an Armada 86 vessel on a seafloor wreck survey together with Carl A. Allen’s Allen Exploration/AllenX team under the auspices of National Museum of the Philippine’s Maritime and Underwater Cultural Heritage unit.
This survey is to explore the San Bernadino Strait for WWII and Spanish colonial era wrecks, the Strait separates Luzon and Samar islands. Considering the allotted duration, its expected that the survey will be exploratory, perhaps identifying and mapping potential wreck sites.
Armada 86 05 is, as I write, repositioning from Singapore’s western anchorage area to the harbour basin where the ST Engineering and Mooreast facilities are located. Previously, Mooreast has provided mobilisation services to the Armada vessels. One of the AllenX vessels arrived in the basin Nov 1st.
Armada 86 02 is enroute to Vancouver while A86 04 and A86 06 remain at the Vard jetty in Vung Tau undergoing trials and completion.
The impact of typhoon Kalmaegi, and a second building storm in the Pacific, may add delay to the project.
At this point, should A86 05 be committed to the resumed MH370 search, the soonest it might get to the sIO is likely to be late December.
The pilot whisperer Richard Godfrey seems to have gone to ground. Wondering what is happening with WSPR I stumbled across this interview he conducted regarding AI171 where he proposes that water egress into the Equipment/Electronics Bay. See: https://42kft.com/root-cause-of-the-air-india-ai171-crash/
@John,
Hmmm, yes, the FIP’s submission to Ministry to Civil Aviation, India, bears some common themes to those espoused at your web link. As Andrew commented, above, tomorrow a preliminary hearing in front of members of the Supreme Court considers a 200+ page ‘writ’ submitted by the FIP and the father of the Air India captain.
[Comments here are closed. Please continue the discussion under the new article.]