For a number of years, Ed Anderson has independently used hydrophone acoustic evidence in an attempt to localize the point of impact of MH370. He has detected an acoustic event that occurred along the 7th arc at a time that is consistent with a large portion of the airframe impacting the seabed. Although it is impossible to definitively relate this acoustic event to MH370, and although his candidate site is further north along the 7th arc than most investigators using other methods think is likely, his mathematical analysis localizing the acoustic event has not been challenged.
Recently, Ocean Infinity (OI) and Malaysia announced that the search centered around the “IG Hotspot” (as defined in UGIB 2020 as 34.23S, 93.78E) is indefinitely paused due to unfavorable seasonal weather in the Southern Indian Ocean (SIO) and due to OI’s other commercial engagements. Ed notes that his candidate site at 8.36S, 107.92E is only 42 miles off the coast of Java, Indonesia, in relatively calm tropical waters, and could be searched before Ocean Infinity’s exclusive contract with Malaysia is terminated in June. Unlike other analytical methods that attempt to reconstruct the flight path and imprecisely predict the impact site, Ed’s method localizes the acoustic event to a fairly small area that could be efficiently searched. That’s my recommendation.
Here’s Ed’s letter:
An open response to ending the current search for MH370 due to weather
March 9, 2026 by Ed Anderson
It is sad that Ocean Infinity has now concluded the current search for MH370. I believe that OI has until June before the $70M contract runs out, and there are still viable candidate sites along the 7th Arc.
My acoustic candidate site off the coast of Java is in relatively calm tropical waters, 42 miles off the coast of Java, at 8.36S 107.92E directly on the 7th Arc. When the Armada vessels go through the Sunda Straight, they pass just 200 miles from the 7th Arc site. The epicenter is very specific, likely accurate to within 2-3 km. A single Hugin AUV could search the broader area in an afternoon. Here is some background detail since this has never been covered by the news media.
The site is indeed over 800 miles from any previous aerial/surface/seabed searches, yet it is an exact match for the SATCOM BTO values and a good fit for all other hard evidence. That may seem impossible, but those previous searches were based on a fundamental assumption that MH370 made no turns shortly after leaving radar, even though it clearly was navigating between waypoints for the previous hour. The assumption is understandable, because optimizing for a single path that best matches the satcom BTO/BFO doesn’t work with additional turns. It opens up too many possibilities without new evidence.
The acoustic evidence is now quite clear. Hydrophones were analyzed by Curtin University in 2014 with two papers published as appendices h and i to the Oct 2018 ATSB Final Report. The first report p 21 accurately detected the bearing to the Java event on the Diego Garcia HA08 array, but it was unfortunately dismissed as a “small earthquake in the Java Trench”. Their focus was on a different signal arriving Cape Leeuwin on bearing 301.6 that crossed the 7th Arc toward the Maldives. That intersection was the northernmost extent of the 2018 OI search. The second Curtin report utilized the IMOS Scott Reef hydrophone, focused on a strong signal arrival in an attempt at triangulation with the 301.6 event. In fact, the signal is instead an exact match for the Java event.
There was no quake cataloged for the 01:15:18Z Java event on the 7th Arc, but it is detectable by at least 40 regional seismometers using public data. The first anomalous aspect is that the HA08 hydrophone signal arrival at 01:59:20Z from Java is one of the strongest of the day. It is at least 10x stronger than strong cataloged Magnitude 4-5 quakes arriving from the Java trench. Another oddity is that all previous quakes over the last century near the Java site have been deep in the subduction zone typically 30-90 km deep. The candidate epicenter is at seabed depth. There was some early confusion on my part over the timing, thinking that the event was due to an underwater implosion of sinking debris as sought by the CTBTO, Curtin, ATSB, and others. The epicenter timing turns out to be much simpler, and matches well with the seabed impact of a large section of sinking MH370. The timing is 56 minutes after the last 00:19:37Z 7th Arc ping, which implies that the large section stayed afloat for over half an hour. A weaker arrival of a signal similar to the Java event in a narrow 1.2 Hz band is just 6 minutes after the last ping. That would be consistent with the core of an engine sinking to the 3400 m depth at the site.
All other active hydrophones in the SIO picked up the Java event. The timing is slightly shifted because the sound entered the 1000 m deep SOFAR propagation channel by reflection off the Java coastline. This is similar to the way the missing ARA San Juan submarine was located off the coast of Brazil. The Cape Leeuwin HA01 hydrophone is mostly blocked from the event by the Australian coastal shelf, but it still picked it up. The other IMOS hydrophone at Perth was not recording at the time, as it samples for 5 minutes out of every 15. Seven French hydrophones near Amsterdam Island were analyzed in an unpublished 22 Dec, 2016, University of Brest report LDO-201608. Six detected the event but it was again classified as an earthquake. A geological survey vessel near Exmouth also detected the Java event at the correct bearing using a huge towed grid of over 6,000 hydrophones, but no surface impact.
Experts in ocean acoustics know that remote surface and seabed events do not typically propagate into the SOFAR channel. The travel outward in rings of convergence separated by broad shadow zones. Only when they encounter terrain that intersects the SOFAR channel can it enter by reflection. The “301.6” event detected by Curtin has timing roughly consistent with the Java event symmetrically reflected off the mid-ocean 90 East Ridge.
There is no dispute that a loud event took place. Any seismologist can verify the event and perhaps improve the epicenter calculation with better tools. The only MH370 expert objections have been about how the sound was generated and whether it can be associated with MH370 vs geological. The odds of an unassociated anomalous seabed event taking place directly on the 7th Arc just as MH370 would have been sinking are difficult to calculate, as there is nothing for comparison.
One low and slow flyable waypoint path to the site is suggested that matches BTO exactly, but there would be others also optimized for BFO. The selected path is at oxygen altitude near holding speed and coincides with seismic infrasound doppler detections of possible flybys at Cocos Keeling and Christmas Island airports matching the ping arc timings.
The Java candidate site has no dependence on the cause of the comms outage, early flight events, motivations of the pilot, or any conspiracies. One speculation is that an impaired plane was unable to safely land but could continue to fly toward daylight as the crew dealt with the damage. This could explain why the SATCOM was partially restored an hour after it was lost.
Drift studies are most dependent on the travel time of the first two pieces of debris found, the flaperon at Reunion Island, and “Roy” in South Africa which also traveled the farthest. Both had barnacles, unlike most of the debris found. 2025 photo analysis of the flaperon shows the barnacles are growing on top of abrasions from beaching. This shifts the previously assumed arrival earlier by about two months. That also matches the studied barnacle growth temperatures with the Austral winter sea surface temps near La Reunion at the time.
Some experts have expressed hope that the Java site would eventually be searched, and it has been added to the CAPTIO-N search map.
OI, Malaysia, and the ATSB are aware of the Java candidate site but until there is media coverage, it may go unsearched.
Now 12 years on, thoughts are with the NOK, and hope that MH370 can be found to bring closure for us all.
More detailed technical reports are at: 370Location.org.
I can only admire such great minds, and truly enjoy this site for more than 10 years now. I miss Dennis Workman , with his whit, intelligence and sharp tongue. So thank you for also educating the lay person on scientific and academic subjects related to MH 370. It never dulls.
@All
Let me start by saying that I support searching the site pointed out by Ed, but with some reservations pointed out below.
Effectively the official search seems to have ended for this year. Sadly, OI did not complete the entire search area, but the majority of the high-probability part of the search area was searched carefully without success. What does that mean for our future strategy?
Statistical considerations point to a significantly reduced probability of finding the crash position in this entire area. That triggers me to reopen previous speculations about different crash positions. That discussion is divided into two parts. The first possibility is that the crash happened in one of the small dark spots, or somewhat further from the 7th arc, or somewhat further south along the arc.
The second possibility is that it happened several thousand kilometers north of the present search area where particularly two small areas have been proposed by Ed and me.
Most other Bloggers will probably discuss the first possibility in the future. At the moment I have no significant new information to add to such discussions. My preliminary impression remains that all the 3 probabilities are small but impossible to ignore entirely. They would have been even smaller if the entire area had been investigated in detail. That would have presented us with a clear conclusion to stop all further searching in this area.
The second possibility is entirely different. Until recently, extremely few believed in solutions so far north, but they are the only proposed solutions with seismic and/or acoustic signals. For my solution there are also atmospheric features related to some type of dramatic event observed by METEOSAT7.
In this unusual situation, I think we need to invent an independent and robust method to determine if the crash happened far to the north or somewhere near the official search area. After thinking about several different possibilities for a long time, I think I may have come up with an entirely new possibility inspired by modern physics-assisted archeology. I will come back to discuss that possibility soon.
In the meantime, I strongly support attempts to persuade OI to search the two spots far to the north. They are both very interesting, but for different reasons. The position proposed by Ed has an extremely small search area. Basically, it could be searched within a few hours.
The position proposed by me has a 5-10 times larger search area, so it may take a few days to complete. However, I think the total probability of success is much higher. Why do I say that? The position pointed out by Ed is very close to Indonesia, so local fishermen would most likely have seen or even collected part of the debris in their nets. There are absolutely no reports of such events. On the other hand, my position is in/near a so-called sea desert, where nobody is going to fish. It is also west of all standard shipping lanes and normal flight routes, so it is expected that nobody would have seen the debris field.
Another problem with Ed’s solution is the time the signal appears. It is close to impossible that MH370 could have remained airborne for so long time, even with an expert in paragliding controlling the final part of the flight. Ed tries to explain the late time with debris hitting the seabed as the main source of the signal, but that is definitively wrong. This is proven by signals recorded by the special arrays of detectors. I received a complete copy of their signals. Using knowledge about the underlying technology (partly invented by myself) it is easy to prove that the entire signal going in the direction of Australia was propagating above the jump layer. That means the signal did not come from a seabed impact. A group of Australian geologists have come to a similar conclusion, but not yet published their results.
Most of you may have overlooked that there is also a clear above-jump-layer signal (confirmed by its dispersion) arriving at HA01 at 00:47:04 with a nice match in heading and time with my position and time. It is included in table 1 of U. Kadri’s Nature paper. The heading is 343.16 degrees. Unfortunately, its delay is incorrect since Kadri assumed the crash happened exactly at the seventh arc and that all signals propagated in the normal guiding mode in the Indian Ocean. The extremely small crossing angle between the arc and the heading vector plus an assumption that the signal propagated in the normal guiding layer explains the entire difference. Without these two assumptions, and using my small offset from the arc, the time fits nicely with Boeing simulations of a crash without pilot control.
Meanwhile…we’re all still waiting for that peer-reviewed WSPR paper from Simon Maskell, now years in the making. With all the attention he craved, you would think he would follow up with something.
This is a fascinating breakdown by Ed, especially the point about the HA08 signal being dismissed as a tremor. If the wreckage is indeed just 42 miles off the coast of Java as suggested, the logistical side of a recovery operation becomes much more feasible than the SIO. On a practical note for those of us following these coordinates closely or even considering independent research trips to the region: since the site is within Indonesian/Portuguese maritime influence zones, does anyone know if having a European tax ID like the one mentioned at https://e-residence.com/de/nifonline/ simplifies the legalities of hiring local vessels or setting up a research entity in nearby jurisdictions, or would we need a strictly local Indonesian representative to handle the paperwork for a Hugin AUV deployment?
@Victor Iannello
Yes, Victor, the manifest lack of any form of independent corroboration of the WSPR-based methodology speaks volumes. More so after all the hoopla surrounding the purported involvement of the University of Liverpool.
And when it comes to MH370, sadly it is not just a passenger-laden B777 that went missing. From some circles, the missing aircraft seems to have have taken a fair swathe of professional integrity and intellectual honesty with it.
As I have said numerous times, that whole endeavour has been an abject delinquency of scholarship.
@Victor,
I am wowed.
Many thanks for your strong recommendation and the boost to continue the search!
Also for maintaining a forum that has enabled in-depth MH370 technical discussions, when most social media platforms are limited to a thought span of 280 characters.
I’ll field critiques the of Java Anomaly in detail, which may take some time and bandwidth.
@370Location. You are welcome.
My plan was to wait until the current search was completed, and if unsuccessful, encourage discussion of other candidate sites, including yours and Paul Smithson’s. However, as you reason in your letter, your site represents an immediate opportunity because of the small area, the geographical proximity to Java, the favorable weather, and because it might be searchable under the current contract.
@Victor,
I’m so glad you decided not to wait!
@Angela,
I’ve considered that a permit would not be needed to search for MH370 in the Indonesian EEZ Exclusive Economic Zone, because it’s not about exploiting their seabed resources. It’s not much different than research ships running bathymetry when in transit, and there are plenty of those tracks in EEZ around the world. As Andrew pointed out, territorial waters are different, and only extend 12 miles from a coast.
@Viking,
Dr Kristensen, I hope to cover everything in your critique at once.
We’ve had several interesting technical exchanges over the years.
For my reference, your 2018 paper is here, revised Apr 2024:
https://arxiv.org/abs/1811.09315
First, it seems a false dichotomy of possible crash sites to divide between either those previously searched, or yours+mine. There are other candidate areas unsearched on the 7th Arc.
I too wondered how many fishing boats might be nearby the Java site to either witness a crash or spot debris. A major clue comes from “lights at night” satellite imagery. There seem to be almost no fishing lights over the abyssal trench. Certainly few compared to the shallow seas. (Lyne has dismissed the Java event as dynamite fishing which seems unlikely.) Wind driven surface currents would have swept debris towards the Sunda Strait in the first three days, then west away from the coast. It would have been fairly scattered at that point. Looking at live MarineTraffic right now, there are no reporting AIS vessels within about 35 km. The nearest ship to the SW is over 200 km away. No ships spotted any MH370 debris even when on lookout. That negative finding should not be cause for excluding acoustic evidence.
You are assuming that the plane was gliding airborne for an extended time to explain the 56 minute delay between the last ping and the seabed impact. That’s very unlikely. The seabed site is an exact match with the 7th Arc. The plane must have entered the water there. A large section apparently stayed afloat, then sank.
BTW, from the strength of the Curtin event which I believe is a reflection of the surface impact at 7th Arc timing off the 90E Ridge, I suspect that the plane entered the water on a WSW heading, which would be about parallel to the waves.
As “proof” that of that mis-timing you are citing private analyses that you invented, for unspecified signals recordings you obtained from “special arrays of detectors”, and unpublished work by geologists. None of that appears in your 2024 paper, so I don’t know what to make of it. I’ll guess you are talking about either the Pilbara or Warramunga seismic arrays that I have examined extensively. Those are a bit too far away to detect the Java event, unless you are very good at seismic beamforming and stacking. At least 40 regional seismometers did detect the event, which is easily validated by seismologists.
Nearly all of the bearings in Kadri’s paper that you cite are wrong. Most of his bearings in the Nature paper table 1 point back toward the 7th Arc. That is inconsistent with work by Curtin, LANL, and myself. His HA01 azimuth at 00:47:04 that matches yours is not coming from 343.16, but 150 to the SE. Here is a plot that removes signal magnitude by focusing on phase match only, and boosts the noise floor to amplify the weakest events:
https://drive.google.com/file/d/1U5zuXKKwaewnaEAAWwDLRk_Af9O34rmP
The 00:47 event is traveling outside the SOFAR channel, indicated by the splatter, and also stronger on the southern pair than the other two, as indicated by the green hue. That means the event was probably very close to the array.
In looking through your paper, you have several mentions of the seismic doppler plot that I sent you from the XMI and XMIS flyby timing. I thought I was in your bibliography for the earlier versions. Have I been redacted?
One key element that separates our candidate sites is the precision of the search area. You have relocated your site twice by about 30 and then 60 km and now appear to be using contrails to define a 10 km radius. I still can’t differentiate them from similar nearby clouds, but have you accounted for a glide or turns below contrail altitude after fuel exhaustion?
Ocean Infinity contract expiry date: does anyone have a definitive date?
Malaysia government’s agreement in principle was trailed in December 2024 and again in February. The contract wasn’t formally concluded until 19 March 2025. That has widely been interpreted/reported as meaning expiry in June 2026 (18 months from first in-principle agreement). However, if duration is 18 months from formal contract completion, it would run until 18 September 2026. Does anyone know?
@ Victor
This is very interesting.
It might be in OI’s interests to extend the search because if the plane or a major part of it is found, OI would presumably not be liable for all the costs of the search ( unless the Malaysian government defined the current/recent search area precisely)
@370Location
Let me try to answer most of your points quickly. Sorry if I miss one or two – I will come back with more information later.
1) You claim that I ignore a lot of other candidate sites. That is partly correct, but for a good reason. Our two sites are the only ones with corresponding acoustic/seismic signals, and that is the reason they are more likely than the rest.
2) I think it is not a very good method to determine the number of fishing boats by looking for lights at night. Most satellites can only see the strongest lights from fishing boats. Such strong lights are only used for fishing in the upper layers, typically near the coast or for special types of fish. Most other fishing is using echo-detection methods. Navigation is by GPS and radar when the coast is no longer in sight.
3) I do not believe the Curtin event is a reflection of your event. Many papers agree that it originates from a tremor near India.
4) When I wrote the last version of my paper I did not have access to the detailed data from the two arrays. Therefore, I could not cite anything specific. I had only briefly been in contact with the group of Australian geologists at that time. Since then they contacted me again, because they had trouble with parts of the data and they had realized that I might have some knowledge that could help them. I suggested some changes in their analysis. That improved the output significantly. One of these changes was to use the typical propagation speed in the above-jump-layer instead of the speed in the normal guiding channel. That had particularly large impact on the quality of the signal output. They where very happy about the improvements, and sent me the revised output plots from both detectors for a long time-span. Since then we have discussed a couple of smaller details some times. When I some day in the future publish the next version of my paper, I will of course thank them and cite their work.
By the way, both arrays are partly towed. Their main application is looking for new potential oil fields, and that is the reason the owners prefer to keep things secret. The geologists did not give me the exact positions, but since I know what shape to look for, I managed to find them in old NASA satellite pictures. They were not so happy (basically a proof that my positions were correct). However, I am not going to leak the positions.
5) The absolutely strongest signal in both detector arrays come from your site. The signal strength is significantly stronger than from the earthquake (my crude memory was magnitude 4.5, but I am probably a bit off). That means we are dealing with an event with an energy larger than a Hiroshima bomb going pretty much 100% into the above jump layer propagation channel. That is certainly not an earthquake or dynamite fishing. It is also not the crash of MH370 unless it carried a powerful nuclear bomb (I do not believe that). The only thing I can come up with as a plausible source is the impact of a coal condrite meteorite, releasing an energy similar to the event in Chelabinsk (Russia) some years ago. A seabed impact of debris from MH370 would release several orders of magnitude less energy (even if you included an explosion of the Li-batteries on board), and most of the energy would go into seabed R-waves. Practically nothing would go into the above-jump-layer mode.
6) Some of Kadri’s headings are wrong, but not the one we are talking about. In that case you keep making the same mistake (leading to almost 180 degree opposite direction). I am now sure it is related to your method of analyzing the signals. It gives much higher weight to simultaneous refreezing noise from an area near Antarctica. I am sorry that I initially accused you of having interchanged two detectors in the analysis. That was not the case, but the outcome is practically the same.
7) I have not redacted you from any places. A long time ago you sent me a plot which is partly in your homepage too. Later (after I wrote version 4) you sent me a lot more. I am certainly going to cite that in the next version. Please note that you specifically wrote to me that it did not contain any signals. That is true for all the upper layers of your plots, but the lowest layer does contain signals. The reason is that on-shore waves from the south make a lot of noise in the other directions while the E-W has only little noise as expected from METEOSAT7 pictures. Take a look yourself.
8) Concerning the exact crash position there are two possible solutions to the INMARSAT equations corresponding to the two types of Boeing simulations. I am sorry that I initially picked the wrong one, since I made the error of trusting the linearized satellite pictures. Sadly the transformation from the original pictures to the linear ones generated an artificial ‘contrail’ exactly to the wrong side at the correct time (extremely unfortunate). One of the Bloggers here (I forgot who it was) pointed out that possibility. I am very grateful for that (I will try to find out who it was and include a citation in the next version). It lead me to return to the original pictures, where the fake ‘contrail’ was absent. Instead I found a real contrail pointing to the other direction. It also had a shadow on the top of the clouds, so I am sure it is the correct one. That is the reason I jumped to the other solution. By the way it fits Kadri’s heading with extremely good precision. In addition there are peaks (in HA01) corresponding to several other propagation modes and expected routes. The problem is that most of them get assigned to practically the opposite direction due to the same problem with the data analysis (also Kadri’s). Unfortunately I am unable to get access to the original data (still secret with a special permission needed). From just a few weeks later this secrecy is lifted. Is that a coincidence or deliberate? I do not know.
More importantly, just after the correct contrail and its shadow meet, and they disappear in the clouds, a mushroom cloud rises a few minutes later.
@Victor
The very last words in OI’s release are: “when circumstances allow”.
This begs a few questions:
1. What circumstance(s) have DIS-allow(ed) completing the planned inner search rectangle ?
2. Would those “circumstance(s)” preclude any further search by OI under the existing contract ?
3. Would those “circumstance(s)” preclude any further search by OI or anyone else under any Malaysian contract ?
4. Would Malaysia even agree to any new contract with anyone – are they “done” ?
Thoughts ?
@ventus45: The wording is written to provide as much flexibility as possible. The search was halted because of weather and other commercial engagements. As the seasonal weather will be progressively worse, the next opportunity would be late in 2026. I don’t think that the current contract would cover that. I have no idea whether Malaysia would extend the current contract with OI, or negotiate a new contract with another party, if that party even exists. It’s possible we’re done. Given the circumstances, Ed’s candidate site is an interesting possibility.
@Viking,
Martin, I’ve taken some time to check over your claims, investigate the 00:47 HA01 event, and generate some plots. Rather than clutter the forum with a bunch of links, the files are shared here:
https://drive.google.com/drive/folders/1-24UfFcdDHKwdiOnK3DU0u-vRM6hJsu7
I turns out that I had investigated the 00:47 event several years ago because it generated a lot of aliases in the plots. I took copious notes and concluded that it was a quake around bearing 150 from HA01 (not close to the array). I’d pegged the origin at around 00:16:18Z. To confirm that, I compared it to two other quakes in rift zones south of Australia. The 00:47 arrival origin is near an M4.5 quake on Mar 7 at 20:36:36Z. Compare the CorrelationWaterfall plot labeled Kristensen with Quake-az157. The green channel is the correlation between hydrophones 2 and 3 in the HA01 array. The peaks are quite close to the 00:47 arrival. Also compare to another cataloged M4.2 quake on Mar 7 at 22:50:07Z, much farther SE labeled Quake-az145-VikibgII-az324. That quake occurs between airgun shots from survey ship Viking II in shallow water off the coast of India on bearing 324. One more correlation peak for comparison is the Curtin event, the strongest event coming from NW around impact time. Note that the correlation peaks are reversed from the 00:47 event.
I also generated a bearing-over-time plots of the two quakes. Both show the correct azimuths, including for the VikingII airgun shots.
The quake noise is not arriving horizontally in the SOFAR channel, which I assume is what you call “above-jump-layer”. The sound waves are technically not above it, but criss-crossing above and below it at higher speed. Even so, the arrival time is within 5 seconds of the expected SOFAR arrival. It is a bit more complex with a quake, because the origin is usually deep – over 15km depth on these. This causes the early arrival centered on the bearing, because the sound began faster through the crust. The uncorrelated noise is also because the sound originated from a wide area around the quake due to the depth.
I think this is definitive evidence that the 00:47Z event originated from a quake south of Australia. The HA01 timing happens to match your site, but that is a mistake made by several other authors. They started with a candidate site, found a timing match on the CTBTO hydrophones, but the bearings turn out to be incorrect.
It’s a bit confusing that you say your correlation peak calculations match Kadri’s, but that you don’t have the hydrophone recordings.
You say that the Curtin event is not a reflection from the 90E ridge based on other papers saying it is seismic. I too tracked that down years ago. There is no cataloged quake near the Maldives to account for the 301.6 event. I made a tentative match to a hydroacoustic event, but it was based on T-waves arriving at island seismometers. I could find no geological origin on nearby seismometers.
There is a curious feature of signals being reflected off of a coastline or ridge. Careful examination shows that the azimuth shifts slightly during the event. This is seen on the Java event refelcting off the coast, on the ARA San Juan signals, and also on the Curtin 301.6 event. It is not seen on ice events or quakes.
There is no “weighting” to my plots to favor events arriving from Antarctic waters. The various streams you see are numerous small events from ice floes well off the coast and calvings on the coast.
If you see some similar collective noise favoring one direction on Christmas Island, it is likely either waves crashing on shore, or noise from the town. Using a single seismometer to match your crash site isn’t enough to triangulate. I’m using dozens.
I use a completely different hydrophone array algorithm than Kadri and others. Kadri finds a negentropy peak, then cross correlates for another peak to (mis)compute the bearing. Instead, I beamform delays between hydrophones and check every bearing at every instant in the plot. That only works after calibrating the hydrophone spacings against reference signals from all directions, but it also allows dozens of different algorithms for measuring signal and phase alignment.
The Java event arriving via the SOFAR channel was absolutely not caused by a Chelabinsk sized event or a nuclear bomb. This may also come from Kadri, who proposed a meteor impact for an ice tremor. You paper also includes his confusion over military ops instead of known survey air cannons. See my site for links to a report where an sea bottom instrument battery pack exploded off the New Jersey coast and was easily detected at Ascension Island thousands of miles away. The signal was clearer than the ARA San Juan implosion. Part of the explanation may be Scholte waves. These may be what Kadri calls AGW. Scholte waves are strongest at the fluid/seabed interface, and propagate along the seabed at speeds between the fluid and crust speed. An event at the seabottom might propagate that way to a nearby coast and then experience efficient downslope T-Wave conversion at the usual depth of around 200m where that happens.
Thanks for clarifying that you have the output plots from towed arrays, not the source data. I was also contacted by an Australian with a plot from a towed grid array of 6,000 high bandwidth sensors. The processing power to crunch all that data must be immense. He detected the Java anomaly reflected off the Java coast. From his accurate bearing and timing I also located the ship. He also did not detect a surface impact.
Well, the opportunity for media to cover the Java candidate site hasn’t manifested so far. Perhaps it’s too dry and technical, or they are put off by peer review, or the experts they checked with pooh-pooh’ed it. Let’s hope that OI will take it seriously based on Victor’s recommendation.
@370Location said: Well, the opportunity for media to cover the Java candidate site hasn’t manifested so far.
Media coverage was never expected or intended, and won’t determine whether or not your candidate site is seriously considered for a new search.
@370Location
I disagree with most of what you write in your latest contribution with a few exceptions.
1) I do not disagree that the 00:47 peak may contain smaller contributions from one or more other sources. However, the dominant contribution is propagating in the above-jump-layer channel. Relevant earthquakes (for MH370 investigations) typically put most energy in seabed R-waves and/or the standard (SOFAR) guiding channel. The mode-type you describe is not identical to above-jump-layer propagation. Instead it resembles propagation in an inner cladding around the SOFAR channel (or just a higher-order mode in the SOFAR channel itself). Such modes also exist in certain optical fiber (and planar waveguide) types. In particular in optical fibers used to make laser weapons. You can find the effect described in a couple of papers published by University of Southampton (UK) 20-30 years ago, before their research got military funding. Please also note that seabed R-waves and above-jump-layer waves have opposite signs of their dispersion. The absolute sizes are comparable (within roughly 20%). The dispersion of the dominant contribution to the 00:47 peak has the same dispersion sign as an above-jump-layer wave, and the integrated dispersion is consistent with the distance to my proposed crash position. How can you explain that as a signal coming from an earthquake?
2) I do not have the original HA01 data, so my own estimates are exclusively made by eye-inspection of published plots as I told you before.
3) Intuitively I would say the Curtin event is not a reflection. One reason to say that is that I found no other reflected signals from that ridge, while I found plenty of reflections from the Sunda trench and from the west coast of Australia. A reflection would also fit poorly with the correlations found by Curtin.
4) Kadri’s method is pretty standard. Yours is alternative (not wrong). In some cases it may be better than his in other cases his is best. However, both methods have troubles with peaks composed of more than one contribution. There are ways to get around the problem (as I briefly described in a contribution from last year), but none of you use such methods. Eye-inspection is actually not so bad as long as the plot is made in an unbiased way, and the contribution you are looking for is the dominant one.
5) I take 100% responsibility for the Chelabinsk proposal. It did not come from Kadri. However, I know he proposed meteorites in other cases.
6) If you write clever software, the processing of 6000 signals does not need a supercomputer.
7) The data plots I received from Australia showed clear signals from at least one ship. However, it could not have been the one you are talking about. I did not look for weak shipping-related signals.
@Victor,
Thanks for your clarity. Sometimes it seems like media coverage and popularity drive the MH370 search. WSPR and 4orbs are clear counterpoints to those cynical thoughts. I’ll try to stick to the technical analysis.
@Viking,
Without knowing which plots you are eyeballing for deductions, I can only guess about some of these claims. If your above-jump-layer means surface ducting above the thermocline, then that’s a very thin layer which may not support long distance propagation of low frequencies. The detectors would need to also be in that layer. The towed surface arrays definitely pick up ship propellers and might be ideal for detecting the surface impact, but the clear plot I saw didn’t. Perhaps as you say they are not tuned for surface sound velocity. They may instead typically suppress it as clutter.
I haven’t investigated the dispersion direction for surface waves. However, I’ve added two custom spectrograms to the batch of files above. One is for the M4.5 quake on azimuth 157, and the other is for your 00:47 arrival nearby on azimuth 150. The frequency dispersions are clearly in the same upward direction. If you are saying it should be the opposite for your above-the-jump deduction, then it isn’t borne out in the data.
The way Kadri described his method, he would pick the strongest correlation peak. LANL may have done the same, as they dismissed HA08 as too cluttered to be useful. Curtin picked multiple peaks and IIRC selected based on expected horizontal slowness. My method shows ALL the peaks, even the weakest. If it has a fault, it’s that it is too finely tuned to distant signals arriving horizontally in the SOFAR channel. Some of my algorithms detune that, which allows seeing local noise like ships passing the array. My plots have another advantage in that each pair has its own R/G/B color channel. Distant SOFAR events are white in color and persist over time. Signals and noise that are mismatched between the pairs show different colors or are very short duration.
As for clever software, my specialty at startups in the 70’s to 90s was developing optimized algorithms for wimpy CPUs and custom hardware, from the 8080 up into the gigaflop FPGA era. That included image capture devices and signal processing. My patent was used by Canon as prior art in defending their use of jpeg in a camera. Two of my early mentors were prominent in the MIT TMRC. I still love an elegant hack 😉
My apologies if I seem to be dismissing your candidate site. I hope it also gets searched. I may get a bit defensive about vague claims that my work isn’t technically sound, and only mean to counter your arguments.
@370Location
Please note that I am not claiming that your position is wrong or that nothing interesting happened there. On the contrary, the most serious problem with your site is that the energy release is too high.
I am certainly in favor of searching your position for two reasons. Most importantly the (per area) probability is high because the area is extremely small. Secondly, it would be interesting to find debris from a meteorite on the seabed, or alternatively from something else with a high energy release (Geiger counter recommended in that case).
My fear is that OI searches your position without finding debris, and then gives up for ever. I do not want the case to end like the failure to solve the murder case for the former Swedish prime minister Olof Palme.
I have several remaining reasons to believe that my position is correct. The strongest of these I have never discussed in public. It can potentially sweep everything else off the table. If you want to get an impression of what it may be, I recommend reading the latest publication about the Olof Palme murder written by two Danish journalists after 20 years of research. It was published in Danish last week, so you may have to wait a couple of months for an English version.
Finally. some more specific points:
1) The jump layer is typically around 13 m thick near Denmark. However, in places like the warmer parts of the Indian Ocean it is several hundred meters thick because of high temperatures and a lot of sunshine from a high elevation angle.
2) The above-jump-layer is actually anti-guiding. That is the reason it has opposite sign of the waveguide-dispersion (material dispersion is negligible). The fundamental mode of the SOFAR channel has extremely small dispersion, so that can normally be ignored. Higher-order modes have dispersion, but I found no theoretical estimates of practical values.
3) You are correct about the outcome of using incorrect speed values for the analysis of the towed array-data. The first plot I received was of that type, and the signal strength from your event was significantly weaker. However, such a high-energy event cannot be suppressed entirely so it was still visible.
4) I estimated the dispersions from Kadri’s frequency diagrams and compared with theoretical values. The agreement is fine.
5) You are right that Kadri’s method catches the heading of the strongest contribution for dispersion-free peaks. However, that is not the case for peaks with dispersion. He has admitted that to me for seabed R-waves, but not yet for above-surface guiding. However, I am sure I will win in the end, because of symmetry-arguments.
6) For the same reasons, you may actually get a wrong dispersion out of your calculations when the software finds 150-degree heading. The problem is that the distance between the two western detectors in HA01 is somewhat too short for discrimination (a more detailed and lengthy explanation is needed, but the problem originates there).
7) For eyeballing I compared different plots of the same signals. The outcome differs significantly for seabed R-waves, but only very little for all other signals including the one at 00:47.
@370Location
Sorry for two small typos:
1) Finally. should read Finally,
2) above-surface should read above jump-layer
@All:
K.S. Narendran voices a sharp opinion on the good, the bad, and the ugly of tracking down MH370.
https://linesabouttimes.com/2026/03/13/mh370-search-bounty-remains-on-offer/
@Viking,
I’m not concerned about typos, but repeating refuted claims as fact. It’s as if you have not examined the signal plots I generated for you:
https://drive.google.com/drive/folders/1-24UfFcdDHKwdiOnK3DU0u-vRM6hJsu7
The numerous ice events arrive at the CTBTO arrays with very large dispersions. Papers have been written on distance determination without an array based on that frequency shift. The SOFAR channel approaches the surface near Antarctica due to temps and salinity, so the signals are generated IN the SOFAR channel. As I’ve noted to you in the past, the dispersion does not shift the bearing. It doesn’t make sense, because the bearing is determined only by the Time Difference Of Arrival between the hydrophones when they are close enough together that the signal has traversed essentially the same long distance path to each one. Ditto for the airgun blasts. They are dispersed, but the bearings are very accurate to the known AIS location of the various ships. There is no azimuthal shift during the frequency shift.
The magnitude of the signal arriving at the arrays is an indication of how close to the SOFAR channel they propagate. That is why even the smallest ice events appear as a steady stream. Earthquakes generate energy on the order of nuclear bombs, but the sound source originates outside the SOFAR channel, making it relatively weak. That off-axis arrival causes a mismatch and slightly shifts the bearing. It makes the signals appear locally across the array to be faster than the speed of sound. Yet the two quakes I plotted are from cataloged location and time, and the signals arrive within seconds of the expected SOFAR speed.
We are not talking about seabed R-waves. Rayleigh (surface) and Scholte (seabed) waves do not propagate at SOFAR sound speed, so they would arrive in a different time window.
I’ll repeat too: The characteristics of the 00:47 HA01 arrival closely match the signatures and direction of the cataloged quakes I plotted for you. It appears to be a weaker uncataloged quake, not a surface impact near Christmas Island.
As mentioned before, surface events over deep ocean tend to travel in concentric convergence rings, separated by shadow zones. They might be detected if a hydrophone were in the narrow convergence ring, but the odds are slim. That is evidenced by my attempts to match arrival timing with a database of millions of lightning strikes. However, when reflected by a long ridge or coastline, where the convergence zone favorably intersects the terrain, the signal can be injected into the SOFAR channel.
Alec Duncan co-authored a book with chapter 6 titled, “Introduction to Sound Propagation Under Water”. Page 202 describes convergence propagation from a surface event. It is openly available online at:
https://www.researchgate.net/publication/364149037_Introduction_to_Acoustic_Terminology_and_Signal_Processing
I have the greatest respect for Alec. He and the ATSB encouraged me to study and pursue the acoustics related to MH370. It unfortunately took me until 2018 to connect the dots, and it’s been quite an education. Had this info been known early on, I am sure the Java candidate site would have already been searched.
@370Location
Interesting link.
I fear Malaysia does not really want to find MH370. The reason is unclear, but it could be that solving the mystery is likely to bring up inconvenient information.
That is the reason I made the comparison with the Palme murder. According to the Danish journalists the Swedish police deliberately messed up the investigation to avoid finding an inconvenient truth.
@370Location
I could not resist the temptation to dig a step deeper in your Java event. After some trouble with the timing I managed to find an atmospheric feature at exactly the correct position. It is in an infrared (IR) picture taken by Meteosat7 at 1:00. At that time there is a nice symmetric ring with roughly 60 km radius around your location, corresponding to a sound event happening roughly 3 minutes before. It may actually have been a bit later, since I have the impression that the recording of the IR pictures took somewhat longer than the visible pictures.
The only plausible explanation is that something passed very fast through the atmosphere at almost 90 degrees sending a symmetric pressure wave out in all directions at the speed of sound. It was certainly not a nuclear explosion, since the heat would have produced a mushroom cloud instead. It looks very much like the Chelabinsk event. There they found (somewhat later) that the core of the object contained a lot of iron. If this object was similar, your feature could well be the iron core (10000 Ton or more) hitting the seabed (and explaining why the event was a couple of orders of magnitude too powerful and happened with a delay). It would be a nice finding (scientifically), but I do not think OI would be particularly happy to find such an object at the seabed.
The bang from the atmospheric part of the event must have been very loud at the coast of Indonesia, and easy to hear at Christmas Island. Did you try to look for such a feature in the seismic signals?
PS: I actually found a very small feature at Christmas Island corresponding to atmospheric sound originating at the correct time from my position. Yours should be extremely easy to find, but I do not have data for such late times.
@370Location
The Chelabinsk event is now estimated to have liberated an energy 30 times larger than the Hiroshima bomb. That puts things in the correct perspective.
https://www.dropbox.com/scl/fi/d3hzcqae5lsed8sdq7s5y/A-Voyage-to_v3-031426.pdf?rlkey=xyezdfi801hsaxl5giawnpuxr&dl=0
https://www.dropbox.com/scl/fo/01d6bci78sddcaiy9ufrd/ALzPuV3ZzgVk0FZFPfd0JqY?rlkey=rdlxlwehnna5qc1v1mhguy3if&dl=0
Hello all,
These are links to an update to my paper “A Voyage to…” While there are no new conclusions or computations, I have clarified the definition and use of the “cost” as a heuristic in adjusting the few relevant model parameters. The other is a folder in which I have posted many of the references, including several that are not specifically cited. I will continue to add references to the folder as I become aware of them.
If any of you have access to reasonably capable flight simulators, perhaps you can try to reproduce my geometric path model. Since I don’t have such a capability, there may be some corrections needed to the scripts I have provided.
I hope that I have addressed the points that have been raised about previous drafts. The link to the spread sheet is in the paper.
@Sid Bennett,
Regarding your cost function for route fitting, I have some comments and suggestions :
1. The probability is not maximized when the observed RMS equals the population sigma. That is because the expected value, with a small number of data points, is not sigma. There is a bias to a slightly smaller value than sigma for both the expected value and for the median. At the median, half the random trials with no route errors present would be lower, and half would be higher. The median of the RMS population is therefore a good choice to use as a “maximum probability” value, which is the goal of the route fitter.
2. The probability is not simply related to the ratio of RMS values, so your cost function is not a probability surrogate. It’s alright to use a cost function in route fitting, but you can’t easily compare the likelihoods of two routes with different values of cost function. If you want to do that, you need to compute the probability the route is consistent with no route errors, and maximize that value in the fitter. Doing that, you can directly find the relative likelihoods of different routes by taking the ratio of their estimated probabilities.
3. The BTO and BFO probabilities are conditional, and so you should use their product to estimate the “satellite data” probability, not a quadrature sum of their inverses.
@DrB
I appreciate your comments and I am sure it would result in a tidier conclusion. But, for a decade I used the heuristic to “optimize” the few parameters needed to characterize the model.
It led to an accurate determination of the FMT time and an accurate determination of the initial azimuth after the FMT. It is not being used to compare one kinemetric model with another.
The real question is whether the congruence of the geometric model and the kinemetric model is sufficient to convince you that this is a correct hypothesis (recognizing that it will only be confirmed by the wreckage).
I am sure that there are other little errors in this work. I am kind of “sloppy” in looking at big problems, but I am usually close 🙂
@Viking,
You’d think someone would have noticed and event the size of the Chelyabinsk bolide. It entered at over 100K ft and shattered windows up to 55 miles away. At 30x brighter than the sun, it was seen over 150 miles away in broad daylight. The sonic booms and airburst were reported over 500 miles away. The infrasound pulse circled the globe twice.
Before the realization that MH370 had turned south, the CTBTO analyzed an infrasound array in Japan for any signs that MH370 exploded, and found nothing. Recovered fragments of the Chelyabinsk core weighed around 1 ton, not 10,000. The mass of MH370 debris would be nearly 150 times greater.
There was a weak event on seismometers 6 minutes after 7th Arc timing, then the larger event at 56 minutes. I suspect the first was a sinking engine core hitting the seabed.
I haven’t seen any of the features you describe on satellite images.
Using a meteorite as an alternative explanation for the Java anomaly is a fantasy that you can pursue as you wish.
Comparing MH370 to the Olaf Palme murder goes into conspiracy theory speculation, which I don’t see as helpful for locating the plane.
@370Location
As far as I remember, the infrasound investigation (and another one using satellites) effectively only covered the time interval from the airplane stopped communicating (ACARS, transponder, etc.) until a couple of minutes after the last handshake. I think the Java effect may also have been smaller than the effect for Chelabinsk, since the recording took place in the opposite hemisphere and on a close to vertical impact, while the Chelabinsk meteor came in at a small angle travelling from Alaska to Chelabinsk within the atmosphere (on the way passing close by Japan).
I also presume the Java meteor had a large solid iron core, and a relatively small outer loose shell compared to the one in Chelabinsk, which was opposite (small core(s), large shell).
It is not so strange that the recovered fragments weighed much less than the original meteor. It is always like that. Even when a small meteor fell near the center of Copenhagen a few years ago, where millions of people searched for pieces. Less than 0.1% of the weight was recovered.
I think you should read the new book about the Palme murder before calling it conspiracy comparing the two cases.
@Viking,
Let’s not devolve into whaddaboutism. A purportedly vertical entry of MH370 due to BFO is nothing like a fictitious bolide grazing the atmosphere.
Instead, the Java anomaly is about dozens of seismometers registering a seabed event directly on the 7th Arc near Java, with all available hydrophones detecting it. The epicenter of the seabed event is exactly on the 7th Arc. The seismic precision is better than the SATCOM. No other candidate comes close.
OI has made it clear that they have concluded their search rather than expanding the previous area where they have been battered by rough seas. They have already spent uncompensated millions searching for MH370 since 2018. Before their $70M contract expires, there is an opportunity to survey a spot in relatively calm waters that is consistent with new and old factual evidence. I hope they will not be deterred by speculation.
@Sid Bennett,
I also think that a simple route was flown after MEKAR, so on that we agree, but identifying that route has proven difficult. Here are my thoughts on a few key issues.
1. After 12 years, we have nothing “positive” to show except physical debris discovered in far-off places. The prediction of crash latitude based partly on the finding locations and especially on the arrival times is uncertain. In almost all cases the arrival date is unknown. Only the discovery date is known for some debris, and even that is unknown for many debris. A major contributor to the crash latitude uncertainty, even with a good drift model, is the assumed or estimated windage and leeway of the debris. Drift model analyses are highly useful in confirming the general crash location in the SIO, but they may not be highly accurate in locating the crash location. We have no prior example of this being done accurately for aircraft debris over long distances and long drift times. So, at best we can say the reverse drift predictions are unproven in this situation. We don’t yet know if they are accurate. Time will tell.
2. On the other side of that coin, on the “negative side”, no debris was found in Western Australia. Potentially, that could help eliminate sections of Arc 7. However, the large uncertainty in predicting the probability that passers-by will beach on that coastline, be found, and be reported makes it impossible to be certain about whether a given section of crash latitude on Arc 7 would or would not result in debris reports from WA. Again, using the lack of reported debris in WA is an untested and unproven method of predicting crash location.
3. Also on the “negative” side, the debris field has not been found by the seabed searches. I believe all satellite data and debris indicate a high-speed crash near Arc 7 with no piloted glide after fuel exhaustion. So, considering the zone within about +/- 25 NM from Arc 7, latitudes from circa 39S to circa 33S have been thoroughly searched without success. By thoroughly, I mean that 95-98% of the area was declared not to contain a debris field.
4. The negative seabed searches are highly informative. They have told us, with a very high probability (> 95%), the crash did not occur near Arc 7 within that 33-39S latitude range. That means that your prediction of 37S and the UGIB prediction of 34.2S are both almost certainly wrong.
5. That leaves us with two choices: (a) the aircraft crashed farther than 25 NM from the arc in a place that has not yet been searched, or (b) the aircraft crashed near the arc but outside the 33-39S searched latitude range. Believing that the crash is much more likely to have been near the arc, I choose the second option as being more likely. In my opinion the debris field is near the arc but outside 33-39S latitude.
6. The area near 39-40S has been proposed many times, including by the original Independent Group in 2014, by me in September 2014, and recently by Paul Smithson. However, that crash latitude fell into disfavor over the years for two reasons : (a) it could not be seamlessly connected to the derived FMT Route, and (b) there was not enough fuel on board to reach it. The second objection was actually the basis by which ATSB abandoned their initial search zones S1 and S2 early on and moved up the arc to S3. However, that decision was based on the ATSB’s assumption that the aircraft had not slowed after 18:22. From 17:24 to 18:22 the air speed derived from radar data was 500 KTAS, which is M.85 at FL350. There definitely was insufficient duel to fly at M0.85 until 00:17:30. However, there was enough fuel to fly at M0.85 from 17:24 to 18:24, and then to reduce the speed to LRC.
7. For the past five months I have been re-examining these objections to a circa 39-40S impact, and especially their underlying assumptions, to see if and how they might be overcome.
8. First on my list was to build a better fuel model which would allow exploring a range of air speed control options. That task is now complete. I was able to build a generalized fuel model which is simple in conceptual terms and highly accurate. I am now able to predict the fuel flow for any combination of 9M-MRO weight, flight level, SAT, and Mach, within about 0.5% (compared to Boeing flow tables for LRC, M0.84, and Holding). This new fuel model has three areas of improvement. First, it incorporates lift and drag coefficients for this aircraft type at a given Mach to predict the thrust needed to equal drag. Second, it accounts for the substantial non-thrust-producing flow. Third, it accounts for the effect of air temperature and Mach on the flows for the thrust and for the auxiliary loads and engine losses. I now am able to predict the flow for speed modes such as MRC, ECON with Cost Index = 52 (the MH370 flight plan), and M0.85.
9. As I showed in Case 9 in UGIB (2020), the known MEFE time of 00:17:30 can be closely matched if the post-18:22 route is flown at LRC with no major descent, but with some fuel flow reductions. The fuel savings required to match MEFE are two : (a) the air packs are turned off at Diversion and never turned back on and (b) there is a small reduction due to electrical equipment being turned off at Diversion. My new fuel model confirms this result. It also quantifies the electrical power reductions needed to be roughly 50 kW per IDG (out of a typical assumed load of about 70 kW).
10. Thus, it is possible for 9M-MRO to have flown at LRC until 00:17:30, but not with a functioning pilot at the end of the flight. That would have been prevented by cabin depressurization for seven hours.
11. It is also possible for 9M-MRO to have flown until 00:17:30 with a functioning pilot aboard the whole time, but not at a speed comparable to LRC. That lower speed necessarily leads to far-NE crash locations, but none of these match the satellite data without multiple maneuvers after 19:41. This complication makes it less likely, in my opinion, and also impossible to predict with sufficient accuracy to enable an affordable seabed search.
12. That leaves the SW end of Arc 7 circa 40S as a potential contender, if the previous objections can be overcome. We know the aircraft could fly at LRC until 00:17:30 (if unpiloted at the end). So, what does it take to reach 40S at LRC? The answer to that question is a very early turn southward from the vicinity of MEKAR, with no lateral offset maneuver. Could that have happened?
13. What drove the insertion of a right lateral offset maneuver circa 18:25 (as shown in UGIB)? The answer to that question is the apparent discrepancy between the 18:22:12 military radar position report and the 18:25 – 18:28 BTOs. For those two data sets to be consistent with no maneuver in between, one of two things must be true: (a) the radar points are too far west by at least 15 NM or have time stamps which are more than 2 minutes too early, or (b) the BTOs then are 100 microseconds too high. Neither of these seemed likely, so in UGIB we inserted a maneuver between 18:22 and 18:25 that would resolve the discrepancy. That was a 15 NM right lateral offset, the purpose of which is unclear.
14. I have recently re-analyzed the BFO data at 18:25 – 18:28. Now I have an improved time alignment which allows compensation of the thermally-driven frequency transient error for all the 18:25 – 18:28 BFOs. This indicates all BFOs then are consistent with a course close to that of N571 with no ongoing maneuver. This new finding further constrains the lateral offset maneuver to be completed prior to the first BFO at 18:25:27. This maneuver takes 3 minutes to complete, and for it to resolve the discrepancy between the radar points and the BTOs means that it must have started after 18:22:12 and completed before 18:25:27. In other words, this 3-minute-long maneuver has to fit in a 3.25 minute wide window. This seems to me to be a highly unlikely event, especially when one understands that the power restoration occurred in the middle of this time window circa 18:24:00.
15. Now we have three seemingly unlikely events to explain the radar/BTO discrepancy : (a) the radar data are very wrong (by 15 NM or 2 minutes), (b) the BTOs are slightly wrong (by 100 microseconds), or (c) a lateral offset maneuver surprisingly appeared in a very tight time window. DSTG opted for (a) and ignored the military radar data after 18:02. UGIB opted by (c), the offset maneuver. What about (b)? Could it be correct?
16. The implication of choosing (a) or (c) is that the starting point for the SIO Route at the FMT is moved by the equivalent of 15-30 NM compared to the best alternative to reach 40S. That is enough to make a difference.
17. I will now select the third choice – that, somehow, the BTOs at 18:25 – 18:28 are too high by 100 microseconds. I do that because the other two choices led to predicted crash locations which we now know are very probably wrong. Option (b) is now, therefore, the only viable option to reconsider.
18. How could the BTOs be 100 microseconds too high, for several minutes? I don’t know. What I do know is that this SDU start-up was unique in two ways. It was done in flight. It was done when the SDU was very cold. There are no 9M-MRO data to assess what the SDU does in either of these conditions. None. So, whenever anybody says they expect the SDU in this case to behave the same as a normal start-up on the ground at normal room temperatures, they are making an untested and certainly unproven assumption. That assumption could be wrong. We know the SDU was cold because the warm-up took longer than usual. It took so long, in fact, that the Built-In Self Test was completed at least 40 seconds before the OCXO reached operating temperature. In every other start-up, the first transmission occurred when the the BIST was completed, the OCXO having already reached the operating temperature. The longer oven warm-up for the 18:24 start-up is a direct indication that the oven was colder than normal when power was restored. This is consistent with the air packs being turned off and the outflow valves being opened at Diversion an hour earlier. The air in the cabin would have dropped in pressure and cooled down considerably, so the SDU would have been getting colder for the hour before 18:24.
19. So, I start my new route assessment by ignoring the 18:25 – 18:28 BTOs, since the other two options have led to search areas in which the debris field was not found. Thus, there is no need for a lateral offset circa 18:22.
20. Then I look at what happened at MEKAR. Was there a turn toward NILAM? That is also an assumption that could be wrong. In fact, the radar data appear to indicate no turn at MEKAR, but rather a continued course of 286 degrees, unchanged from the bearing from VAMPI to MEKAR. How could that happen? Well, it’s easy to do if you just put VAMPI and MEKAR into the waypoint table, rather than all of N571. Then when you get to MEKAR, a route discontinuity occurs and the lateral navigation thereafter holds the same magnetic heading. Basically, the aircraft flies right through MEKAR with no turn before or after.
21. Now, you only need one turn circa 18:31:45 to 188.9 degrees initial bearing with no maneuvers thereafter. This ends up near 39S on Arc 7. One way of accomplishing this is to insert a DIRECT TO waypoint of 45S85E.
22. Such a route has very high BTO and BFO probabilities, both peaking near 38.7S, and there is sufficient fuel (within a fraction of one percent of average predicted flow).
23. This route PDF is quite narrow (only about 1.5 degrees in crash latitude) because (a) it requires the route be seamlessly connected to the 18:02 location with no delays and (b) it allows only one turn after VAMPI.
24. There may be a sliver of this route PDF which lies just beyond the SW edge of ATSB’s searched area. Could ATSB have just missed the debris field circa 39+S, by only a few miles?
@370Location
I am not trying to prevent searching of your site, but there are simply some things which do not add up, if it were a plane crash causing the signals. As an example the time between your weak signal and the strong one is roughly 50 minutes.
Experiments with large cannon-balls indicate a sinking time to the relevant depth of 5-20 minutes. It is a large interval, but if the time is around 50 minutes, it must be smaller pieces with moderate density. These would absolutely not be able to produce a large signal when hitting the seabed. The observed signal is extremely large, so that story is simply impossible.
However, the cannon-ball interval is consistent with the time I come up with (slightly less than 18 minutes), if the iron piece(s) is/are irregular in shape and with rough surface. That is possible.
On the other hand, I have no idea what caused your weak signal.
I suggest you look for the atmospheric sound signal in the nearest seismic detectors.
A 60 km radius pressure wave does not come from a small event. The only comparable one I remember filmed was the explosion of Fukushima-1. After the earthquake in Japan I was just waiting for it to happen. I sent out warnings to all friends in the area to get away to safe distance 3 days before the explosion.
@DrB
Your comments are cogent and I will spend some time reflecting on them before I respond. I am glad to see some movement from the state of discussion that we have settled into. May I also suggest that if some of the route was programmed by waypoints, VNAV could have been used to do a step climb. I am assuming that this is a very experienced pilot and he was trying to go as far as he could.
I should mention that I an now re-looking-as the EOF (2nd flame out) and should have a draft for comment within a week.
Nicely put and explained Dr B.
Couple of thoughts. The SDU getting so cold would require a cabin temp where arctic clothing would be needed, donned while wearing an oxygen mask. Certainly the flight deck temperature could be lifted and some pilot foot etc direct heating applied but that would entail some engine air bleed and added electrical power, costing fuel.
While alternatively the pilot might have set subsequent course changes before IGARI were he to have elected to self-hypoxiate then, the change to LRC you envisage at the turn south would be needed still.
Observing fuel-stretch sensitivity there will be extra drag from open outflow valves. Also there is any difference in drag from pitch and yaw trim optimisation, if un-piloted, from what that your modelling assumes.
Even more finickity there is the accuracy of engine fuel efficiency correction, commonly treated as invariant across the power range and compressor blade cleanliness plus how accurate some other general assumptions are. Fuselage skin cleanliness, fuel actual calorific value and actual passenger weights with carry-on baggage being others).
Even the differences in aircraft weight resulting from velocity relative to space that I had a go a while back, might intrude.
Of the above, based on the effect of RAT intake drag has on fuel consumption (commonly overlooked when considering maximum glide range, APU running), I would guess that the cruise speed drag of opened outflow valves might have the most appreciable effect.
What LEP longitude please?
Second last para, first line: ‘APU’ vice ‘RAT’ pse.
@DrB
G’day Bobby,
Thank you for the outstanding first principles-esque review, and thank you for couching it in terms that we punters in the cheap seats can understand.
In the spirit of “no sacred cows”, there are two other assumptions that have been critical in shaping the search areas that should almost certainly also be reviewed/re-examined. They are:
1. That the final 0019 UTC log-on request was related to a depowering/repowering event associated with fuel exhaustion.
2. That the BFOs derived from the 0019 UTC log-on request indicate a rapid, and rapidly increasing, rate of descent.
On point 2, as you noted, there are no 9M-MRO data to assess what the SDU start-up looks like when initiated in-flight. In the case of the 0019 UTC log-on request, we have an SDU undergoing its second in-flight start up.
Regarding the notion that the debris field has been just missed in the course of the prior search campaigns, I’ve been of that opinion for some time. On the basis that missing the edge of the field by just a metre means missing it completely, I suspect (without evidence) that it will eventually be found tantalisingly close to a previously searched area.
Any old how, thank you again for kick-starting going back to the drawing board.
@Sid Bennett
RE: “May I also suggest that if some of the route was programmed by waypoints, VNAV could have been used to do a step climb.”
A step climb point entered against a waypoint only affects the performance predictions, it does not cause the aircraft to climb. The climb cannot be executed without further pilot-action, even if VNAV is engaged.
To execute a step climb, the pilot must select the higher altitude in the MCP altitude window and then, either:
a. Push the altitude selector to execute the modification, or
b. Enter the higher altitude in the VNAV CRZ page and press EXEC to execute the modification.
In both cases the aircraft will climb immediately, not at some other defined point further down track.
Step climbs cannot happen if we assume the pilot was incapacitated or dead.
See page 6-35 of the Honeywell Pilot’s Guide – Flight Management System (FMS) for the Boeing 777, Rev. 1, July 2014.
@Viking,
You are again jumping to conclusions while clearly missing prior discussions about the Java anomaly.
At the top of this very page Victor quoted my open response which says:
“The timing is 56 minutes after the last 00:19:37Z 7th Arc ping, which implies that the large section stayed afloat for over half an hour.”
One dense piece of debris that would sink very quickly is the engine core, which is on a breakaway pylon intended to detach in a crash. Turbine blades might cause some drag, but they would also trap entrained water between them, reducing drag. The 5+ minute
estimate is about what I calculated for terminal velocity to 3,400 m depth, depending on drag. Also, the algorithm that detected the weak signal may have had an inherent delay. It appears the weak event was at around 00:27:04Z which is about 7.5 minutes after the last ping.
The conversations you missed were about reports by Tom Kenyon estimating the energy of a wings-forward fuselage section of the plane impacting the seabed. He estimated only empty mass of the fuselage at 55 tons but unfortunately he omitted the weight of entrained water both inside and out. The internal pressure volume of intact 9M-MRO is over 750 cubic meters. The front third would contain over 250 metric tons of seawater. The total mass would be over six times the value Tom used to find a Magnitude 0.7 seismic equivalent.
As I wrote above, the hydrophone signal strength is more related to how well the event is conducted into the SOFAR channel. We can assume that most of the wreckage ended up on the seabed, and not all at once. Yet only two signals stood out. The noise level did go up around the time of the first weak event, but it would be more difficult to assign that noise to an epicenter.
I used a novel method to find the two signals seen on the regional seismometers while eliminating environmental noise sources. The recording from each detector was shifted by the TauP arrival delays for P and S waves from the anomaly epicenter. That aligned them all to origin time. An STFT was applied essentially making a spectrogram for each recording. Then NMF (Non-Negative Matrix Factorization) was used to decompose the STFT
into the fundamental frequency components (basis) and when they occurred through time (weights). The weights were then sifted to exclude isolated noise, and stacked to show the signals common to all. The key basis function shows a narrow frequency peak at 1.2 Hz common to both events.
So, essentially I have already done the search for your imaginary bolide.
The sonic boom from a bolide would take about 90 seconds to reach the surface. The fall of the core would take about four minutes, then 5+ minutes to sink. I’ll leave it to you the determine the odds that an invisible and silent huge bolide 1/300 the mass of a sinking MH370 fuselage fragment could land directly on the 7th arc at just the right time to mimic a seabed impact.
Unlike you, I rarely use the word “impossible”, but I think it may apply to your bolide.
@Viking,
Another note on the delayed seabed impact. A large floating section would also take time to reach terminal velocity as the entrained air became compressed. Sinking ships accelerate as they go down, as can be seen by the speed of the last part to go under, but eventually reach a terminal velocity based on their drag coefficient.
@370Location
I can only conclude that we disagree on many details of the interpretation of the Java event. First of all, I do not believe that a very large section of the airplane containing one or more large and high-density parts would have survived the crash in one piece, and certainly not stayed afloat for a significant amount of time. I may be wrong on this point (it is not impossible), but the probability is low.
Concerning the fall towards the bottom, the water inside should be counted if it is well confined, but on the other hand the weight of the total volume of water corresponding to the volume of the falling object should be subtracted (according to the law of Archimedes). This means that the energy released upon seabed impact must be relatively small. That is in serious disagreement with the very large signal recorded.
Using Archimedes correctly, would instead reduce the estimate to around 0.4 seismic equivalent. I think even that number is too large. The signal dwarfs a magnitude 4.5 earthquake. This means 4-5 orders of magnitude disagreement (at least). Most of the signal recorded near Australia did not arrive in the SOFAR channel, so if you are right, it would have gone through at least two mode-conversions on the way. That does not happen without significant intensity loss, so the disagreement may be as much as 6 orders of magnitude.
Indeed I agree that it would be unlikely that a meteorite fell on the 7’th arc almost simultaneously with MH370, but it is not impossible. There are more of these impacts than previously believed.
What other event than a meteor (excluding MH370 and a nuke) could you imagine causing such a large signal? The only other thing I can come up with would be the sinking of a nuclear submarine. I am quite sure we would have gotten some information about such an event if it happened. The last one I can remember with a large seismic signal was the Russian Kursk. I have a faint memory that it produced roughly a magnitude 2.5 signal (without release of any nuclear energy, but including the explosion of a couple of conventional torpedo’s with a wrong design), so it is not completely impossible.
Perhaps OI could get funding from the US Navy in case it were a sub from China in the same way as somebody got funding to find Titanic on the condition that they found a missing Russian sub first?
Finally, I am happy that you provided the exact time for the weak event. It may help identifying its source.
@Andrew
https://aerospace.honeywell.com/us/en/about-us/news/2019/10/inside-fms-step-climbs-capabilities
Please let me have your assessment of this article.
@370Location
I think I have located the small signal in both the (partly towed) detector arrays. The times and heading data from the arrays indicate that it happened roughly 270 km east of the large signal if I use your latest time.
If your previous time was correct the distance would be roughly half as long.
I can go to highest magnification and do complete triangulation if you want me to. I guess both numbers would shrink a bit (perhaps 10%).
However, I am quite sure the origin of the weak signal is different. The strong event fits extremely well with your data.
@Sid Bennett
The article refers to Honeywell’s NG FMS. The B777 uses a ‘legacy’ Honeywell FMS that does not have an automatic step climb function. The behaviour of the B777 FMS is described in the Honeywell manual I mentioned earlier.
@370Location
More progress. I also found the signal from the air pressure wave in the nearest of the two arrays. The other one was turned off at that time. The fit is good with the French satellite data.
The detailed shape of both signals (air and sea) indicate a meteor came in from a slightly northerly direction.
I think there is no need to speculate more about a nuke or a sunken nuclear submarine. It must have been a meteor. The weak signal was most likely unrelated.
@Andrew
Thanks. When did the NG issue and was it a software only update. Also, do you know of any other changes that might be relevant to the MH370 analysis. Getting accurate info is a good part of the problem 🙂
@Mick Gilbert,
Yes, the 00:19 restart of the SDU was the second example of this type of event. Unfortunately, we don’t have any means of independently checking the BTOs for accuracy, so it does not assist in understanding if there is a bit of additional delay.
That 00:19 restart was also different in that the SDU was only without power for about a minute (per ATSB), so the OCXO did not cool down much. That probably reduces (or eliminates) the transient frequency errors affecting the BFOs then. So, maybe the best we can do is ignore that effect in this case and treat the BFOs normally, which results in the high estimated rates of descent (lacking any other mechanism to explain the increasingly negative change in the BFOs).
I have never heard of a different explanation for the SDU restart at 00:17 besides loss and restoration of electrical power. It is possible that an engine restarted briefly but that would be highly unlikely after 00:17:30, so the APU automatic start is the only option that makes sense to me. MEFE must occur within a few minutes of 00:17:30 at the range associated with Arc 7, so I think the depletion of engine fuel is highly likely to be the cause of the loss of electrical power to the SDU.
@David F,
There was an active pilot circa 18:24. Otherwise the Left Main Bus could not have been re-powered then, which re-started the SDU. The pilot could have survived on the flight deck for at least an an hour after the 17:24 Diversion turn-back with the air packs off and the cabin depressurized. If the aircraft crashed south of 34S on Arc 7, I don’t believe it is possible for the pilot to be alive at the end of the flight. There’s simply not enough fuel to achieve that range at the required air speed without the air packs being off, no matter how much of the electrical equipment is de-powered at Diversion.
There were maneuvers after 18:24 requiring pilot actions. The FMT was commanded to change course to reach the SIO, and this does not appear to have occurred at a waypoint on N571. It also seems likely that the air speed was slightly reduced and the altitude was increased, requiring pilot action then.
The 18:40 phone call BFOs are highly informative but difficult to interpret. They don’t match either the N571 course and speed (as do the 18:25 – 18:28 BFOs), nor do they match the final course and speed toward Arc 7. Their stability is greater that what would normally result from a fairly rapid descent at 2,500 fpm. On the other hand, a lower ROC during a climb at about 500 fpm may be sufficiently stable. Possibly the 18:40 BFOs are a result of a new course at about 190 degrees after a turn off N571, plus an ongoing climb to a higher altitude (as from FL350 to FL375, for example) at about 500 fpm.
One turn off N571 to 190 degrees after 18:29, with an ongoing climb at 18:40, matches all the 18:40 – 00:19 satellite data with one exception – the 19:41 BFO. That one is odd, for some reason. I don’t know why it is about 7-8 Hz higher than the prediction using a SIO Route with no maneuvers. In UGIB we hypothesized a small rate of climb then. However, the reason for it and the means to perform it are unclear. Here we run into another commonly-used “hidden assumption” – that there were no maneuvers after 19:41. Even if there were, that doesn’t necessarily require a pilot to be functioning then. Maybe there was a slight change in course after 19:41, at a waypoint being passed between 19:41 and 20:41. At that point in time the “next” waypoint would automatically be selected and determine the destination and the final course. The pilot could simply have entered two (or more) waypoints before the 18:40 phone call.
Longitudes of “Plan B” Route LEPs : 85.6E – 86.6E
@Viking,
The Java anomaly was first presented to the ATSB in 2018 after the OI search was over. I took the findings public in 2019. It has been debated and discussed here in the years since. The discussion resulted in a significant realization that the seismology was not a complex interaction with an underwater implosion in the SOFAR channel. That was predicted by experts at the CTBTO who called on scientists for in depth study. It was instead a far simpler seabed impact that confirmed the initial epicenter calculation.
You are raising old objections based on new false assumptions. We have had several email exchanges in support of your candidate site, where you had ample opportunity to express any doubts about mine. Still, I will take this as an opportunity to explain my findings as needed.
Several contributors here have expressed that the Java site be searched because it is so specific, but usually after higher priority sites. While waiting for that moment, I have continued my acoustic research by acquiring new data sources and testing new approaches. I have sought to avoid cherry picking and confirmation bias by running the same algorithms on other days and against noise. That has caught errors more than once. Calibrating against known events has been key to getting reliable results.
My analysis has weathered high profile but miscalculated reports from Kadri, which have only fueled conspiracy theories and piling on by others who cherry pick from his unintended misinformation.
And now there is a unique opportunity where OI does not want to risk rough seas on an expanding search area without solid evidence, yet their $70M contract to find MH370 is good until at least June.
I have been working on the raw data for over a decade with dozens of algorithms and thousands of plots, which gives far better insight than looking at someone else’s published report and coming to a different conclusion.
If my work has suddenly inspired you to new revelations for your candidate site in the last few days, then more power to you. It appears that a search of your broader area will have to wait, so I would ask that you take the time for in-depth analysis and testing.
A bolide does not make sense for the Java anomaly. A meteorite remnant large enough to register a seabed impact would have been detected on infrasound arrays around the world. I encourage you to obtain the restricted raw infrasound data from the CTBTO. Kadri has access through his university, and you might collaborate or obtain access through yours. In particular, the Cocos Keeling Island array has never been properly examined:
https://370location.org/2018/08/cocos-island-infrasound-may-be-key-to-locating-mh370/
The entrained water is not about its weight dragging the fuselage down. Water is obviously neutrally buoyant in water. It is the enveloped mass of the water inside the fuselage that adds to the energy impact at the seabed. Navy reports on the physics of sinking ships are available. The fuselage is clearly not high density as you seem to require for a high energy impact.
To reach the Java site the plane was likely piloted, and the pilot had hours to contemplate how to handle entering the water if he was unable to land on a runway. Any ditching attempt was clearly unsuccessful, but the pilot may have put the plane in a steep stall for minimum entry speed. Impacting along the crest of a wave belly first may have split the fuselage at the wing root. I recall seeing an airliner crash vid where the plane went off the end of the runway, then cracked in half as the front stretched past an embankment. It is not far fetched to allow that a forward section of the fuselage sank slowly until completely submerged.
I have already explained why quakes are relatively weak on hydrophones. Hand waving about needing 4-5 more orders of magnitude is incorrect. Air cannons on multiple survey ships are clearly detectable. They put out about less than 200KW-seconds of energy to produce around 235 dB sound events at the surface. That’s equivalent to
the strongest of lightning strikes. Detection of the survey ships comes and goes as they venture into shallow water or otherwise favorable terrain to be reflected into the SOFAR channel. Any survey ships over deep water are undetectable via the SOFAR channel. Only ships propellers near the array (or terrain) are heard. Here’s a link to the report where a small lithium battery pack explosion was clearly heard 8,000 km away:
https://www.researchgate.net/publication/228404702_The_long-range_detection_of_an_accidental_underwater_explosion
I don’t doubt that you have a plot from a towed survey array that you cannot share. If you think you are seeing a weak event 270 km from the Java site, then consider that the azimuth on HA08 and the towed array plot that I have point to a reflection-conversion point on the Java coast, typically at 200m depth. The timing of the signal also matches the added travel time to and from the coast to the detector.
If you were to obtain permission to share it privately or put me in contact with the source, I might be able to dig deeper.
Please be aware that the weaker event was detected using 40 stacked seismometers to subtract local noise, but was not detected deep in the survey noise with only three distant hydrophones. Again, the weak event was found based on seismic timing to the same epicenter as the 00:15:18Z Java event, so not unrelated.
@370Location
I keep getting the impression that you think I am trying to cast doubt on your strong event. You have the timing and position of that one 100% correct. We disagree on its absolute strength, but do agree that it was the strongest signal in the Indian Ocean for at least 24 hours.
As an old spectroscopist (master project on molecular spectroscopy) we always had a saying that everybody could agree on line positions, but intensity was much more difficult to agree on, so it is not surprising to me that this gives trouble.
On the other hand, the French satellite pictures tell a lot about what happened. Practically nothing is in the visible picture. The (far) IR signal presents a nice ring as expected for a pressure wave (local compression heats up the gas, so it emits IR light). The WaterVapor (WV) signal is entirely different. The area passed by the pressure wave gets extremely foggy due to condensation. The WV picture shows exactly this effect. The fog is mostly under the normal cloud cover (also expected). Finally, the old type of IR detectors on board METEOSAT7 were quite slow (as I told you). Therefore the IR picture was effectively taken a bit later than the other two. Therefore the radius of the fog area is smaller (roughly half) of the radius of the ring. Everything is consistent with a meteor impact. In principle it could also have been a hypersonic missile, but as far as I know only China had such missiles in 2014, and only a few prototypes. I cannot imagine they would have fired such a prototype missile across Indonesia. Only North Korea does such stupid things (and typically over Japan).
You say that you used the combined signal from 40 seismometers to extract the weak signal from the background noise. If there were reflections involved (as you say) that becomes a delicate task with many free parameters. Therefore, I trust the recordings from the two arrays much more. Each of them measures the heading, signal shape and time to good precision, and the two results agree nicely with each other. I know the distance (and direction), so those are not free parameters. It also makes me suspicious that you initially got the time wrong. That is what could happen if the reflections were interpreted incorrectly.
The two signals are unlikely to be reflections, since there is no primary signal, and the signal looks clean (totally symmetric spot in contrast to the seabed impact which is somewhat asymmetric). I guess it came in at an angle around 75 degrees (instead of 90 degrees as I initially guessed).
Concerning the atmospheric pressure wave there is a similar signal with the same moderate asymmetry at the expected time. Sadly, detector array 2 was turned off before the signal arrived, so I only have it from one array. It is much weaker than the water-guided signal, but still visible.
I need another 10-15 minutes of signal-recording from CI to find the same signal there. Given the high intensity and arrival from a northeasterly direction (with low noise) it should be easy to see. I hope you will give it a try.
@Viking,
I don’t know what you’re seeing in images from a French satellite. I don’t recall you ever sharing them. Images in your report can’t be objectively examined for subtle features because you drew lines on them. If you are seeing a pressure wave from space like the Tonga explosion, it would definitely have registered on infrasound arrays. Nothing met the threshold of automated detection for MH370.
I’ve tried to clarify the difference between the hydroacoustics and the seismics. Seismometers detected the seabed impact directly. The CTBTO hydrophones are located at SOFAR depth, and are sensitive to even weak distant signals traveling horizontally in the SOFAR channel. For that to happen with the Java anomaly, the seabed noise needed to travel upslope to the the continental shelf (likely as Scholte waves at the fluid/seabed boundary) then by T-wave conversion back downslope into the SOFAR channel. I simplified that by calling it a reflection off the coast.
The weaker signal was only detected seismically. No hydrophones and no reflections.
The various seismometer recordings, including two on Christmas Island, are publicly available from the ds.iris.edu service and others. The easiest way to access them is with a python package called obspy, and possibly obspyDMT for gathering and correcting the data.
@Dr B.
Thanks. In the generally accepted scenario it was unlikely that anyone other than the pilot was active at 18:25 but supposing the SDU had been very cold then as you raised, whoever it was, the same ought to have applied to them. Still I suppose they could have wrapped themselves in blankets; oxygen mask notwithstanding.
Understandably, you have assumed now and also in the UGIB LEP that a pilot’s priority must have been range, though all the same I think still there was a fair chance it would be concealment. A few miles extra range would hardly be a good trade for the uncertainty of leaving the aircraft to descend where that and its impact might be observed and heard; and not necessarily lead to flotsam minimisation, thence increasing the prospect of that being spotted by fishing boats, satellites or washing ashore.
To me an active pilot would be able to monitor the leg south and control the descent yet without reducing range if he closed the outflow valves – not just restore them to auto, close them fully. That would lead to not only reducing their drag to normal but to less than that, nominally extending fuel beyond your estimate.
As to whether from time to time he might select a pack on briefly for warmth and to re-pressurise so that he could discard his oxygen mask, that would depend again on how concerned he was to maximise range. Maybe blankets and the mask would continue to do.
Thus even though not needed for navigation, a pilot electing to continue to the end need not shorten your calculated range by so doing.
As to the destination of around 39S, 86E, his steep descent would be 15-20 mins before sunrise (at sea level), according to a web calculator I have found. A night descent, or at least impact, is what I speculated recently to be his aim and even though in the event he may overshot into dim light, for my part that tends to add to the general case you put.
My earlier postulation was that he would retain a fuel reserve in case he found a need to abort a descent and try further on, including some for an APU start for descent control: he being unaware of tank residual fuel being accessible at MEFE in level flight. That need for a reserve could be met if, having shut the engines (or remaining engine) down before MEFE, he had glided, only selecting APU start when about to dive, for dive control. That would be after he had ensured, from a reduced altitude, that there were no lights below.
After starting, the APU would have energised the SDU a minute later, after dive commencement, though unbeknownst to him.
The above is of course just another hypothesis but I hope you and others find it of interest.
As to whether the glide would be long enough to complicate final-transmission SDU cooling implications (your second para to Mick Gilbert), it might, though it would have needed to be at high speed to realise the distance needed for compatibility with final BTO’s and of course commenced after the 6th arc transmission.
Being about to enter my 88th year, being slow, well out of touch and mistake prone, I acknowledge my use-by date is well past, so will contribute only very sparingly henceforth.
And today, you never know, maybe even with no editorials.
The skills, devotion, generosity and persistence of those contributing to this international blog and Victor’s balanced leadership of it are admirable and remarkable.
@David F: Thank you for the kind words and I am very impressed with your continued contributions at a time in life when many are doing a lot less.
@Sid Bennett
Honeywell’s NG FMS entered service during development of the B747-8, which was certified in August 2011. The system was approved as a retrofit for the B747-400 shortly afterwards and was later certified for Embraer’s range of E-jet aircraft and several types of business jets made by manufacturers such as Gulfstream and Dassault Falcon.
The NG FMS required hardware changes, including new FMC units with more memory and faster processors. It was not just a software update. The system features a completely redesigned modular architecture that allows easier upgrades and better adaptability to different aircraft types. I don’t have a full list of its features, but enhancements include support for RNP 0.1 approaches, FANS compliance, WAAS-LPV approach capability, 4D trajectory management for improved ATC flow control and reduced separation, a larger navigational database, enhanced performance predictions and the planned step climb/descent function.
I was a little loose with terminology when I described the B777 FMC as a ‘legacy’ system. It is actually quite advanced in the sense that it is integrated with a range of other functions into a system known as AIMS (Airplane Information Management System). Other AIMS functions include thrust management, flight data acquisition, data communication management, central maintenance computing, airplane condition monitoring and primary displays. Some of the capabilities found in the NG FMS were incorporated into the B777 FMC function over the years, but the planned step climb/descent function was not included. The best description of the B777 FMC capabilities that pertain to MH370 may be found in the Honeywell B777 FMC manual and the B777 FCOM. There is no mystery to it!
@370Location
Your description as reflections off the coast did indeed confuse me, so that is one disagreement less.
Now you have admitted that it takes at least two mode-conversions to reach Australia. That is another one disagreement less.
I am willing to send you a copy of the relevant IR picture plus the corresponding reference picture. Without the reference picture it is extremely difficult to analyze the picture. If you go to 325% enlargement it is easy to see the circular pressure-wave imprint near the bottom of the screen (without moving the picture on a standard labtop screen).
Please send me an email to confirm your interest, since there has been some confusion about addresses in the past.
@DrB
In the same spirit as you have shared your thoughts on possible alternatives, I have commented on them where I can and have suggested some of the possibilities for further investigation. We have explored a number of blind alleys over the years. But the plane is real and we need to find it.
[Comments are in breckets]
1. After 12 years, we have nothing “positive” to show except physical debris discovered in far-off places. The prediction of crash latitude based partly on the finding locations and especially on the arrival times is uncertain. In almost all cases the arrival date is unknown. Only the discovery date is known for some debris, and even that is unknown for many debris. A major contributor to the crash latitude uncertainty, even with a good drift model, is the assumed or estimated windage and leeway of the debris. Drift model analyses are highly useful in confirming the general crash location in the SIO, but they may not be highly accurate in locating the crash location. We have no prior example of this being done accurately for aircraft debris over long distances and long drift times. So, at best we can say the reverse drift predictions are unproven in this situation. We don’t yet know if they are accurate. Time will tell.[ Agreed. It should have been weighted less than the DTSG-defined search area due to the uncertainty in the model. I don’t know if anyone has made a formal combination of the two probability dietributions.]
2. On the other side of that coin, on the “negative side”, no debris was found in Western Australia. Potentially, that could help eliminate sections of Arc 7. However, the large uncertainty in predicting the probability that passers-by will beach on that coastline, be found, and be reported makes it impossible to be certain about whether a given section of crash latitude on Arc 7 would or would not result in debris reports from WA. Again, using the lack of reported debris in WA is an untested and unproven method of predicting crash location. [OK. But it is a secondary issue.]
3. Also on the “negative” side, the debris field has not been found by the seabed searches. I believe all satellite data and debris indicate a high-speed crash near Arc 7 with no piloted glide after fuel exhaustion. So, considering the zone within about +/- 25 NM from Arc 7, latitudes from circa 39S to circa 33S have been thoroughly searched without success. By thoroughly, I mean that 95-98% of the area was declared not to contain a debris field. [Negative side for what? I disagree with the hypothesis that the plane crashed within +/-25 nm of the 7th arc. If we are not looking in the alternative area based on the glide hypothesis, we will never find the wreckage. Still working my way in detail through the scenario.]
4. The negative seabed searches are highly informative. They have told us, with a very high probability (> 95%), the crash did not occur near Arc 7 within that 33-39S latitude range. That means that your prediction of 37S and the UGIB prediction of 34.2S are both almost certainly wrong. [Do not agree. Current search doctrine depends strongly on the Boeing simulation whose initial conditions are poorly described in public documents and on the validity of the 2nd BFO value after second engine flame out. 4000-14000 fpm in 8 seconds.]
5. That leaves us with two choices: (a) the aircraft crashed farther than 25 NM from the arc in a place that has not yet been searched, or (b) the aircraft crashed near the arc but outside the 33-39S searched latitude range. Believing that the crash is much more likely to have been near the arc, I choose the second option as being more likely. In my opinion the debris field is near the arc but outside 33-39S latitude. [(a) Yes to that. (b) Even the DSTG analysis which is statistical in nature favors the current focus on the searched latitude range along the 7th arc. What would the negative search results suggest in a Baysian analysis where there were two different possible scenarios after the 2nd engine flameout?]
6. The area near 39-40S has been proposed many times, including by the original Independent Group in 2014, by me in September 2014, and recently by Paul Smithson. However, that crash latitude fell into disfavor over the years for two reasons : (a) it could not be seamlessly connected to the derived FMT Route, and (b) there was not enough fuel on board to reach it. The second objection was actually the basis by which ATSB abandoned their initial search zones S1 and S2 early on and moved up the arc to S3. However, that decision was based on the ATSB’s assumption that the aircraft had not slowed after 18:22. From 17:24 to 18:22 the air speed derived from radar data was 500 KTAS, which is M.85 at FL350. There definitely was insufficient duel to fly at M0.85 until 00:17:30. However, there was enough fuel to fly at M0.85 from 17:24 to 18:24, and then to reduce the speed to LRC. [(a) Recently I showed that the post-FMT route can be firmly connected to a waypoint route as far back as Penang. (b) This is marginal and not in my mind determinative. If the pilot had tried to select FLs during the period between the top of climb and the FMT, could any fuel have been saved? The final FL would be whatever would give the most economy. ]
7. For the past five months I have been re-examining these objections to a circa 39-40S impact, and especially their underlying assumptions, to see if and how they might be overcome. [Keep going.]
8. First on my list was to build a better fuel model which would allow exploring a range of air speed control options. That task is now complete. I was able to build a generalized fuel model which is simple in conceptual terms and highly accurate. I am now able to predict the fuel flow for any combination of 9M-MRO weight, flight level, SAT, and Mach, within about 0.5% (compared to Boeing flow tables for LRC, M0.84, and Holding). This new fuel model has three areas of improvement. First, it incorporates lift and drag coefficients for this aircraft type at a given Mach to predict the thrust needed to equal drag. Second, it accounts for the substantial non-thrust-producing flow. Third, it accounts for the effect of air temperature and Mach on the flows for the thrust and for the auxiliary loads and engine losses. I now am able to predict the flow for speed modes such as MRC, ECON with Cost Index = 52 (the MH370 flight plan), and M0.85. [This would be a useful improvement of our understanding of the various weighting factors.]
9. As I showed in Case 9 in UGIB (2020), the known MEFE time of 00:17:30 can be closely matched if the post-18:22 route is flown at LRC with no major descent, but with some fuel flow reductions. The fuel savings required to match MEFE are two : (a) the air packs are turned off at Diversion and never turned back on and (b) there is a small reduction due to electrical equipment being turned off at Diversion. My new fuel model confirms this result. It also quantifies the electrical power reductions needed to be roughly 50 kW per IDG (out of a typical assumed load of about 70 kW).
10. Thus, it is possible for 9M-MRO to have flown at LRC until 00:17:30, but not with a functioning pilot at the end of the flight. That would have been prevented by cabin depressurization for seven hours.[ I am now assuming no pilot input from 18:40 and perhaps slightly earlier.]
11. It is also possible for 9M-MRO to have flown until 00:17:30 with a functioning pilot aboard the whole time, but not at a speed comparable to LRC. That lower speed necessarily leads to far-NE crash locations, but none of these match the satellite data without multiple maneuvers after 19:41. This complication makes it less likely, in my opinion, and also impossible to predict with sufficient accuracy to enable an affordable seabed search. [A functioning pilot requires too many assumptions to simulate. The alternative restricts the options and makes it tractable to think about.]
12. That leaves the SW end of Arc 7 circa 40S as a potential contender, if the previous objections can be overcome. We know the aircraft could fly at LRC until 00:17:30 (if unpiloted at the end). So, what does it take to reach 40S at LRC? The answer to that question is a very early turn southward from the vicinity of MEKAR, with no lateral offset maneuver. Could that have happened? [I think the evidence favors a FMT at IGOGU at 18:40. A 15 nm offset maneuver was a standard procedure for a plane in that area of the world that has lost communications. Later it was changed to 5 nm. I used 10nm recently for convenience as the 15 sec resolution of Barry’s spread sheet program is not granular enough to choose.]
13. What drove the insertion of a right lateral offset maneuver circa 18:25 (as shown in UGIB)? The answer to that question is the apparent discrepancy between the 18:22:12 military radar position report and the 18:25 – 18:28 BTOs. For those two data sets to be consistent with no maneuver in between, one of two things must be true: (a) the radar points are too far west by at least 15 NM or have time stamps which are more than 2 minutes too early, or (b) the BTOs then are 100 microseconds too high. Neither of these seemed likely, so in UGIB we inserted a maneuver between 18:22 and 18:25 that would resolve the discrepancy. That was a 15 NM right lateral offset, the purpose of which is unclear.[ I would have to go back and re-do that specific portion of the analysis with better granularity. ]
14. I have recently re-analyzed the BFO data at 18:25 – 18:28. Now I have an improved time alignment which allows compensation of the thermally-driven frequency transient error for all the 18:25 – 18:28 BFOs. This indicates all BFOs then are consistent with a course close to that of N571 with no ongoing maneuver. This new finding further constrains the lateral offset maneuver to be completed prior to the first BFO at 18:25:27. This maneuver takes 3 minutes to complete, and for it to resolve the discrepancy between the radar points and the BTOs means that it must have started after 18:22:12 and completed before 18:25:27. In other words, this 3-minute-long maneuver has to fit in a 3.25 minute wide window. This seems to me to be a highly unlikely event, especially when one understands that the power restoration occurred in the middle of this time window circa 18:24:00. [See my answer to 13. This time period saw the restoration of power to the SDU and meant that a person was active in the cockpit. The offset maneuver is a simple manual keypad entry. Could be done and executed in seconds.]
15. Now we have three seemingly unlikely events to explain the radar/BTO discrepancy : (a) the radar data are very wrong (by 15 NM or 2 minutes), (b) the BTOs are slightly wrong (by 100 microseconds), or (c) a lateral offset maneuver surprisingly appeared in a very tight time window. DSTG opted for (a) and ignored the military radar data after 18:02. UGIB opted by (c), the offset maneuver. What about (b)? Could it be correct? [This is beyond my purported expertise 😊. But I have used(c) as it is a reasonable match to the BFO data.]
16. The implication of choosing (a) or (c) is that the starting point for the SIO Route at the FMT is moved by the equivalent of 15-30 NM compared to the best alternative to reach 40S. That is enough to make a difference. [The FMT cannot be later than 18:40. IGOGU is the waypoint turn. (Alternatively there could be a DIRTO ISBIX entered somewhat earlier, but our solution is very sensitive to such changes. I would prefer to go back to some more detailed analysis of the last radar contact through 18:30 period.)]
17. I will now select the third choice – that, somehow, the BTOs at 18:25 – 18:28 are too high by 100 microseconds. I do that because the other two choices led to predicted crash locations which we now know are very probably wrong. Option (b) is now, therefore, the only viable option to reconsider. [Not with a glide.]
18. How could the BTOs be 100 microseconds too high, for several minutes? I don’t know. What I do know is that this SDU start-up was unique in two ways. It was done in flight. It was done when the SDU was very cold. There are no 9M-MRO data to assess what the SDU does in either of these conditions. None. So, whenever anybody says they expect the SDU in this case to behave the same as a normal start-up on the ground at normal room temperatures, they are making an untested and certainly unproven assumption. That assumption could be wrong. We know the SDU was cold because the warm-up took longer than usual. It took so long, in fact, that the Built-In Self Test was completed at least 40 seconds before the OCXO reached operating temperature. In every other start-up, the first transmission occurred when the the BIST was completed, the OCXO having already reached the operating temperature. The longer oven warm-up for the 18:24 start-up is a direct indication that the oven was colder than normal when power was restored. This is consistent with the air packs being turned off and the outflow valves being opened at Diversion an hour earlier. The air in the cabin would have dropped in pressure and cooled down considerably, so the SDU would have been getting colder for the hour before 18:24. [One of the differences in a restoration of power in flight is that the SBU is computing the Doppler correction in real time, while on the ground the value is pretty static.]
19. So, I start my new route assessment by ignoring the 18:25 – 18:28 BTOs, since the other two options have led to search areas in which the debris field was not found. Thus, there is no need for a lateral offset circa 18:22. [For much of the time I have been doing simulations, I ignored the period prior to the FMT. The latest simulation gives best results as a smooth waypoint turn at IGOGU from a track offset from N571. I am a bit burnt out with the spreadsheet, so I won’t go back to it until I finish the glide scenario.]
20. Then I look at what happened at MEKAR. Was there a turn toward NILAM? That is also an assumption that could be wrong. In fact, the radar data appear to indicate no turn at MEKAR, but rather a continued course of 286 degrees, unchanged from the bearing from VAMPI to MEKAR. How could that happen? Well, it’s easy to do if you just put VAMPI and MEKAR into the waypoint table, rather than all of N571. Then when you get to MEKAR, a route discontinuity occurs and the lateral navigation thereafter holds the same magnetic heading. Basically, the aircraft flies right through MEKAR with no turn before or after. [But there is a pilot. He (for sure HE) was necessary for the restoration of SDU power and is unlikely to have permitted a route discontinuity at that time.]
21. Now, you only need one turn circa 18:31:45 to 188.9 degrees initial bearing with no maneuvers thereafter. This ends up near 39S on Arc 7. One way of accomplishing this is to insert a DIRECT TO waypoint of 45S85E. [Yes, just like I added an arbitrary waypoint after ISBIX. The key is whether the kinemetric model yields a good result. If you can give me a lat/long, time (which I can vary) of the DIRTO and Mach, I could easily try it.]
22. Such a route has very high BTO and BFO probabilities, both peaking near 38.7S, and there is sufficient fuel (within a fraction of one percent of average predicted flow).[see 21]
23. This route PDF is quite narrow (only about 1.5 degrees in crash latitude) because (a) it requires the route be seamlessly connected to the 18:02 location with no delays and (b) it allows only one turn after VAMPI. [Same is true of a 18:40 FMT.]
24. There may be a sliver of this route PDF which lies just beyond the SW edge of ATSB’s searched area. Could ATSB have just missed the debris field circa 39+S, by only a few miles? [Too early to speculate.]
@David F,
I don’t believe that achieving maximum range was a priority for the pilot. That was certainly not the initial priority at Diversion, because the speed was increased from the Flight Plan to M0.85 (with about 500 kts of ground speed based on the radar tracks). Doing that would minimize the chance of being followed and intercepted, but at the cost of burning more fuel and reducing range.
Although turning the air packs off at Diversion increases range, I think the reason for doing that was to incapacitate the other people on board. We also know that the speed was reduced below M0.85 later in the flight because that’s the only way MEFE can occur as late at 00:17:30. At M0.85 for the remainder of the flight, MEFE would occur at about 00:01 with the air packs remaining off and at 23:56 with the air packs turned back on circa 18:24.
It is possible that at Diversion circa 17:24 the air speed control remained in ECON mode but the Cost Index was increased from 52 to about 250. That is simple to execute, and the air speed would go to M0.85 at that time, with the Mach automatically decreasing to about M0.82 as the weight was reduced by fuel consumption. So, it’s possible that the pilot only changed the Cost Index once, circa 17:24. Doing that would give him the desired speed boost, while sacrificing endurance and range.
After the other people were incapacitated, the pilot could have closed the outflow valves to save a tiny amount of fuel and perhaps allow the cabin to become slightly warmer. However, it doesn’t seem possible that he repressurized the cabin, based on when fuel exhaustion occurred. On one point we disagree You seem to think a pilot might “occasionally” repressurize and survive until MEFE. I don’t think that is possible, because of the extreme physical effort required to breathe in a low-pressure environment, even with a pressure-demand oxygen mask. It’s just not possible for any pilot to do that and survive for seven hours. And why would he? That makes no sense, either. For all SIO Routes flown at LRC or higher until MEFE, there’s not enough fuel to sustain cabin pressure. For all crash locations south of 32S it is highly likely the pilot was deceased at MEFE and there was no piloted glide.
My guess is that the final speed, flight level, and waypoints were entered into the FMC before about 18:35. No pilot actions were needed after that was done, except removing his oxygen mask.
@All,
I need to add a fourth item to my list of explanations for the possible discrepancy between the 18:22:12 military radar location and the 18:25 – 18:28 BTOs:
(a) the radar data are very wrong (by 15 NM or 2 minutes),
(b) the BTOs are slightly wrong (by 100 microseconds),
(c) a lateral offset maneuver surprisingly appeared in a very tight time window, or
(d) the BFOs are very wrong (by 60 Hz).
For Item (d), the 18:25 – 18:28 BFOs are strongly affected (by up to 130 Hz!) by the OCXO thermal transient during oven warm-up after a SDU power outage and restoration.
A sharp left turn between 18:22 and 18:25 can produce a match to the BTOs then, but the BFOs, even after corrections are applied, are now off by about 60 Hz. I don’t know any way for the BFOs to be shifted by 60 Hz, especially so at 18:28, after the thermal transient effect had died out.
Despite a great deal of effort by me and my colleagues, we have been unable to come up with any sound technical explanation of a 100 microsecond shift in the SDU processing delay or a (temporary) 60 Hz shift in the BTO Bias. Therefore, at present I must reject (b) and (d) above.
Regarding (a), there is one way to perform an independent check on the accuracy of the military radar data. There was another flight (UAE343/EK343) which was also flying NW on N571 about four minutes behind 9M-MRO. If we had the data from that same military primary radar at Butterworth for UAE343 in the vicinity of MEKAR circa 18:26 we could compare those locations with the ACARS and ADS-B data for that flight. Surely if the military radar locations of that airliner matched its independently known location, we would know the MH370 radar points were similarly accurate in the same location four minutes earlier. I have requested these data from ATSB. We’ll see if they can be provided.
So far, we can find no direct evidence that the military radar data are deficient in any way. Every test we have done has confirmed its accuracy in time and space.
Thus, if (a) is not the answer, then that only leaves (c), the lateral offset maneuver. Such maneuvers are uncommon but are required in certain non-standard situations. They are simple to execute. At first glance it is surprising that such a maneuver might occur in such a brief (3-minute) time window. However, it turns out that they may not be randomly timed at all. One action is required to execute the lateral offset maneuver. A second action is needed to restore IDG power to the Left main AC Bus, which then re-starts the SDU. Both actions would have occurred soon after the aircraft passed MEKAR (and exceeded the maximum range of the Butterworth radar), and both actions could easily have been done within a few seconds. The pilot would not have known that, even though the In-Flight Entertainment (IFE) System was turned off, the SDU would be re-powered and would automatically transmit handshake signals to log onto the satellite network some three minutes later. The fact that the lengths of the offset maneuver and the SDU start-up time are both about three minutes is coincidental, but the actions to begin each were logically correlated and may have been done close together in time.
If we now accept the lateral offset maneuver, Item (c), as being the most likely of the four options, portions of Arc 7 become more difficult to reach from this more Northerly location within the available time and fuel. This is especially constraining for Arc 7 near 40S. I have recently demonstrated the feasibility of reaching 40S on time and within the available fuel if a southward turn off N571 occurred very early – near MEKAR – instead of a right lateral offset maneuver. However, one cannot do that and match both the BTOs and BFOs then. So far, no one has demonstrated the feasibility of reaching 40S after the offset maneuver with full consistency with BTOs, BFOs, and fuel flows. There is enough fuel at LRC, but the distance is too great for that air speed. At M0.85 there is enough speed but not enough fuel. Maybe there is an intermediate ECON speed setting which is adequate in both speed and fuel. We shall see.
@Sid Bennett,
I appreciate your commenting on my 24 points.
You said: “I disagree with the hypothesis that the plane crashed within +/-25 nm of the 7th arc.”
To prove that the crash could have been more than 25 NM from Arc 7, please demonstrate it is possible for a pilot to have survived to reach Arc 7 and then perform a manually piloted glide. To do that, you have to show a route for which there is sufficient fuel for the air packs to be on for all but an hour or so after Diversion. Alternatively, please provide an example of a large modern airliner staying aloft for an extended unpiloted glide after fuel exhaustion in secondary mode. Lacking one of those, the crash was not far from the Arc.
You asked: “What would the negative search results suggest in a Bayesian analysis where there were two different possible scenarios after the 2nd engine flameout?”
If you believed a glide was possible, then you would keep the far-from-Arc unsearched areas as part of the remaining search area with some assigned probability.
The DSTG simulations did not consider this glide/no-glide choice. Nor did they properly consider the fuel required for different routes. It’s wholly inadequate, in my opinion, to simply say “there was enough fuel to reach this end point”. There not only has to be enough fuel to reach that point along the simulated path, but there can’t be too much. If the route does not have MEFE at 00:17:30 it is wrong. Many routes they “accepted” could have been rejected because the fuel was too little or too much.
@DrB
“Regarding (a), there is one way to perform an independent check on the accuracy of the military radar data. There was another flight (UAE343/EK343) which was also flying NW on N571 about four minutes behind 9M-MRO. If we had the data from that same military primary radar at Butterworth for UAE343 in the vicinity of MEKAR circa 18:26 we could compare those locations with the ACARS and ADS-B data for that flight. Surely if the military radar locations of that airliner matched its independently known location, we would know the MH370 radar points were similarly accurate in the same location four minutes earlier. I have requested these data from ATSB. We’ll see if they can be provided”
Regarding your paragraph, perhaps the ATSB was not privy to any Malaysian radar data at all from 18:03 onwards. Perhaps EK343’s predicted path was worked out after 18:07 the same way I did, a straight forward, distance over time calculation.
https://mh370.radiantphysics.com/2017/10/12/simulator-data-from-computer-of-mh370-captain-part-1/#comment-8310
@DrB
Arrival time at Vampi for EK343 on the date in question is predicted as approx 02:16:51 if that is EK343 continued along N571 at the same speed from its last known location at 02:07.
From Vampi, the predicted time it would take is approx 5 minutes 21 seconds for EK343 to be at its possible location (6 23 29.26N 96 52 6.31E) along the N571 at 02:2212.
The estimated approximate distance between the predicted EK343 and the predicted the MH370 locations at 02:22:12 is 60.18KM.
That means it would take EK343 at it’s still estimated speed along the N571 approximately 3 minutes 53 seconds to be at MH370’s last known predicted radar position.
https://mh370.radiantphysics.com/2017/11/25/possible-mh370-debris-seen-in-aerial-search-in-spring-2014/#comment-10295
@Joseph Coleman: Back in November 2024, I used the UAE/EK343 radar data from the Butterworth and Langkawi radar sites and compared it to the Malaysian military data for MH370. My estimate was a spacing of 34 NM at 18:22:12.
I created this graphic:
https://www.dropbox.com/scl/fi/r5gstrzhhdnfqsoqovsm9/2024-11-26-Radar-Data-for-MH370-and-EK343.png?rlkey=uiquwit0zhfq409vyagnbgc50&dl=0
I explained the MH370 v. EK343 comparison as follows:
https://mh370.radiantphysics.com/2024/03/05/ocean-infinity-proposes-new-search-for-mh370/#comment-37412
For those interested in the flight path of EK343 relative to MH370, this graphic may be of interest, where the path of EK343 was calculated from the SSR data sets from the Butterworth and Langkawi radar sites, which were obtained through a back channel.
At 18:22:12, which is the last radar point for MH370 in the Lido Hotel slide, MH370 is 10 NM past MEKAR on N571, and EK343 is around 24 NM before MEKAR also on N571, for a distance between targets of around 34 NM.
Some observations:
1. EK343 was tracking on airway R467 between the VKL VOR and GUNIP, and then joined airway N571 towards VAMPI and MEKAR.
2. EK343 was out of range of the Langkawi radar site just after passing MEKAR.
3. Malaysia’s military radar should have had no problems distinguishing between the MH370 and EK343 targets due to the relatively large spacing, which was around 34 NM at 18:22:12z.
4. The targets from the Lido Hotel slide are close but not exactly positioned in the figure due to difficulties in accurately extracting the coordinates from the original photograph.
5. As we’ve known for years, the data from the Lido Hotel slide is noisy and sparse, which may explain why Malaysia has been reluctant to share it.
@Joseph Coleman,
ATSB is asking DSTG and Malaysia AAIB for Butterworth primary radar data of EK343. ATSB does not have it now, and DSTG probably does not. We’ll see if Malaysia AAIB is willing to provide it. We may also get the MH370 PSR data from 16:42 to 17:29.
@Victor
As I can see from your graphic for the EK343 track, Did Langkawi SSR start from 01:51 or then about? And Butterworth SSR ended just before Vampi. So after Vampi in particular is it only Langkawi SSR for a possible comparison between the EK343 path and the said MH370 Military Radar track.
@DrB
We can only Hope.
@Victor
Looking at your graphic too, did Butterworth SSR for EK343 end at approx 0 2 1 5 LOCAL?
@Sid
I question if the Right offset is exactly correct, I think it could have been a Left jog over to B466 to ANOKO after MEKAR and before NILAM. There was air traffic coming down N571 so it is possible pilot wanted to get off of N571. I do feel the pilot may have been saving fuel, adding 15-nm Right/north adds total 30-nm to the path to SIO, seemingly that could be a mistake forced by unexpected air traffic, but I can “live with it” if I have to accept it. That is to say, the 180s path with “loiter” still works but I have to give up on flying as far around Indo FIR if I have to waste 30-nm closing off a offset to the north.
@Dr B.
I agree that rather than range being the aim, it and fuel consumption were consequences.
Comments:
• Depressurising by using the outflow valves only would result in fuel consumption above that available, while selecting the packs off instead would not.
• Complementing that by opening the outflow valves beyond their auto setting would shorten depressurisation time. However since the cabin crew would have access to oxygen cylinders a speedier depressurisation might not make much difference to the time to total hypoxia, so this step well could be left aside.
• In the later long leg south, what I described was his use of bleed air to top up the cabin pressure, outflow valves closed. Such would be needed only to overcome leakage, so sporadic rather than the continuous you supposed. Rather than that you raised the alternative of a pilot choosing instead to endure continuing cabin depressurisation, since maximum range was not. About that I agree with your, “..why would he?”
Likewise though, closing the outflow valves and topping up from time to time, as above, rather than just selecting packs, again begs the same question, range not being his aim.
• As you say, fuel constraint does lead to the conclusion that either there was no pilot, though while much less likely, the aircraft might have consumed less fuel elsewhere than your calculations would indicate.
• On that, while you have high confidence in the accuracy of your fuel flow calculations there could be an accumulation of minor variables adding to a fuel usage probability better expressed as a probability than a categorical figure. I have listed some of those and add that the assumed atmospheric conditions, including temperatures and wind speed and direction, might be probability peaks themselves rather than spot on.
So far as I know they are estimates.
Re: Depressurization Strategy
Keep in mind MH370 had an unusual high bleed air at take-off with one circ fan off (per ACARS), which we have discussed here some years ago without resolution as far as rationale. I presume that could have role in faster depressurization by holding outflow valves somewhat more open that usual. I presume the initial logic would be open outflow valves fully (they are slow acting, so couple minutes to open) and cut off bleed air.
@TBill
I am just going where the data seems to take us. As for motivation…..
@TBill,
I raised this with @Andrew who, as I remember, said that most likely that off- selection was to help with windscreen demisting while taxying then overlooked at take off, and probably it would have been spotted and corrected later in the climb.
However while off, pressurisation air flow would have been higher as you say and with that, fuel consumption.
@David
@TBill
In humid conditions, the recirculation of moist air through the air conditioning ducts causes mist to form around the outlets in the cabin. That effect can be reduced by turning off the recirculation fans. There is an EICAS memo message to remind pilots that one or more fan switches are off, so it may well have been a deliberate act if a switch was left off for take-off.
@370Location
I checked a few more signal details from the two array detectors:
1) The angular dispersion (as observed in the nearest array) of your strong signal fits nicely with expectations from the mode-conversion and the shape of the slope of the seabed.
2) The time dispersion of your strong signal fits somewhat less perfect, but I suspect it is slightly longer because part of the signal is going through the above jump layer instead of the SOFAR channel part of the way.
3) The angular difference between the middle of the dispersion curve for the strong signal and the (non-dispersive) weak signal is roughly 13 degrees for the nearest array. The most distant array value is roughly 1 degree smaller. Using these values together with the absolute arrival times, I still find a distance between the weak and the strong event near 270 km.
4) There is a small signal corresponding to direct propagation of the sea surface impact signal (causing the strong event at the seabed roughly 17 minutes later) through the SOFAR channel at an angle corresponding to the lower end of the angular dispersion curve for the strong signal arriving at 1:13 at the nearest array. These values are in good agreement with the ring observed at 1:00 using an IR picture from METEOSAT7.
5) There is an extremely small signal in the nearest array corresponding to direct arrival of the atmospheric pressure wave at the expected time. However, this signal is too small to use as a proof.
It would be highly interesting to look for signals at Christmas Island for signals from the events. The only one predicted within the signal segment you sent me should arrive as a weak seabed R-wave just before 00:30. There are 3 very weak signals between 00:29 and 00:30, but they are too weak to analyze in detail.
Most other signals are expected to arrive after the end of the signal window I have, typically soon after 1:15. There is some risk of signal overlap.
@Viking,
You are drawing conclusions from hydroacoustic plot images that you won’t share, using vague terms that are counter to the raw data I’ve analyzed. You conflate three different types of dispersion: frequency, angular, and time. You are referring to a weak signal I detected only on seismometers at the same epicenter as the Java anomaly, yet you claim it’s 270 km distant on secret hydrophones. I’ve tried to infer what you’re claiming, but I cannot decipher your point #4 related to seeing a ring on a METEOSAT7 image that I previously shared. Even with advanced contrast enhancement, I’ve never seen the contrails or ring you claim are visible. I no longer know what you’re looking for on XMI and XMIS seismometers, but have described how you can plot the data yourself.
If you can share the data, plots, and French satellite images that you’re drawing conclusions from, maybe it will start to make sense. At this point, it has become a word salad of claimed evidence and proof.
@DavidF
Re: MH370 Bleed Air/Fuel Consumption
There is a point/counter-point the higher MH370 bleed air, on the one hand reduces thrust, on the other hand, reduces power consumption with the circ fan off. My pay grade is too low to net that out.
@370Location
Let me try to help you with the different items:
1) Frequency dispersion is effectively absent for the fundamental mode in the SOFAR channel. For the rest of the channels there is frequency dispersion. This dispersion is all of the type waveguide dispersion, since material frequency dispersion is very small for the wavelengths involved. Waveguide dispersion is related to the effective filling of the waveguide. For instance the longer the wavelength is compared to the waveguide core diameter (or effective core diameter in case of propagation above the jump layer) the more it samples the material outside the core. A small wavelength compared to the diameter is effectively only inside the core material. The amount of the wave outside the core also depends on the refractive index step between the core and the cladding (or in other words on the speed difference between the two materials). For normal guiding the speed is smallest in the core and faster outside. The larger the speed difference is, the more the propagation is confined to the waveguide core.
2) Angular dispersion is predominantly due to mode conversion happening at different places. Other types (e.g. due to thermal differences, which can lead to things like self-focusing (or defocusing) if the wave-propagation heats the material) are also known, but they are too small to matter here.
3) For your strong signal, mode-conversion between seabed R-waves and SOFAR waves predominantly happens during reflection from the northern seabed slopes of the Sunda strait. Since this slope is located along a curved line (locally inward curved) it is possible to get reflections from different places reaching an array detector located near Australia. Phase matching conditions further complicate the local amplitude of mode-conversion, so some places reflect exclusively into the SOFAR channel while other places also reflect into the above-jump layer or into higher order SOFAR modes.
4) Time-dispersion is generated in the same process, since parts of the signal reflected in different places along the steep side of the Sunda trench travel different lengths (and in different modes) to reach the same end goal at different times. I measured these position differences for a couple of contributions, and came up with angular patterns very close to those observed by the array detectors. The angular dispersion fits nicely, proving validity of the approach, while the time dispersion only fits if I assume some of the places also reflect part of the intensity into the above-jump layer.
5) The local inward curving of the steep part affectively makes it work as a focusing lens, so the array detectors receive more intensity than naively expected. That is not the case in other detector positions and/or for different origins such as that for your weak signal.
6) A seabed event is also able to excite small amounts of SOFAR waves directly without a sloping undersea landscape. Typically it is only seen for very strong signals, of if the seabed has suitable surface ripples effectively working as a so-called long period grating (providing phase-matching). Such a grating makes mode conversion in the forward (outward) direction opposite to the slope-induced conversion. The effectivity is frequency-dependent, so the long-period-grating also works as a filter exclusively passing certain frequencies but not other frequencies. This could be the reason for several different peaks from one event.
In addition to dispersion effects, you claim that you shared a METEOSAT7 picture. I am not sure which one you are talking about. I downloaded all the pictures from a meteorological home-page long time ago. Most of them are now deleted at that place.
I am willing to share relevant French METEOSAT7 pictures. The most important IR picture with the pressure-wave ring also contains several contrails from other flights partly in/out of Jakarta. In addition, the Mushroom cloud I assign to be generated by a combination of atmospheric instability and the crash of MH370 is still visible at 1:00 in the same IR picture. If you look very carefully, it is also possible to see remains of 4-5 older Mushroom clouds. These were triggered during a short time-interval 1-2 hours before the crash of MH370 by a well-known re-trigger mechanism appearing when the morning sun starts hitting the top of old thunder-clouds. That mechanism came to an end roughly 1 hour before MH370 crashed nearby. The crash triggered one single Mushroom cloud at a meteorologically unexpected time, but fitting nicely with the crash-time. The trigger mechanism was the large splash of hot seawater up into the lower part of the cloud.
I use signals from the southern seismic detector at Christmas Island that you previously sent me, The signals are only visible in the red curve (at the bottom of the stacked curves).
@TBill. Use of engine bleed flow is very inefficient in power use but continues owing to the complexity, cost and weight of alternatives, I gather.
Scaling from Andrade and Tennings’, ‘Design of the Boeing 777 Electrical System, Fig. 4, the recirculation fans, with ‘Lav/Galleys’, consume 2.1 kW on the ground. Assuming that the recircs take the bulk of that and have the same loading in the air, around 0.4 kW for each?
Either way, very small when compared with the fuel consumption rise due to the added 67 lbs/min bleed air needed to replace each. That will total about 0.6%, at 0.3% each, say 188lb/min, in the climb (see below). At an engine thermal efficiency of 30%, that would amount to about 300 kW of power as a comparison, though of course less in the cruise.
The outflow is at a constant volume rate as the aircraft ascends, at least once cabin pressurisation starts, mass flow rate decreasing with the decreasing density of cabin air as cabin altitude rises.
@Andrew. Revisiting the anomalous 435 lb/min ACARS second, and last, Engine Health Report total pack bleed rate recorded in the climb, the applicable Aircraft Maintenance Manual’s Schedule 1, minimum, indicates that should be 307 lb/min. Indeed as you know, that is consistent with an earlier MH371 flight where flow was 311 lb/min at about 2000 ft lower.
In MH370 a fan switch (two fans) off would increase the 307lb/min by 67 each, to 441 lb/min, pretty close to the 435 above; though still the cause could be something else entirely.
Continuation of the consequential increase in fuel consumption is made the less likely observing as you have that there would have been that EICAS warning (and there might well have been a warnings too had, instead, outflow valves failed/jammed/been selected to manual). Also though, continuation would rule out reaching the UGIB LEP, noting its consistency otherwise with other indicators and earlier surface and bottom searches and imagery.
However there was some circularity in the mutual support entailed: why the increased fuel consumption wouldn’t extend into the cruise – because it couldn’t have. With the aircraft wreckage not yet having been located near that UGIB LEP to date, it might be worth bearing in mind that the continuation of that excess fuel consumption just might have continued, the warnings being overlooked in distracting circumstances.
Incidentally I notice from that climb HER the fuel flow difference between engines, i.e. 15,583 lb/hr left, 15,786 right: 1.3%.
Attributable just to right engine relative inefficiency at that power level do you think? Seems a lot of added fuel cost and yet be seen as economic. Just engine overhaul costs?
@TBill. Erratum:
‘2.1’ kW in line 4 should 21 kW, ‘0.4’ kW in line 5 should read 4 kW, though the that the outcome is unaffected.
@David F
RE: “Attributable just to right engine relative inefficiency at that power level do you think?”
Probably. The right EGT is also higher, so the engine is obviously working harder to produce the required thrust.
The extra fuel burn is insignificant in the short term compared to the cost of an engine overhaul, which runs into millions of dollars. Barring any other issues, I think the usual practice is to keep monitoring the engine parameters and pull the engine off the wing once the EGT margin reduces to the limit specified by the manufacturer.
@370Location
The imperfect fit of the time dispersion triggered further investigations from my side. Initially I thought it was due to part of the signal reflected into the above jump layer. However, that would yield too large time dispersion to fit the data. After writing my last contribution I investigated the possibility that the conversion went into a cladding mode of the SOFAR channel instead. That would fit better to the observed data, but it has the disadvantage that most intensity would likely get stripped off when passing the opposite side of the Sunda trench.
Therefore, I looked again at the signal received by the nearest array, and now in maximum resolution. That showed a signal composed of two different contributions, roughly 1.5 minutes apart and with somewhat different angular dispersion. The last signal had almost no time dispersion. If this separation was due to cladding mode excitation the angular dispersions would have been more similar, and the time dispersions would have been practically identical.
The details instead pointed to two different contributions coming from the same place, but around 1.5 minutes apart.
After some speculation, the only way I could come up with such a scenario would be if there were two separate impacts. One impact event where low- and high-density parts separated upon impact with the sea could not fit. The low-density elements would fall much more slowly to the seabed than 1.5 minutes difference indicated.
I then went back to look at the satellite pictures again and realized that I had overlooked two very important details. First of all, there were two separate rings formed roughly 30 minutes apart, but close to each other in position. The first one was photographed at 00:30. It was small and irregular corresponding to a couple of smaller impacts very close to your proposed time of 00:27. That would agree with expectations for a small, low-density meteorite. Then a new ring formed shortly before 1:00. That one was more regular, considerably stronger, and elongated in east-west direction. When I looked previously, it looked almost circular because of left-over from the previous ring. This second impact fits roughly in time and angle with my previous claims. The impact time must have been around 00:58.
I had also overlooked that both impacts happened roughly midway between my suggested impact position (270 km east of yours) and your position (same as the seabed impact).
Now things started looking consistent with a more complicated, but also more likely scenario. A meteor composed of two parts impacted on earth at a much lower angle than I had initially expected – comparable to the one in Chelabinsk. The meteorite was torn apart by the gravitational field of Earth, whereby the low-density part hit the atmosphere before the high density-part (most likely iron, – far too strong to be pulled apart). I had overestimated the released energy considerably because of the focusing effect in the Sunda trench and the double-signal in the array. Most likely the released energy was around 1000 times smaller than the one in Chelabinsk, making it a very common (daily) event, but still liberating a total energy more than 100 times higher than a plane crash.
The two impacts took place near each other and at similar angle but roughly 30 minutes apart. Due to the high horizontal velocity (from east to west), the pieces impacted on the seabed quite far from their surface impact. Since the low-density part came first but sank slowly to the seabed they happened to impact roughly simultaneously with the seabed with the lighter one 1.5 minutes after the heavy one. The heavy impact liberated an energy roughly 100 times expectations from a plane crash, while the light impact liberated something like 4 times expectations from a plane crash.
Now you probably think this is an unrealistic and unnecessarily long explanation, but I think I can prove it. If you look at the red signal from Christmas Island, it has several relatively small peaks. Most of these come from my position (due to the crash of Mh370?), but one of the strongest at 00:35:30 does not fit at all into that picture. It consists of a slightly stronger peak followed by a couple of small ones (smaller fragments?). The time fits with SOFAR channel signals coming from an impact at the center for the first irregular circle observed by METEOSAT 7 near your proposed time (00:27).
I expect a much stronger signal to appear from the impact of the iron part of the meteorite on the surface of the sea. It should arrive at the southern detector at CI around 1:06, plus-minus 1.5 minutes. Sadly, your time series only goes to around 00:49.
Most importantly, this scenario explains all signals (array, seismic, satellite) I have access to, and also why you think it was MH370 crashing at your position. The energy released on the surface at 00:27 is consistent with a plane crash. It happens within the expected time-interval, but a bit far from the 7th arc. However, the seabed impact of both parts happened at the 7th arc, but roughly 50 minutes too late, and with 100 times too high energy release. Fortunately, the array detectors had sufficient resolution to resolve the details and solve the mystery together with the French satellite pictures. The seismic detector at CI provides confirmation of the model, particularly if we also find a signal around 1:06.
@Viking – Regarding “If you look very carefully, it is also possible to see remains of 4-5 older Mushroom clouds. These were triggered during a short time-interval 1-2 hours before the crash of MH370 by a well-known re-trigger mechanism appearing when the morning sun starts hitting the top of old thunder-clouds. That mechanism came to an end roughly 1 hour before MH370 crashed nearby. The crash triggered one single Mushroom cloud at a meteorologically unexpected time, but fitting nicely with the crash-time. The trigger mechanism was the large splash of hot seawater up into the lower part of the cloud” –
From review online, a mushroom cloud seems to require intense heat and energy. To quote AI “Mushroom clouds form immediately following a large explosion, such as a nuclear detonation, volcanic eruption, or massive chemical blast, when a buoyant, intense fireball of hot gas rises and cools. The rising hot air draws up dust and debris, creating a stem, while the cooling vapor at the top spreads into a cap.”
I am not clear how you say there were remnants of 4 to 5 mushroom clouds before MH370. Was there such huge thermal impact or other factors before the crash?
@Viking,
The point of Victor posting this page about the Java candidate is that it’s a unique opportunity for searching a specific viable site before the OI contract runs out.
I’ve looked back at our exchanges, and see that you’re rehashing old claims and adding new ones, refuting any analysis that doesn’t support your candidate site.
Last year, you seriously suggested that a megaton meteorite actually hit flight MH370, which is beyond comprehension.
Bolides leave a trail of dust particles and ionization in the stratosphere, not a ring or mushroom cloud. You only saw a second ring after perceiving a second event from a secret acoustic plot. Even a huge remnant meteorite would be less than 1/300th the mass of plane debris hitting the seabed, even less than one engine.
You apparently have spectrograms and azimuthal intensity over time plots for two hydroacoustic arrays. If so, you should be able to triangulate.
You may have experience with fiber optics, but the analogy falls apart when dealing with the SOFAR channel. There is no cladding layer in the ocean. You claim that there is no frequency dispersion in SOFAR propagation, but common ice events and survey airshots are well dispersed. Quakes clearly show temporal dispersion due to arrivals from a wide area around the epicenter, but not angular dispersion. Signals reflected off of broad terrain are unique in showing angular dispersion, but also show temporal dispersion from multiple echos. A reflection from concave terrain can even focus the acoustic energy in amplitude and time.
Your reference to a weaker event is unrelated to mine if it comes from a different direction, though you may be seeing a reflection. From now on, I would prefer to dissociate any of the events I’ve detected from whatever you’re perceiving.
I now see that when you refer to French satellite images, you’re talking about METEOSAT7. I posted a series of images for you in 2021 (with a corrected source link). The first link also has the seismic plot you are currently requesting.
Attempting to refute your various claims seems pointless, as you just keep repeating them and piling on more.
I’d prefer to focus on what OI stands to gain by a quick search of the Java site.
@ST
The short answer is that the atmosphere in the area was very unstable for two reasons. It was close to the intertropical convergence zone at that time of the year (where heavy thunder-clouds and tropical hurricanes form/persist), and the seawater was extremely hot while the atmosphere was starting to cool down.
About two weeks later a tropical hurricane peaked in strength between Christmas Island and this area.
@370Location
I agree it would be good if OI could search one of our two sites. However, I am increasingly sure your site only contains fragments from a meteorite. As I said many times, the released energy is too high to be consistent with the crash of MH370.
The nuke idea was mostly a joke from my side – just to illustrate a large energy discrepancy.
I would certainly love to do triangulation, but the focusing effects (and reflections) make it difficult. As an example, the nearest array did not even see the earthquake, because the energy was focused away from it. The other array did see the earthquake, but with somewhat reduced intensity compared to expectations. Unfortunately this second array was shut down a few minutes before the arrival of the mode-converted seabed signal from your event.
In addition, the signals I see from your 0:27 event are indeed likely to be reflections.
In arctic regions the SOFAR channel will have some dispersion, but I explicitly ignored thermal effects with only one example (self-focusing) to avoid making the explanation incredibly long.
The SOFAR channel is likely to have higher order cladding modes for some of the relevant wavelengths. It is not necessary to have an inner cladding to guide such modes. Propagation losses are likely to be significantly higher, particularly near sharp transitions in temperature, salinity or depth.
As far as I know METEOSAT7 was built at a factory near Paris (visited it a few years before they made that satellite), and was a French satellite.
@DrB,
@Victor,
@Paul Smithson,
DrB said in part and in para (6) on March 15, 2026 at 4:57 pm
concerning “The area near 39-40S” and “recently by Paul Smithson” that this region had fallen into disfavour for two reasons the second being “(b) there was not enough fuel on board to reach it.”
The only “way” that this region could have been reached is if the fuel flow rate was significantly reduced compared to a normally functioning passenger carrying flight.
For the purposes of discussion and developing some outer bounds, and after some review, we can apply two initial sets of “limits”. For the first, the approximations, or estimates, presented in the 2020 “UGIB” for air conditioning support non thrust producing fuel flow are most likely an underestimate. For the second, it would seem that Paul’s estimate for the fuel cost of engine bleed for cabin pressurisation of 3% in Paul’s “1pp summary Jan 2026” may be getting close. But the cost of cabin pressurisation also incurs a measurable electrical load for fans (and chillers ?), for example. This may, or may not, be covered by Paul’s allowance for a lower IDG load of 0.8%. I suspect that 0.8% is an underestimate of normal passenger support electrical loading. Consider, for example, passenger support cabin lighting.
So, I started passenger air supply analysis from somewhat “first principles” which has been interrupted. If I can find time and concentration to proceed with this in the future then it may even be of value to know how many passengers the preceding Flight MH371 of 9M-MRO was carrying.
For background I found a document meant for estimation of airport ground support facilities required to support the B777, which helps.
The aim of my attempts was to determine if there was sufficient information available to confirm or deny Paul’s extension of the flight path to his eventual end of flight.
All,
I have continued to refine and exercise my integrated model of GDAS weather, fuel flow, and 4-D aircraft path. From take-off to 18:20 I update all parameters with a time step of one minute. This is sufficient to model the fuel flow accurately during the early portion of Flight MH370. After 18:20 each model time step is five seconds. This is sufficiently precise to accurately model all turns, using the turn rate dependence on speed and bank angle, and on the limits on bank angle and its time derivative. I can follow geodesics or tracks or headings, although I have focused on geodesics between waypoints since that is the normal long-distance navigation method used by airline pilots.
A recent finding is that the observed ground speed (507 kts) required to match the 17:21 – 18:25 radar and satellite data is best matched by M0.86 at FL350. I now have the ability to accurately predict the fuel flow at higher speeds, and there is a significant increase in fuel flow. This increase amounts to a 1.5% net reduction in total flight fuel compared to using LRC (as we did in UGIB) during this post-Diversion leg from 17:24 to circa 18:45.
A second finding is that this high speed indicates a new possibility for matching the 18:40 phone call BFOs. The route leading up to 18:40 starts with a slight right turn circa 18:21 at MEKAR to follow N571 toward NILAM. Very shortly thereafter, a 15 NM right lateral offset is executed off N571. This path is entirely consistent with all 18:25-18:28 satellite data (including all BTOs and all BFOs then) and the military radar track out to 18:22:12. Next, when abeam IGOGU circa 18:38, a left turn is made toward VOBX (the Indian Naval Air Station in Baaz).The phone call occurs during this leg from IGOGU to VOBX. Note that when departing N571, the lateral offset is canceled when abeam IGOGU, and the track proceeds around IGOGU and directly to VOBX. What happened between 18:45 and 19:41 is still a mystery, because we have no observational data during that hour. It is certainly possible that MH370 flew NW at least as far as IGREX and/or VOCX (Car Nicobar AFB) before turning south.
“Plan A” Search Zone
I call the areas thoroughly searched by ATSB and Ocean Infinity the “Plan A” Zone.
The Arc 7 latitude range of the Plan A Zone is 33S to 39.5S.
The seabed in this area has been searched for a debris field with area coverage ranging from 95% to 98% within about 25 NM of Arc 7 with negative results.
It is possible to reach many parts of this area with a route which is consistent with all the radar and satellite data, given sufficient fuel. The fuel is marginally adequate to inadequate, even when significant fuel savings are applied to achieve the known endurance. This includes having the air packs off from circa 17:24 through MEFE at 00:17:30 and allowing 50 kW of electrical loads reductions per IDG. The electrical loads reductions can save 0.4% of fuel, but my new, higher estimated flow for the first hour of Diversion costs an extra estimated 1.5%. Thus, compared to the UGIB(2020) fuel estimates, there is a net requirement for another 1% in fuel for the routes to the Plan A Zone. This makes it less likely that such routes are flyable, and even less likely that they were flown with a functioning pilot at MEFE.
The negative Plan A search results to date imply that the probability that the debris field is near Arc 7 between 33S and 39.5S is only 2-5%. Therefore, from a practical perspective, we must abandon Plan A.
Almost certainly ATSB and UGIB were wrong in their predicted crash locations. In addition, the case for believing no pilot could have survived until the end of the Plan A routes is bolstered by the fact that no known fuel savings measures beyond the ones I have already applied can make up the larger fuel shortfall with the air packs on (to provide the pressurized cabin required for pilot survivability). Therefore, the probability that a piloted glide occurred in the Plan A Zone is extremely low, so it is also quite improbable that the debris field is outside the searched area near Arc 7 in the Plan A Zone.
It is not worthwhile, in my opinion, to continue searching the seabed in the Plan A Zone. The chances of success there are too low to justify the cost.
“Plan B” Search Zone
The question now is if and where to search next.
I have evaluated the area immediately SW of the end of ATSB’s search area at 39S. I call this the “Plan B” Search Zone.
Is it worthwhile to search the Plan B Zone? The short answer is no.
The problem with the Plan B Zone is that it is even farther away than is the Plan A Zone. Thus, it is not possible to fly to the Plan B Zone with the time and fuel available. An additional 2%-5% of fuel is needed beyond the savings measured already assumed, depending on the air speed control method. I have evaluated the fuel needed for speeds of Cost Index (CI) =0 (i.e., MRC), CI=52 (the Flight Plan), CI=150 (LRC), CI=250, M0.84, and M0.85. I did a parametric study of each speed mode, maximizing endurance and maximizing range by adjusting the flight level. There is no speed which simultaneously matches the known endurance to MEFE, matches the range to the Plan B Zone, and matches all the satellite data. The basic problem is that the SIO Route must begin after the Right Lateral Offset maneuver and connect to it with no gap. That’s why I built my path integration model – to assure the connection between the FMT Route and the SIO Route is flyable.
Thus, Plan B is also not worth searching. ATSB did a good job in extending their seabed search area down range just beyond what was flyable. Kudos to them.
“Plan C” Search Zone
So where should we look next? I am a firm believer in the validity of Arc 7, so it must be on or very near Arc 7. If the debris field is not southwest of the searched area then it must be northeast of it.
I have recently begun exploring routes which end on Arc 7 at latitudes north of 33S. This is a work in progress, but I have found a reliable and simple method of excluding most of it from consideration. I call the excluded area the Plan C Zone, and it extends northward of 29S latitude on Arc 7.
Why 29S? The answer lies in the BFOs for Arcs 4, 5, and 6. Simply put, north of 29S you can’t match those BFOs at any speed (even within +/- 7 Hz) on any bearing at any speed. The observed BFOs are higher than what can be achieved in level flight in the Plan C Zone. Of course, you can match any BFO given a large enough rate of climb (ROC) or rate of descent (ROD). In my opinion, it is highly improbable that this occurred at exactly 21:41, 22:14, and 00:11, or that a constant-vertical-speed descent was flown for almost 3 hours. I also note that there are auto-pilot routes (to other parts of the Arc) which match the handshake BTOs and BFOs with a single geodesic with no turns and no altitude change from 19:41 to 00:11. Thus, there is no need to employ artifices to match a single or a couple of satellite data points. All you have to do is look elsewhere than the Plan A, B, and C Zones.
“Plan D” Zone
I call the region between 33S and 29S the “Plan D” Zone. This is simply the gap between the NE end of the searched area at 33S and the SW starting point of the Plan C exclusion zone at 29S.
Is the debris field in Plan D? Maybe. Here is what I think I know:
1. The seabed has not been searched here, either near or far from the Arc.
2. There is a simple route ending there which matches all the satellite data, is flyable, and for which the fuel is adequate for MEFE between Arcs 6 and 7.
3. The BTO and BFO residuals have normalized mean squared values of 0.7 and 0.9 sigmas. The fit is excellent and consistent with the expected parameter noise and the number of degrees of freedom without contributions from route errors.
4. The speed is the Flight Plan setting, with Cost Index = 52, and the Flight Level is 390.
5. The fuel is consistent with the air packs being off from 17:24 to circa 18:42, and with the electrical load reductions occurring at 17:24 until fuel exhaustion.
6. The fuel is adequate for the cabin to have been re-pressurized prior to the pilot becoming incapacitated.
7. However, in my opinion the possibility of an extended glide is not supported by the in-flight high-speed departure of right-wing control surfaces and the absence of the IFE log-on after the Arc 7 handshake. Therefore, I think future searchers should continue to prioritize the area within about 20 NM of Arc 7.
8. A crash in the Plan D Zone would explain why no debris were reported in Western Australia and why the “Roy” debris was found in South Africa at an early date.
9. An argument against Plan D is that the 2014 aerial surface search did not detect low-windage floating crash debris in the Plan D Zone, at an estimated 70-100% probability level.
10. In addition, Broken Ridge crosses Plan D, so some of the seabed will be very difficult to search because of steep terrain.
Overall, I think the Plan D Zone is worthy of more study and scrutiny.
@Dr B. Re Plan D.
You say that the seabed hasn’t been searched. I believe it has.
https://sea-technology.com/latest-search-for-malaysian-airlines-flight-mh370-concluded
The link shows the chart of areas planned for the extended 2018 search.
OI has publicly stated that Area 3 was searched, implying everything south of 28.5S has been covered at that swathe width. I have been unable to identify a definitive statement that Area 4 [to 26.5S] was completed. I note that Captio’s shape file extends all the way to 25S though I’m unsure of the data source.
You say that the BTO and BFO fit is “excellent”. Unless there is a manoeuvre posited between Arc 2 and 6, I don’t see how it can be. Please post BTO and BFO residuals.
@paul smithson
I agree with your points. I also found higher residuals in the relevant area without manoeuver(s).
In addition, some people pointed out (a couple of years ago) that the area agrees poorly with the general beaching pattern (not with the individual events, but with the overall statistical pattern).
Finally, there were flights and ships going through parts the area relatively often (e.g. to/from Perth). Why did nobody see any debris?
@paul smithson, @DrB: During OI’s search in 2018, the seabed that was scanned extended northward along the 7th arc to 24.84S latitude, and inside the arc for a width of 21 NM and outside to 24 NM. The contour of the area searched was provided privately by Ocean Infinity.
I don’t understand why this area would be a priority unless the recommendation is to search further out than 24 NM, and I would question this decision.
@DrB
Re: Fuel Use to 18:25
I assume Jean Luc et al have made some good new analysis ideas of the radar data. I realize unfortunately it is late in the game, our spirit of cooperation among groups is lacking due to conflicts.
CAPTION feel all power was cut, which in addition to Bleed Air off, could be a significant fuel savings. I personally wonder if Rt IDG could have been cut off at IGARI or later to save fuel and possibly to depower DFDR.
These unknown factors could complicate fuel models. I prefer to think in terms of a range of fuel use to 1825, with focus on the “worse case scenario” of what would be the most fuel left over at 1825, which as a placeholder, I think of MRC from IGARI to 1825. I am not sure with the unorthodox operations if it possible to perform better than MRC or worse. I am not opposed to taking debits on fuel use where it is justified, but so much is unknown.
@Victor, thank you for that confirmation. I agree with you.
Dr B has discounted “Plan B” (south of 39.5) on grounds of fuel feasibility, stating: An additional 2%-5% of fuel is needed beyond the savings measured already assumed…Thus, Plan B is also not worth searching.
That conclusion rests on a highly-questionable assumption – that such fuel savings are not achievable. According to my fuel modelling – the detail of which I am happy to share, I reach the conclusion that fuel exhaustion occurs at 00:07 (right) and 00:15 (left) employing the following model parameters:
1.PDA +1.5%, shared L and R as 0.445% and 2.555% respectively [ref Dr B’s engine data referred on the blog]
2.Temperature adjustment +3% per 10 degree TAT above ISA standard
3.Electrical load adjustment, both IDGs -0.8% in view of anticipated electrical load reduction of roughly 2/3 and estimated associated fuel burn.
4.Bleed air adjustment -3.4% both engines, in view of expectation that bleed air was turned off at 17:21 to achieve most rapid depressurisation possible – and never switched on again. This figure is in line with public domain estimates of “typical” specific fuel consumption impact of bleed air for passenger jets. The ballpark figure is driven by max configuration passenger load for the aircraft type. Additionally, the estimate can be triangulated from other data sources, including the amount of electrical power required to substitute bleed air for “more electric aircraft” designs, as well as the % thrust (N) associated with bleed air – which in turn has an approximately linear relationship with fuel burn. ALL of these numbers point to the same estimate range of 3%-5%. A careful reading of Boeing’s response via ATSB to Bobby’s enquiry stops a long way short of a ringing endorsement of UGIB’s 1.6% fuel saving attributable to “packs off” and it would be imprudent to interpret it as such.
5.Residual fuel. Thanks to @Andrew, the aircraft-specific residual fuel quantities are known from the weight & balance documentation are 122kg in feed lines and engines, not counting the 188kg in the manifold, presumed inaccessible with x-feed closed. My model assumes that there is residual fuel of about 60kg per engine beyond “gauge 0”.
6.The 150 second delta predicted vs “expected” final fuel exhaustion 00:17:30 represents 0.6% of total model time (from 17:07) and is well within the error margin of the fuel model ~1%.
My path model assumes:
M0.821 between 17:07 and 17:21 at FL350; then climb to FL360 from turnback onwards with speed:- 17:21 to 17:36 at M0.76, M0.79, M0.84 in steps, to match radar profile; 17:37 – 17:52, M0.87 (MMO); 17:53 until end M0.84.
For non-M0.84 speeds, I used an earlier version of Dr B’s fuel model. This applies to the period 17:07 to 17:52. From 17:53 onwards, I used fuel burn interpolated for aircraft gross weight and altitude from the Boeing M0.84 tables, applying the adjustments described above to the “base model” fuel burn.
@All: In the wake of the failed search in 2018, I endeavored to find an automated route that ended north of the searched area. In particular, I discovered an automated route that included a turn to the east and a deceleration, which is required for more northern routes, as routes following geodesics and near constant speeds will not satisfy the BTO data.
What I found was a route that included a leg from Car Nicobar Airport (ICAO: VOCX) and towards Cocos Island Airport (ICAO: YPCC) with a possible intention to land there by loading the RNAV Z RY 15 approach. The turn to the east occurred as the approach course was intercepted, resulting in an End of Route and continuation of the flight at a constant magnetic heading. The deceleration occurred because there was no descent from the cruise altitude of FL320 and the VNAV speed schedule followed what would be expected as if the plane descended below 10,000 ft (240 KIAS) and at lower speeds on the approach, limited by the Minimum Maneuvering Speed (MMS) of around 210 KIAS.
The details can be found here:
https://mh370.radiantphysics.com/2018/06/08/an-mh370-flight-path-ending-further-north-on-7th-arc/
This path crosses the 7th arc around 22S latitude, which was never searched.
When drift models suggested an impact point closer to 34S, I didn’t pursue this possibility. In light of the recent search results, perhaps the drift model should receive much less weight, and routes like this one should be considered.
@DrB I agree with you about the Plan D zone, and am interested to hear what flight routes and crash sites you identify in your work in progress.
@Victor. The path above involves a BFO error of 6Hz at 19:41. I haven’t calculated the 1840 BFO residual for that specific path, but I’d expect it to be larger still. The BTO fit is satisfactory, but is achieved by means of multiple changes of speed and direction between Arc 2 and 6 – whereas, as we are all aware, the most parsimonious path solution produces superior BTO fit with no manoeuvre.
No one can declare the route hypothesis “impossible” but I would describe it as “contrived”. I certainly wouldn’t rate it as more plausible or promising than 39.6S.
@Paul Smithson: If you look at the parameters for the updated path from July 18, 2018, the BFO error at 19:41 is 4.7 Hz Meanwhile, the RMS error for the BTO data is 25 us.
The path implies there was an ongoing descent at 18:40z and a holding pattern near Car Nicobar (VOCX).
Yes, there are maneuvers, but they would automatically occur if VOCX to YPCC was the leg in the FMS (a leg between two airports) and the RNAV Z 15 approach for YPCC was loaded. In that sense, the path is relatively simple.
Readers here know that I am not a big fan of routes that rely on “hunches”, which this path does, i.e., a planned landing at Cocos Island that went wrong, resulting in a ghost flight. However, if we assume automated flight after 19:41 AND we wish to find possible crash sites north of 25S latitude, this becomes a candidate, as “straight” paths won’t satisfy the BTO data.
We can point out the weaknesses of this path, such as the timing discrepancies for the floating debris discoveries in East Africa, but this path does have sufficient fuel and is in an area that was not searched before. Unfortunately, we don’t have a lot of options left.
At this time, I am not proposing this as the next area to search. Rather, I am presenting it as a candidate that could be weighed against other proposals.
If the goal were to identify candidate path hypotheses north of 25S then I see where you are coming from. But why would you even “go there” in preference to the best-fitting BTO segment of arc (39-40S), that is compatible with a highly plausible, waypoint-compatible path, hasn’t been searched? I get that some folks prefer to ascribe significant evidentiary weight to fuel, drift, BFO. If everything triangulated neatly, that would be fine – but it doesn’t. The BTO data p-max is around 39.3S. The BFO p-max is around 36-37. The drift – who knows – it depends which model parameters you cleave to … so anywhere between about 20S and 40S. The fuel – depends on assumptions regarding plausible fuel savings. So up to roughly 32S without fuel-saving measures, up to about 40S with fuel saving measures that I have set out above. In other words, we have a bunch of somewhat incompatible predictors. If you try to triangulate them, you end up somewhere in the zone that has been searched. If you believe that the BTO evidence is the most solid of all (ascribe greater weight), and that there are clear, empirically justifiable reasons why the drift, fuel and BFO inferences were misleading to a greater or lesser extent, then prioritising the 39.6S solution is a no-brainer.
Meanwhile, I’d argue that the best latitude discriminator of all is the null-result of the search. In my interpretation of the end-of-flight evidence, that rules out every latitude from 25S to 39.5 with 99% confidence.
@Paul Smithson: We don’t agree on your p-value analysis and what the drift, fuel model, and BFO data suggests. I think everybody also understands that you believe your area near 39.6S is the best to the extent that searching there is a “no-brainer”.
The purpose of my comment was to encourage alternative thinking. The fact that you believe we should not be doing this is noted.
I have no trou le at all to “alternative thinking”. What I find difficult to fathom is obdurate refusal to accept persuasive evidence.
@Paul Smithson: We all understand what you believe to be true. There is no need to continuously repeat it.
It is not as if I’m insisting the earth is flat.
@All
PART A: I note with some disquiet, that the “tone” of recent exchanges here is tending towards the potentially unpleasant. We all know that there are clearly fundamental differences of opinion held by many regarding the relative “weight” that should be applied to different data elements, and therefore, the inferences to be drawn from them are different, which ultimately drives people towards different segments of the arc. That is “known”, and I thought accepted. There is no need to get “snarky”.
PART B: The elephant in the room.
OI unexpectedly and abruptly terminated their search for MH370 in the Austral Summer of 2026 “at the halfway point”, without explanation, and when they had just revisited the ATSB sonar targets test site, and there was a good 8 to 12 weeks of “the Austral Summer search season” (the best “weather window”) still available. The then sea state in the MH370 search area was temporary (not unexpected in the overall scheme of things, and it passed, as expected) such that the conditions were again suitable for searching only two weeks later, which would have meant only an additional stay of one week or so in Freemantle (since it would have taken five or six days to transit back to the search area) if it had continued the search for MH370.
But it did not stay in Freemantle for that additional week. Instead, OI totally blindsided everyone when they suddenly and unexpectedly re-deployed Armada 8605 to American Samoa, (and it is still there) on a task that was clearly “on the books” or “on the to do list”, but which was clearly neither time nor weather dependent, since presumably, it had been INITIALLY planned for Armada 8605 to go there and do that AFTER it had finished the 2026 Austral Summer search for MH370.
It therefore appears that going to American Samoa when they did (presumably TWO MONTHS EARLY) was most likely a fortuitous and conveniently plausible excuse “to get out of town pronto” than anything else, which begs the rather obvious question, why ?
The “why” becomes even more important, when one considers the wording of the OI press release that announced the “conclusion” (OI’s word) of the search, with particular reference to their last paragraph. QUOTE: “Although this phase of the search has concluded, our commitment has not. We’re continuing to work with the Malaysian Government in the hope of being able to return when circumstances allow.”
My question is, why has everyone ignored any substantive discussion of the means of, and the probable the implications of, those final three confounding words, “when circumstances allow”.
All,
Here is a link to a one-page explanation of how the Arc 6 BFO constrains the MH370 crash latitude:
https://drive.google.com/file/d/155jQXE-CB0rBDi-w09aNI1grrEvOzE50/view?usp=sharing
This analysis is done assuming level flight. Arc 5 also shows a similar effect. There simply is no course or speed in level flight which can match those Arc 5 and Arc 6 BFOs for a crash latitude North of 29S. I also don’t think that a constant vertical speed climb could have been maintained for a number of hours to create those BFO errors.
Victor’s YPCC Route confirms my analysis and my conclusion that a crash north of 29S is highly improbable. His 00:11 BFO error considerably exceeds the acceptable limit, and therefore his proposed 22S crash latitude is inconsistent with the satellite data.
@Paul Smithson,
Thank you for the link to the OI map. Regarding how far north OI has searched, I have not found any post-search report from OI in which they state exactly how far they went. Their weekly progress report link no longer works. The map at the link you provided looks to me like a search planning map. Maybe they finished all of their planned areas, or maybe that’s an incorrect assumption we are making.
Perhaps Victor has a map of actual searched areas from OI. It would be very useful at this point to have solid information on this, including the lower and upper latitudes.
Regarding the post-Diversion speed, I find that M0.86 at FL350 from 17:24 to at least 18:42 provides a good match to all the radar tracks and the satellite data, including the 18:40 phone call (with no climb or descent). I would suggest you also continue the high-speed until after that first phone call. At FL350 with a weight of 210-220 MT, the ratio of M0.86 flow to M0.84 flow is 1.073. So, the Diversion Route fuel flow jumps up quite a bit for more than hour.
I will reiterate that reaching 40S has two major deficiencies, which are interrelated. First, the distance is too far, starting after the right lateral offset, to get there on time at speeds of LRC or slower. Also, there isn’t enough fuel to fly there even at LRC by about 1.5%. Going faster burns a lot more fuel, so you just can’t get there from the right lateral offset location.
Here are the results of a trade study I did for Plan B. There is a good route fit at S38.7, but it does not extend past ATSB’s search limit of S39.5, so there is no Plan B area left to search.
https://drive.google.com/file/d/1BlDrSNhWgcVKiJal4qQ01eCDZ9i1D_Ap/view?usp=sharing
It’s not enough to run a fuel model with unvalidated fuel savings guesses, while fitting routes unconnected to where the aircraft was circa 18:41 (which is the last reliable estimated position I have). You have to connect the SIO Route to the FMT Route to demonstrate flyability. When you do that, which is what I have been doing since November, you will find that you can’t reach 40S.
Regarding my initial proof-of-concept Plan D Route, it is at FL388 with a Flight Plan speed of CI = 52. It has nine BTORs ranging from -45 to +43 microseconds. It has 12 BFORs ranging from -5 to +3 Hz when I remove a small slope of 1 Hz/hr, and with which I use a sigma of only 2.3 Hz. If you don’t allow for any OCXO drift, the BFORs range from -7 to +2 Hz, and then you need to use a larger sigma circa 3.5 Hz. I have emailed you the Arc positions and residuals. There is a FMT prior to 19:41, and at 19:41 the bearing of the geodesic is 175.69 degrees at 93.545E, 4.787N. Arc 7 is at 31.73S.
@Paul Smithson. Re your Mar 25th 11:10 AM and its “ballpark” 3.4% fuel saving from selecting bleed air off, here is another look at that.
In my Mar 22nd 1:33 AM for @TBill, I mentioned that the AMM indicated that 134 lb/min of bleed air when added to the 307 lb/min scheduled at that climb altitude (some 22K ft) would lift the fuel consumption rate by approximately 0.6%.
The scheduled bleed air flow rate in a cruise at 30K ft would be 280 lb/min vice that 307. Assuming that the same or very similar fuel consumption penalty per lb of that 134 lb/min increment would apply to the 280 lb/min total, selecting bleed air off would result in a 280/134 X 0.6% fuel saving, or 1.25%.
Were the cause of the anomalous climb pack flow flow rate (in fact 435 lb/min) uncorrected, that fuel savings from a bleed air off selection would rise to (280 + (435 – 307))/134 X 0.6%, or 1.83%. While that assumes the 134 lb/min in the climb is independent of altitude that would be the case were a fan switch being off the cause, as has been supposed. That is implicit in the AMM.
However even this difference may not help in your case. The UGIB flow rate for the cruise assumes already that the cause of that anomaly, in the climb, had been corrected already.
Back to my post minimisation.
@DrB:
The OI weekly reports are included in a Dropbox depository created by Rupp/Labrat:
https://www.dropbox.com/sh/0addl5zq91dpfve/AACXYUVvGSvBJPgrgd8OuD9Ba?dl=0
In report number 18 (29 May 2018), they had completed the area up to around 27S and were working on the final part that extended to around 26.5S, but apparently the search was further extended after that.
@Victor Iannello:
“During OI’s search in 2018, the seabed that was scanned extended northward along the 7th arc to 24.84S latitude, and inside the arc for a width of 21 NM and outside to 24 NM. The contour of the area searched was provided privately by Ocean Infinity.”
Is 24.84S the 7th arc spot or the northern end of the box? I’ll just note that if the 7th arc limit was exactly 25.0S, and with 21.0 NM inward, the northern end of the box would be at around 24.83S according to my calculation (and conversely, 24.84 would translate into 25.01).
@All
An academic debate has re-emerged that I am following closely but cannot contribute much. However I would like to state my thoughts…
The areas of the seventh arc between 33-39s that were searched without success fit the relatively simple theory of pilot murder-suicide, where a mentally and/ or emotionally broken man decided to commit a heinous act and ended up in a painless manner as far away as possible in an area that was almost impossible to search. (Honestly, I have always had a problem with the theory of prolonged suicide).
But if we are talking about areas north of 25s, much closer to mainland, civilization, airports, shipping lanes… it is necessary to include MOTIVE, and we have to think about a much more complex scenario that would include, say, a sophisticated hijacking gone wrong… if I am not mistaken, DennisW advocated that thesis at one time, may his soul rest in peace.
ventus45 said: I note with some disquiet, that the “tone” of recent exchanges here is tending towards the potentially unpleasant.
The fundamental problem is nobody has presented a search zone that is without issues that significantly lower the success probability.
@eukaryote: The 24.84S latitude is on the 7th arc.
@Vlad.BG
You bring up an excellent point – the motive.
When an English translation of the new book on the Palme murder appears, you should read it. After reading the whole book, I am increasingly convinced about the similarities in method and motive.
The title of the new book is: Murder Box
@Victor
I am pleased to read about your thinking out of the box. A 22S solution is one of the few remaining possibilities I find difficult to eliminate.
At the moment, I can only think about the non-observation of such an event by the two array detectors as a way to eliminate it. However, I prefer two independent methods, and I cannot come up with a second one.
@Vlad.BG
I would mostly agree with your summary of the 33-39s search. I would say it was pilot deliberate with focus on passive or deceased pilot (straight LNAV/no maneuvers).
It was probably not until 2017-2018 that a consensus emerged that a true B777 “deceased pilot/unintended ghost flight” would in all likelihood be a curved heading with wind impacts, thus an exact straight LNAV great circle or True Track path generally needs to be a “deliberate” ghost flight preset to fly straight. But back in 2015, the first ATSB Fugro/Bayesian search at 38s focus had more hopes it could also be a unintentional ghost flight as an over-flight/discontinuity.
I see it as deliberate, with motive, with some maneuvers, active pilot, not a strictly straight path, probably at least 100-nm off Arc7, some fuel probably remains at Arc7 (because if MH370 was headed to 30s area there is more fuel due to shape of the Arcs).
@TBill. Your portayal of DSTG flight control mode is incorrect. They did not specify which mode ended at the 38S mode. However, path model analysis is unambiguous that it must be TT or LNAV to achieve low BTO residuals, the latter producing superior fit.
@Vlad.BG: Thank you for your comments.
I think what you observed is accurate. Essentially, impact sites further south can be reached with straight paths and lend themselves to mathematical analyses without regard to motive or intentions. Impact sites north of what was already searched are curved and require an associated scenario to explain (or guess) the maneuvers. That’s the challenge and that’s why localizing a search area to the north is so difficult. I don’t see a way around this.
@Victor
I qualitatively agree with your categorization in two types of solutions. However, it does not mean that northern solutions must be smoothly curved. In some cases it is possible to find solutions with 2-3 manoeuvers and no gradual curvature or gradual height change(s).
If everybody is going to agree on only one solution we must initially find an independent method to distinguish between northern and southern solutions. I have a novel proposal for such a method.
I think the method may actually be able to distinguish between 3 different categories of solutions:
(I) Endings south of 23S
(II) Endings between 23S and 11S
(III) Endings north of 11S
I think we should let science decide which of the 3 categories is the correct one. That would be the best way to move on.
All,
Here is a link to a short paper on the FMT Route, including a new explanation for the 18:40 phone call BFOs.
https://drive.google.com/file/d/1jaT6vKBS7l6MgSG9pA3TqPRswoQALhWz/view?usp=sharing
@Viking said: However, it does not mean that northern solutions must be smoothly curved.
If you look at the path I described, there is a distinct turn to the east when the plane intercepts the final approach course to runway 15.
All,
Here is a link to a short paper on the MH370 bearing at 19:41 on Arc 2:
https://drive.google.com/file/d/1iw1o7J_Gi0RV9AAZLKbDDgIEP-rflExf/view?usp=sharing
Is anyone privy to his flight paths programmed into his home flight simulator that he deleted? He planned something out.
If not, just have to keep pushing.
@Julie:
https://mh370.radiantphysics.com/2017/10/12/simulator-data-from-computer-of-mh370-captain-part-1/
I’m not into the number crunching but was a project mgr on half billion $ transit projects. My main question…
Assuming pilot murder/suicide, why would a person that was obviously cunning, deceitful, and highly calculating fly a straight line path after turning south? A man that successfully hide his true intentions, made his plane go dark, straddled two countries at low altitude, and had to know a huge search would ensue. After all that deception, why would he just fly straight to the end? Doesn’t line up with his immediate prior behavior. My personal opinion is he was performing evasive maneuvers to the end.
@Viking,
You said: “I think the method may actually be able to distinguish between 3 different categories of solutions:
(I) Endings south of 23S
(II) Endings between 23S and 11S
(III) Endings north of 11S
I think we should let science decide which of the 3 categories is the correct one. That would be the best way to move on.”
Science just eliminated your second two categories. See my previous post on March 25 :
https://mh370.radiantphysics.com/2026/03/10/a-candidate-site-for-mh370-that-can-be-searched-quickly-and-soon/#comment-40396
I have demonstrated that crash latitudes north of 29S cannot possibly match the 00:11 BFO in level flight or in a descent. Only an ongoing climb at 00:11 (and at 22:41, as well) can do that, and this seems to me to be quite improbable.
@DrB said: I have demonstrated that crash latitudes north of 29S cannot possibly match the 00:11 BFO in level flight or in a descent.
However, the previous search results rule out impacts close to the arc to a northern latitude of 24.84S latitude. So either the plane was in a controlled glide that extended past what was searched or it searched and missed.
All,
Here is an updated version of what I think I know about the crash location:
In the Plan A Zone (33S to 39.5S):
• The fuel is marginally adequate (at a 1.5% shortfall compared to my nominal fuel flow model), and only then if the air packs were never turned back on after 17:24.
• For Plan A Routes, it is highly unlikely a pilot was alive at MEFE, and so a piloted glide far from the Arc is also highly unlikely.
• The negative search near Arc 7 (-25/+32 NM) eliminates Plan A with a very high probability (both near and far from Arc 7).
• There is no justification for continuing to search in Plan A (as OI apparently concluded).
In the Plan B Zone (South of 39.5S) :
• It is extremely unlikely that the fuel is adequate (at a 2% to 5% shortfall) to reach beyond the 39.5S limit of the ATSB search.
• No plausible route south of 39.5S has been identified which is consistent with all the satellite and radar data.
• There is no justification for searching in Plan B.
In the Plan C Zone (North of 29S) :
• The fuel is more than adequate to reach Plan C and to maintain cabin pressurization.
• However, no route has been identified which is consistent with all the satellite and radar data.
• A simple analysis of the Arc 6 BFO demonstrates that no crash north of 29S can be consistent with the satellite data.
• OI has searched part of Plan C (-21/+24 NM from 31.5S up to 24.8S) with no detection.
• There is no justification to search further in Plan C.
In the Plan D Zone (29S to 33S) :
• The fuel is adequate to reach Plan D and possibly to maintain cabin pressurization after Diversion.
• A piloted glide cannot be ruled out because the fuel may be adequate for cabin re-pressurization.
• So far, only one potential route has been demonstrated to be feasible.
• More work needs to be done for Plan D crash latitudes, including investigating a post-19:41 lateral maneuver, such as a late turn to a Destination waypoint or a Route Discontinuity shift to Magnetic Heading navigation.
• OI has searched -21/+24 NM from 31.5S up to 24.8S with no detection, covering all of Plan D near Arc 7.
• There is a slim possibility that an unpiloted glide might have impacted just outside the OI search width.
• There is a possibility, at an unknown probability, that a pilot survived until MEFE and performed a dive/glide farther from the Arc.
• At present, Plan D appears to be the best choice should additional searches be conducted, but more work needs to be done to firmly establish feasible route(s).
• It is possible that only small portions of Plan D have feasible routes, and this result would at least allow prioritization of subzones within Plan D.
@DrB, @VictorI
As you know I revised my recent paper to accept the fact that at a route discontinuity would result in exiting LNAV, and proposed a distant waypoint to replicate a great circle path.
An alternative method of maintaining continuous LNAV navigation to the 7th arc, without specifying a distant southern waypoint, would be to:
Insert four additional pilot-defined waypoints after ISBIX of the form ISBIX/186/700, ISBIX/186/1400, ISBIX/186/2100, and ISBIX/186/2800, entered on the LEGS page with the heading reference switch in the TRUE position. These four entries, each referencing ISBIX as the origin fix with cumulative distances of 700 nm to 2,800 nm, are consistent with the place/bearing/distance entry format documented in the Honeywell B777 FMS Pilot’s Guide (D201201000013, Rev 1, July 2014, Table 10-1). The resulting concatenated path approximates a rhumb line at 186°T rather than the great circle to 43S88E, crossing the 7th arc approximately 9.3 nm east of the great circle route.
This would achieve the result of making the path insensitive to crosswinds. This could be done at any time before encountering ISBIX, perhaps in the time between VAMPI and IGOGU.
Also, I hope to have a draft of my Glide hypothesis in a few days.
@DrB, @VictorI
As you know I revised my recent paper to accept the fact that at a route discontinuity would result in exiting LNAV, and proposed a distant waypoint to replicate a great circle path.
An alternative method of maintaining continuous LNAV navigation to the 7th arc, without specifying a distant southern waypoint, would be to:
Insert four additional pilot-defined waypoints after ISBIX of the form ISBIX/186/700, ISBIX/186/1400, ISBIX/186/2100, and ISBIX/186/2800, entered on the LEGS page with the heading reference switch in the TRUE position. These four entries, each referencing ISBIX as the origin fix with cumulative distances of 700 nm to 2,800 nm, are consistent with the place/bearing/distance entry format documented in the Honeywell B777 FMS Pilot’s Guide (D201201000013, Rev 1, July 2014, Table 10-1). The resulting concatenated path approximates a rhumb line flown in 700 nm great circle increments at 186°T rather than the great circle to 43S88E, crossing the 7th arc approximately 9.3 nm east of the great circle route.
This would achieve the result of making the path insensitive to crosswinds. This could be done at any time before encountering ISBIX, perhaps in the time between VAMPI and IGOGU.
Also, I hope to have a draft of my Glide hypothesis in a few days.
Sorry. Slight amendment to the second pos. Please read it. Thanks.
@Sid Bennett: Two points:
1. On some variants of the B777 FMS, the distance is limited to 511 NM. I don’t know if this applies to 9M-MRO.
2. Are you sure this creates a path along a rhumb line? It may be that each leg is flown as a geodesic (approximate great circle) with an initial bearing equal to the course that is input, and the path along the geodesic is flown until the distance is reached, changing the track as it progresses along the leg to remain on the geodesic. Perhaps there is not much difference over legs of 700 NM at a course of 186 deg.
@VictorI
I don’t have a FMS simulator, but ….
If you concatenate a series of 186/700 waypoints, the path between waypoints is a geodesic. The initial azimuth is reset at each waypoint. In the case of an almost meridional path, the difference bwtween the great circle and this approximation is small. Of course, there is no reason that the initial azimuth could not be changed by waypoints, but this seemed a simple way to do it.
If the maximum distance were to be 511nm per leg, another leg would be needed. An experienced pilot would know his plane.
@DrB: You wrote “A recent finding is that the observed ground speed (507 kts) required to match the 17:21 – 18:25 radar and satellite data is best matched by M0.86 at FL350. I now have the ability to accurately predict the fuel flow at higher speeds, and there is a significant increase in fuel flow. This increase amounts to a 1.5% net reduction in total flight fuel compared to using LRC (as we did in UGIB) during this post-Diversion leg from 17:24 to circa 18:45.”
Can you please elaborate how the match is better compared to the previously assumed LRC (at high FLs if I remember well)?
@DrB:
You variously write:
“A simple analysis of the Arc 6 BFO demonstrates that no crash north of 29S can be consistent with the satellite data.”
“There is no justification to search further in Plan C.” (North of 29S)
“I have demonstrated that crash latitudes north of 29S cannot possibly match the 00:11 BFO in level flight or in a descent.”
You have in the past said that the Java site is “impossible” due to mismatch with your past fusion of BFO and drift.
Looking over your recent 1 page Arc 6 BFO exclusion report, your assumed parameters retain a clear bias away from latitudes north of 29S. You assume level flight at 480 knots (though noting that the BFO is not sensitive to speed above 400 kt). High speed flight at Mach 0.8 passes the arcs far too early at low latitudes. I have provided two flyable low and slow sample paths to the Java anomaly endpoint, with a 6th arc speed below 325 kt.
You restricted the track headings to 140-200 degrees, roughly SE to SSW. A plane arriving the Java site on the 7th Arc at 8.36S would have flown over Java past Jakarta on those tracks. My 6th arc track between waypoints from Christmas Island is on a track of 84 degrees. My direct path at even slower speed is 47 degrees.
Part of the reason for my picking the waypoint path was that the time between 6th and 7th Arcs was shorter than expected for a direct path going diagonally between the arcs. The plane could instead have climbed after passing the Christmas Island airport
then later picked up speed between the 6th and 7th Arcs.
Your chart shows BFO errors around -35 Hz for the Java latitude. My sample path was based on BTO Arc timing only, but later checking of the BFO showed a worst case of all the arcs was at Arc 6 with +19.2 Hz. I used the spreadsheet recommended by the IG at the time to check the BFO. That’s quite a discrepancy.
We have already spent years searching for MH370 based on the possibly false assumption that the plane made no more turns after the FMT. Early papers from INMARSAT (like Appendix B of the final report) showed plausible BTO/BFO fit paths that included a path near the Java anomaly, but BFO was considered unreliable. Further BFO analysis focusing on the areas searched may have refined it, but possibly still within a narrow context. I think it will be a mistake to rule out as “impossible” any flight path north of 29S based on BFO with narrow assumptions.
Perhaps it is time to review other new evidence showing the flaperon barnacles began growth after beaching at Reunion. That changes old assumptions used in the drift modeling.
We really don’t know what the attitude of the plane was during most of the flight. If it was being piloted, we should not assume it was flying level at high altitude at Arc 6.
@Niels,
You said: “Can you please elaborate how the match is better compared to the previously assumed LRC (at high FLs if I remember well)?”
The average ground speed using the radar track from 18:00:25 to 18:13:12 is 507 ± 5 kts.
The average ground speed using the radar track from 18:13:12 to 18:22:12 is 508 ± 5 kts.
In UGIB we used LRC at FL385 with an estimated 10 kts of tailwind. The intent was to use the best-matching speed for which we knew the fuel flow.
The tailwind varies with altitude. My current path model, which uses GDAs winds and temperatures at the location and interpolated with altitude and time, shows a good match to 507 kts of ground speed with M0.86 at FL350. It shows LRC at FL385 to be near 500 kts ground speed.
The 18:40 BFOs are much better matched using M0.86. Using LRC, you are clearly short of the turn needed around IGOGU.
@370Location,
You said: “I think it will be a mistake to rule out as “impossible” any flight path north of 29S based on BFO with narrow assumptions.”
The only assumption I make about the BFO data is that it is valid. With that assumption, all Arc 6 locations north of 29S are incompatible with the BFO then.
No one has yet demonstrated that ANY BFO is not well behaved and predictable, if one knows the flight parameters at that time. Even the brief warm-up transients are quite repeatable.
Claiming that your proposed location is feasible is only possible by ignoring the BFO data. You are the one making an unwarranted assumption with no basis in fact. We have compared BFOs during many flights, and those data are found to be reliable in every case. There is no reason for you to assume otherwise.
@370Location,
I ran BFO predictions at Arc 6 at 8S with speeds as low as 250 kts and at bearings as low as 47 degrees. The BFO errors always exceed the maximum allowable error. So, my statement is true for your routes. They are incompatible with the BFO data.
@DrB:
My sample route is just one of many possible paths to the Java endpoint. I only intended to show that an easily flyable path matched BTO so used a constant low altitude. CAPTIO showed an optimized BFO path up to Cocos Keeling island airport, and I have no doubt that the remaining portion could also be optimized for BFO.
Here is the V17.0 spreadsheet from Nov 2019 that I last used for BFO computations along my path. It appears that I had put in a track of 84 at 343 kt for the 6th Arc with an experimental 1,075 fpm climb for a null BFO residual:
https://docs.google.com/spreadsheets/d/1BymuFkMxVkfOawd_ssWYphtJ2ruf7YJe/
Am I using it incorrectly? If there is a better spreadsheet for BFO calculations that can be shared, I’d very much like to plug some numbers into it.
@DrB
Thank you. Did you systematically search a certain range of FLs and thrust settings for the interval mentioned to best match with the radar track?
Could you perhaps share the new 18:30 UTC position and total aircraft weight? I would like to see how it would impact my recent calculations.
For the moment I stay focused on the “plan A” zone. I’m not sure how literal we can take the coverage percentages, especially for the low probability detection areas. Just try to imagine a towed sonar scanning under quite shallow angles over often sloped seabed. My pragmatic view is that if something is hard to find, you might need to look more than once in the same place, especially if you have a better search method (read much more advanced technology) at hand.
@370Location
You might find this interesting.
@ventus45
Great article – particularly the section with the red ink. Do you have an email address of the author?
I did not know about that observation. It basically proves that either my position or that investigated by Ed is correct. Everything else is very unlikely as origin for the debris.
@DrB
Victor and Ed have already described 3 different possibilities that you may not be right about your conclusion concerning the BFO value at 00:11.
There is also another possibility. If the temperature inside the cabin went outside the specifications of the stabilized oscillator inside the SDU, the value could have been significantly off. Even smaller deviations from normal temperature would have had some impact (0.3Hz/K) as I pointed out some time ago.
@Viking
https://www.linkedin.com/in/desmond-ross-6799365/
MH370 – 10th Anniversary
@ventus45: I disagree with just about everything he says in this article except this one:
Anyone who claims to positively KNOW the location of the downed aircraft is either delusional or simply attention seeking!!!
It’s always been a game of probabilities.
@viking: @DrB is correct that in my hypothetical route over Cocos Island Airport (YPCC), the BFO error at 00:11 is 12 Hz, which may be beyond what can be expected for the correct route.
@viking, @DrB, @370Location: The route over Cocos Island Airport ending had a BFO error of 12 Hz at 00:11z, which may be beyond what can be expected. However, the route assumes a groundspeed of 354 knots and a track of 145 deg true at 00:11z, which was close to the alignment with RY 15. The reason for the large BFO error is because at this time and location (-21.4,103.0), the BFO would require a groundspeed/track combination such that GS*cos(track) is around 170 kn, which means the track had to be between northeast and northwest rather than southeast.
But if we are going to assume there were pilot inputs to explain the no-find results close to the 7th arc for latitudes south of 25S, how can we rule out maneuvers before fuel exhaustion that satisfy the BTO and BFO values for routes north of 25S? It seems completely arbitrary to allow pilot inputs after fuel exhaustion but no pilot inputs before fuel exhaustion.
If we allow pilot inputs, I’m not seeing a way to properly bound the problem to a manageable search area without following hunches. On the other hand, with no pilot inputs, I’m not seeing a way to define a search area that has not already been searched.
@ventus45
I tried to find the original report from the 6 sailors to CH authorities, but got strange results. The report had recently been deleted from CH archives.
A brief resume existed in another place, but the information was somewhat misleading. The coordinates clearly indicated that they were near Jakarta and sailing away from the Sunda Strait rather than approaching the Sunda Strait. Later on the same day they were on the other side of Indonesia when making further observations. Since the original report is gone, I am unable to see if it is only some kind of typo, or if the report is internally inconsistent.
If the observations were indeed made so far up (near Jakarta), I think it had nothing to do with MH370. It was most likely just waste floating in the water.
However, there could in principle be some kind of typo. If they were indeed approaching the Sunda Strait from the south, the observation would fit with either Ed’s position (if near the coast of Indonesia) or with my position (if it were near CI).
Without the original report, I am unable to judge if any of these speculations may be relevant.
@370Location,
Here is a link to my BTO and BFO calculator from years ago:
https://docs.google.com/spreadsheets/d/1a5K_8u5HOqz__EjE_-WsPx37FN-XECLP/edit?usp=sharing&ouid=118002676424453150092&rtpof=true&sd=true
I have said that my “exclusionary rule” for crash latitudes north of 29S (which I call the “Plan C” Zone) is valid for level flight or an ongoing descent. I am unaware of any plausible counter-example route which is flyable. With a substantial rate of climb (such as your 1,075 fpm) one can match the 252 Hz BFO at Arc 6. However, such a ROC is not sustainable at cruise altitudes. In addition, Arc 6 is not the only time when there is a mismatch in the BFOs. The same is true at Arc 5. So, if you want to match all the BFOs you have to have high ROCs coincidentally occurring briefly at multiple unknowable points in time along the route or over a long period of time (i.e., for hours, which is not possible). Both options seem highly unlikely. Therefore, it is highly unlikely that MH370 crashed north of 29S (in the Plan C Zone).
@All
If the 6 sailors made the observations one day prior to the reported date (March 11 instead of March 12) it fits remarkably well with my position.
It may be pure speculation but it is a funny detail. Could there have been a typo in the report?
I might consider regenerating it from the Internet archive.
@DrB: Your exclusion of paths north of 29S implicitly assumes a range of track values at 00:11z. As I said above, a track between northeast and northwest would satisfy the BFO with level flight. If we allow the possibility of pilot inputs before or after fuel exhaustion, there is a large range of acceptable flight paths.
One approach might be to accept that the crash occurred close to the 7th arc based on the downward acceleration of 0.67g and no IFE log-in and just continue the search northward from 24.84S latitude.
@Viking, @370Location, @Victor,
Victor is correct in saying that the Arc 6 BFO could be matched with a roughly NE to NW bearing at Arc 6.
So far, I have tested my “exclusionary rule” for the cases of no turns or one minor turn (of less than 90 degrees) after 19:41. For bearings between +/- 45 degrees, which probably require multiple turns after 19:41, or at least one turn of more than 90 degrees, there may be matches to the Arc 6 BFO and the Arc 5 BFO. The question is whether or not there is a flyable route from circa IGOGU at 18:40 that can match all the BTOs and especially the Arcs 5 and 6 BFOs.
For example, at 8S and 460 kts, a bearing of 15 degrees would get the Arc 6 predicted BFO 7 Hz below the measured value. However, at 17S a bearing of 45 degrees at 460 kts can match the Arc 6 BFO. That bearing is almost parallel to Arc 6.
If a “plausible” route, even one with a large turn or several turns, can be identified which matches all the satellite data and is consistent with the fuel available, and which ends in an unsearched area, that zone may be a more probable area for future searches than any area south of 24.8S (which has already been searched near the Arc). By “plausible” I mean it should utilize existing waypoints, including possibly a route discontinuity.
@Niels,
Yes, to identify the post-Diversion speed setting, I searched Flight Levels between 350 and 400, and speeds of Holding, MRC, CI52, LRC, M0.84, CI250/M0.85, and CI500/M0.86. There are, of course, multiple combinations of speed and altitude which produce the same true air speed.
My predicted 18:30 parameters are: Latitude 7.24N, Longitude 95.57E, Weight 208.4 MT.
Regarding the possibility of the debris field being missed in previously searched areas, I think our best guess for that number is simply the percentage of the search area which has no data or low-quality data. This is reportedly in the 2% – 5% range. So, in my simplistic view, the probability of the aircraft crashing there is the same number, and the probability it did not crash there is 100% minus that number, or 95% to 98%. That’s sufficiently high, in my opinion, to look elsewhere, especially in areas which are unsearched near Arc 7. We know that all parts of Arc 7 north of 24.8S are unsearched, so they are potential candidates. However, we need to demonstrate a plausible and flyable route which matches MEFE and the satellite data. Finding that north of 24.8S, if it exists, will probably require adding a major turn (>90 degrees) or several turns after 19:41. That’s what the negative seabed searches have done – forced consideration of more complex routes. The assumption of simple routes (with no major turns) led us to Plan A. That didn’t work out. Now we must relax our constraints to look for more complex routes, especially those which end in unsearched territory.
When assessing the validity of the BFO at 00:11 with respect to end latitude, it is helpful to understand the different contributions to the BFO and how they depend on latitude. At this time the satellite 3-F1 was approaching the equator, so the Doppler compensation calculation was close to being correct as far at the aircraft motion is concerned. The biggest contribution to the BFO was the motion of the satellite. How did these depend on latitude?
I ran a quick set of calculations for a range of latitudes from -40 to -5 degrees at longitude 94 degrees (it depends very little on longitude). I also took an aircraft velocity of 470 knots tracking due south (maximiizing the effect.) I find that the BFO contribution from the satellite motion decreased from 57 to 15 Hz, a change of 42 Hz, while the BFO contribution from the aircraft motion decreased from -11 to -16 Hz, a change of just 5 Hz. Thus the BFO is quite sensitive to latitude. This sensitivity is one reason that routes ending far to the north have such difficulty matching the pattern of BFOs.
Since there is a “large number” (INMARSAT report) of possible plausible trajectories, amoung many options, why not factor in the option that Captain Zaharie Ahmad Shah was willing to go to Australia with a complex route at a lower flight level?
@sk999: In your example, you kept the N-S component of ground speed constant. To get acceptable BFO fits for northern paths, the N-S component of ground speed needs to be positive (or at least a lot less negative). Bobby’s plausibility criteria exclude northern tracks.
At the latitude of 21.4S at 00:11z, I calculate the BFO as 246 + 0.034 GS(kn) * cos(track), with a match with the measured value at GS(kn)*cos(track) of 170 kn. Others have already proposed northern paths with a turn towards the north towards the end.
A turn in the path is not without disadvantages, but at this point, EVERY option that’s left has a disadvantage.
@All
While you are reviewing a number of proposed locations on the 7th arc, does anyone have further questions of comments on “A Voyage to…”. I am dealing with the period after00:11 separately.
@DrB
Thank you for sharing the estimates for 18:30UTC.
Regarding the possibility that the debris field was missed: I imagine it is especially tricky to assign probabilities when data quality is low.
In the southern half of the S33-S36 deg range there are significant patches with low confidence coverage (estimated at 7.8% of all the area). Most of it has never been re-scanned by OI (also not in 2018 as far as I know).
It’s the highest probability latitude range based on “simple” path BTO/BFO data fit, so I would be in favour of eliminating gaps or uncertainties there first, before moving to other areas.
@DrB
Thank you for your Excel (BTO/BFO/fuel/etc) tools.
@Gregg
“why would he just fly straight to the end?”…exactly.
CAPTION/CAPTIO group (Jean Luc et al) has always been willing to embrace active pilot, but they are not really going anywhere with the concept except a nearly straight south path. There needs to be a vision of the motive/where the aircraft was possibly heading. Which I submit the home sim data intersection with Arc7 is the best/only evidence we have, and I am pretty sure it works.
@Ventus45:
I looked into the Swiss tourist report of Sunda Strait flotsam years ago. The most notable item was a wallet, which should have been solid evidence. Nullschool models show current was flowing out of the channel into the SIO, and surface winds were WNW across the channel. Google Earth Pro shows a shipwreck at -5.8324 104.8326 that appeared after 2013, but I couldn’t pin the date of the event. (OpenAI-GPT-OSS-20B hallucinates it with multiple news headlines, but recants when asked for references.)
@sk999:
Thanks, Steve, for looking into the BFO constraints.
@Victor:
Thanks always for the bigger perspective.
Rather than choosing between hunches, I believe a forensic approach of exploring new evidence is the best path forward.
For example, recordings from the infrasound array at Cocos Island were prematurely dismissed due to a time zone error. If the plane flew anywhere near Cocos, that could reveal more clues to the flight path. As custodian, Australia could release that data to the public, independent of CTBTO contract restrictions. Australia could also examine coastal defense radar from two installations on the north end of Christmas Island.
@DrB:
Thanks much for your useful spreadsheet. The numbers I get are similar to v17 when plugging in my waypoint path values at the 6th Arc. A climb of 1150 fpm nulls the BFO. If the plane were not on autopilot, that shallow 200 ft/mile 1.9% climb might not have been intentional as fuel was about to run out. Of course a steady climb would not be feasible throughout the flight. I don’t have a model for fuel exhaustion to run the Java path, but Boeing simulated minimal turns at each arc for a similar path with fuel exhaustion only slightly farther over Java. I’m encouraged by your shift on Plan C from “impossible”, to “unlikely”, to “Now we must relax our constraints to look for more complex routes.”
What’s important at the moment is whether OI might see some consensus for doing a very specific search in calmer waters before their contract expires. Expanded searches may need to wait for new findings and future contract negotiations.
All,
My focus has shifted as a result of OI’s negative search results as far north at 24.8S.
This new information, that it appears highly likely the debris field is NOT near Arc 7 from 39.5S to 24.8S, implies that it must be north of 24.8S if one also believes the debris field is highly likely to be within 20 NM of Arc 7. I believe this is true based on the end-of-flight BFOs and the cessation of transmissions. So now I should look north of 24.8S, but how far north? The the implication of the negative seabed searches is that a late major turn might have occurred. That would explain the negative searches and put the now-most-likely location north of 24.8S.
In 2014 it was reasonable to see if simple routes with no major turns would match the satellite data. There were many that came close, and so a seabed search effort was mounted, based primarily on DSTG’s Bayesian analyses and on individual routes proposed by many of us still participating in this forum.
There were also nagging differences in certain BFOs. Obtaining a completely satisfactory match to all the BFOs proved elusive. In particular, three BFOs were problematic.
One problematic BFO was at 18:40. In particular, this minute-long sequence of BFOs seemed to show an ongoing maneuver because it appeared to indicate systematic changes during that minute. The stability (smoothness) of the BFOs also did not seem consistent with an ongoing descent at a high ROD. I believe I now understand how these BFOs occurred (during a 156-degree turn from abeam IGOGU to VOBX to IGREX), as I posted previously. That turn appears to discriminate the outbound waypoint to be IGREX, because no other nearby waypoint can produce the “upturned end” seen in the 18:40 BFOs.
The second problematic BFO was at 19:41. Closely matching it with a course with no significant turns after 19:41 proved impossible. In UGIB we proposed a small ROC at the end of a climb. My recent studies have shown that a bearing of circa 150 degrees is needed to match the 19:41 BFO. It now looks like a leg from LAGOG to BULVA will fit both the Arc 2 BTO and the BFO rather well. I note that LAGOG was one of the waypoints in the “home simulator” route. I also note that BULVA is reached at a consistent time to match the 19:41 satellite data using a VOBX to IGREX to VOCX to LAGOG to BULVA path.
Then it is clear that a 30-degree, right turn toward the south may have occurred in the vicinity of BULVA, because otherwise there would be a large overshoot of Arc 3 at 20:41. We know from Inmarsat’s paper and from UGIB that a nearly-due-south path between Arcs 2 and 4 is called for. On possibility is that the next waypoint after BEDAX is POSOD.
The third problematic BFO is at Arc 6. A bearing of 84 degrees matches the 252 Hz BFO at a latitude of 24S (just north of the zone excluded near Arc 7 by the seabed searches). With a BFOR of +7 Hz, the predicted BFO is 252 + 7 = 259 Hz and the matching bearing is 53 degrees. With a BFOR of – 7 Hz, the predicted BO is 252 – 7 – 245 Hz and the corresponding bearing is 115 degrees. Thus, we see nearly eastbound bearings being needed, from 53 degrees to 115 degrees, at 25S on Arc 6. So, how does the bearing change from south to east between Arcs 4 and 6? By a large turn to the left. Does this turn occur before or after Arc 5?
Here is a short paper showing a notional waypoint route to YPCC at Cocos Island between Arcs 4 and 5.
https://drive.google.com/file/d/1t_vQcsKlLaewFG-NrckchumE5lFPczz7/view?usp=sharing
This partial route matches all the satellite data through 21:41. I have not been able to match Arcs 5 and 6. Maybe someone can figure this out. I would also be grateful to receive information on aviation waypoints in the general vicinity of Cocos and Christmas Islands, BEBIM, and RUNUT, if any other nearby waypoints exist.
@TBill:
Thanks for the response. Another question. The designated search, which has been 95+% searched, is based on no active pilot input after the turn south, with the exception of possibly some pilot intervention at the very end? If there was an active pilot intent on deception does that invalidate the previously searched area? If there was an active pilot are we done to hunches?
@DrB:
I look forward to where this BFO analysis of multi-turn paths may take us.
A wide area around 7th Arc was also well searched by air from 19.5S to 21.5S allowing for drift from the ULB pings. That’s the northernmost boundary of any SIO searches that I’m aware of.
I have an RNAV.kmz collection of global waypoints found over a decade ago. I then weeded out points related to the Indian Ocean and MH370. The two files are here:
https://drive.google.com/drive/folders/1XjqcSiTOTvhBMrgvlBnL2KYR3CQs5Viw
@DrB
It seems to me that any discussion of the early portion of the flight after NILAM ought to at least address Kate Tee’s sighting. Whether you agree with it or not, it should be part of the conversation.
@370Location,
Thanks for the list of SIO waypoints. I’ll see if it has any I have missed.
The ULB detections were very probably real, but none were from 9M-MRO’s data recorders. The detected signals were too strong, were not stationary in location, and some had the wrong frequency. They were probably pingers attached to drifting fishing nets.
Regardless of the validity of the pinger detections, all surface searches are valuable information about the crash location on Arc 7.
@Sid Bennett,
I don’t know which aircraft Kate Tee saw, or when she saw it. My current route is too far from her ship’s track to match her description.
@Gregg
In my view, we have an assumption problem of (1) no maneuvers to Arc7, (2) fuel exhaustion at Arc7, (3) crash not at a planned location, (4) meaninglessness of pilot’s home sim data, (5) etc. Relax those assumptions, and a number of different solutions come into play, including the home sim data inferred flight path to 29-32s. I consider home sim data as evidence, because it looks to me potentially where MH370 actually flew.
Part of the rationale for ruling out such solutions, in the past, has been the excess residual fuel at Arc7 which conflicts with assumptions above. Even if MH370 did fly this way to 29-32s, the next problem becomes the apparent “getaway” flight segment after Arc7: heading and distance flown.
All,
Here is a link to a research note describing a new route ending at 23S which matches all the satellite data :
https://drive.google.com/file/d/1-80fdowspSSu0X99K9LNWIg16n_Xoyuz/view?usp=sharing
@DrB
In your March 29th post, 4:58 pm, you wrote:
“In particular, three BFOs were problematic”; the 18:40 UTC BFO being one of them.
I was checking the BFO for my hypothetical paths that I used to estimate fuel consumption between 18:30 UTC and 19:41 UTC, and actually it seems it is quite easy to reproduce the 88 Hz BFO (18:40) for example assuming after 18:30 track changed to about 200 deg and Mach number to around 0.82 at FL 400. Actually the 200 deg is in the direction of the most probable 19:41 position that I found in my analysis, connecting to a FL 370, CTT 182, LRC path after 19:41. I’m not saying this is what happened, I just want to point out there are reasonable alternatives to explain this BFO value compatible with a rather simple set of assumptions. Typically these are the “early turn” scenario’s.
To my limited knowledge and correct me if I am wrong.
With, for example, the spreadsheet program (SK999-Satellite Model) developed by Yap Fook Fah Associate Professor, School of Mechanical & Aerospace Engineering, anyone can calculate the BFO when entering the six components of the position and the speed of an aircraft at a given moment (time of passage to the arcs, thanks to the BTO).
If the calculated BFO value differs by less than 7Hz from that measured by INMARSAT (uncertainty resulting from INMARSAT analysis from recording of other flights), then the position and speed of the aircraft are considered as valid.
Therefore, with the BTO and BFO, one cannot calculate the positions and speeds of the aircraft, but it allows to check the plausibility of a trajectory based on operational and technical considerations.
In the INMARSAT Report, the selected trajectory is presented in Figure 17, As an Example, Flight Path and, in the Conclusions, as A potential flight path.
There is a “large number” of possible plausible trajectories.
In 2018, the original CAPTIO team, I was part of, assumed an active pilot until the very end and came with
https://www.mh370-caption.net/wp-content/uploads/MH370-PlausibleTrajectory-3.4.pdf
and assumed that Christmas Island was targeted.
I left the CAPTIO team because I thought that, even being smart, a group of amateurs could not come up with a trajectory convincing enough to orientate expensive research.
Indeed, there are two issues: the fidelity of FMS simulators and the Cost Index selected by the airlines.
Flight Management System algorithms are treated as proprietary, because they sit at the core of certified avionics and directly affect safety, performance, and regulatory compliance. Manufacturers such as Honeywell, Collins, Thales, and Airbus guard these algorithms closely.
The actual Cost Index values used by each airline, on each fleet, and under each operational scenario are not published.
Cost Index directly reflects an airline’s internal cost structure. Revealing it would expose sensitive details such as:
• The relative weight the airline assigns to fuel cost vs. time‑related costs (crew, maintenance, schedule pressure).
• How aggressively the airline optimizes fuel savings or on time performance.
• Operational strategies for delay recovery, network efficiency, and fuel hedging.
In the Investigation Report, there is no information at all on Honeywell Flight Management System algorithms or on MH370 Cost index.
Without these data, only Boeing, Honeywell, Inmarsat and Malaysian Airlines could carry any new accurate enough scenario to end up with a trajectory precise enough to locate the wreck. But this is no longer their business, and Boeing has other fish to fry.
If one keeps willing to build a scenario and trajectory, on one hand, the BTO/BFO constraints must be relaxed and, on the other hand, the research zone is going to be large.
Back to CAPTIO/CAPTION: on his own, Jean-Luc Marchand came with a new trajectory:
https://www.mh370-caption.net/wp-content/uploads/MH370-PlausibleTrajectory-CAPTION.pdf
assuming that Learmonth was targeted and as well, the assumption was the pilot in command until the very end.
I do not have either competence or willingness to compare it with the “CAPTIO” one.
Ed Anderson’s theory, MH370 Acoustic Analysis for Implosion, is based upon Australian ATSB: operational-search-for-mh370_final_3oct2017 Page 123
And Appendix J: Seismic and hydroacoustic analysis relevant to MH370
In 2019, Ed Anderson came up with a trajectory
https://370location.org/2019/02/a-consistent-mh370-waypoint-path-to-a-specific-7th-arc-location
With such a complex trajectory, the assumption is that the pilot was in command until the very end, but for which target?
There is another interesting theory from Randy Rolston Seventh Capricorn
https://mh370report.com/
supporting MH370 crash point east of the 7th Arc between S23° and S24°.
With Richard Godfrey and Ed Anderson, I dared to comment
Richard Rolston, as well, writes: Without concrete evidence, it was presumed that the autopilot had been active for the last 5 ½ hours of flight, that the crew was incapacitated due to hypoxia, and that no one was controlling the plane at the end of the flight.
Lately, Jean-Luc Marchand decided to team up with Patrick Blelly and to support his trajectory
https://www.mh370-caption.net/index.php/captain-blellys-trajectory/
Captain Blelly’s innovative assumption is that the person in command did not wait for the fuel to be exhausted and probably decided to shut down voluntarily the last running engine timely before running out of fuel.
Therefore, the pilot was in command until the very end, but without any target, a suicide?
I do not have either competence or willingness to compare it to other “south” trajectories.
As far as I know, all the South trajectories assume that the pilot was not in command after the Northwest of Sumatra and committed suicide.
As far as I know, the No find, No fees OI /Malaysian Government was not made public.
No information was given on the area to be searched and on which assumption?
Had the wreck been found, whatever the credit claimed by those whose end of trajectory would have been the closest to the wreck, it is unlikely OI would have shared both the glory/PR and 100 K US dollars.
Anyway, the wreck was not found at the end of these south trajectories, including the INMARSAT “example”
Isn’t it time to forget suicide as the motive for the hijacking?
None of the few cases of pilot suicide occurred in this way. There may be countless reasons for suicide and ways to do it, but this scenario seems unlikely. No known element of the private and professional life of Zaharie Ahmad Shah, the respected captain, explains such suicide. It was fabricated that Zaharie Ahmad Shah had money problems and/or a mistress. The investigation report describes him as a respected professional. Within his family, he was the “uncle” to whom advice is turned.
With the assumption that Zaharie Ahmad Shah was willing to go somewhere, it could be possible to build a trajectory piloted until the end, with relaxed BTO/BFO constraints.
@370Location
You said “Google Earth Pro shows a shipwreck at -5.8324 104.8326 that appeared after 2013, but I couldn’t pin the date of the event. (OpenAI-GPT-OSS-20B hallucinates it with multiple news headlines, but recants when asked for references.)”
I asked Claude, an AI assistant made by Anthropic: “Google Earth Pro shows a shipwreck at -5.8324 104.8326 that appeared after 2013. Can you find the date of the shipwreck?”
Answer:
“The coordinates -5.8324, 104.8326 are near the Sunda Strait/Lampung coast. Let me check the shipwreck log more specifically and look at the exact location more carefully.Searched the webSearched the webInteresting — a result mentions a shipwreck at Kedu Beach, Kalianda, Lampung. Kalianda is very close to those coordinates. Let me investigate this further.Searched the webSearched the webThe coordinates -5.8324, 104.8326 place the wreck very close to Tanjung Putus, near Kalianda, South Lampung. One result mentioned “Shipwreck Tanjung Putus Lampung.” Let me look into that specifically.Searched the webSearched the webThe coordinates (-5.8324, 104.8326) are actually near Kalianda/South Lampung, not Tanjung Putus (which is further north near Bandar Lampung). Let me try a more targeted maritime incident database search for that area.Searched the webSearched the webThe 2018 Sunda Strait tsunami caused by Anak Krakatau is very relevant — it damaged hundreds of boats in the Kalianda/South Lampung area. But Google Earth showed this appearing after 2013, which could mean anything from 2014–2018+. Let me check if the 2018 tsunami caused a notable shipwreck near those exact coordinates.Searched the webSearched the webThe 2018 tsunami was very destructive to the Kalianda/South Lampung coast and damaged 350+ boats. However, Google Earth imagery showing the wreck appearing after 2013 doesn’t necessarily mean the 2018 tsunami — it could be any year after 2013. The coordinates -5.8324, 104.8326 place the wreck specifically near the Kalianda coastline. Let me try to look at vessel incident databases more directly, or check MarineTraffic for any known incident near those exact coordinates.Searched the webSearched the webLet me try the Shipwreck World database directly, which has GPS-based records, to check if anything near those coordinates is listed.Fetched: Wreck Finder – Maps & GPS Location CoordinatesFetched: Wreck Finder – Maps & GPS Location CoordinatesThat’s a map-based UI that requires interaction. Let me try one more targeted approach — looking at this from the angle of the 2004 Indian Ocean tsunami, which also affected the Lampung coast, or checking for specific vessel incidents in that exact sea area using a maritime incident database.Searched the webSearched the webI’ve done extensive searching but unfortunately cannot find a specific, confirmed record for the exact shipwreck at -5.8324, 104.8326. This is in the Teluk Lampung (Lampung Bay) area near the Kalianda coastline of South Sumatra. Let me try one more targeted search using the Indonesian local media.Searched the webSearched the webAfter extensive searching, I haven’t been able to find a specific, confirmed record for the shipwreck at exactly -5.8324, 104.8326. This is a very specific location in Teluk Lampung (Lampung Bay), near the Kalianda coastline. Let me be transparent about what I’ve found and what the most likely candidates are.
Unfortunately, I wasn’t able to pin down the exact date of this specific shipwreck from publicly available records. Here’s what my research did establish:
Location context: The coordinates (-5.8324, 104.8326) place this wreck in Teluk Lampung (Lampung Bay), near the Kalianda coast of South Sumatra — a busy but regionally modest maritime zone. It’s not on the heavily-trafficked Merak–Bakauheni ferry route itself, but nearby.
Most likely candidate events (after 2013):
The most significant maritime disaster to affect this exact coastline after 2013 was the December 22, 2018 Sunda Strait tsunami triggered by the collapse of Anak Krakatau. The tsunami struck around 21:27 local time after large parts of the southwestern side of Anak Krakatoa collapsed into the sea. Wikipedia Hundreds of boats and over 500 houses were damaged FloodList, with Kalianda being one of the directly affected areas. A vessel beached or sunk during this event could appear as a new wreck in subsequent satellite imagery.
Other possibilities include routine local vessel losses — smaller cargo ships, fishing vessels, or ferries that went aground in the area between 2014–2018 — which often go unreported in English-language maritime databases.
Why I can’t give you a precise date:
Indonesian local maritime incident records for smaller vessels in this area are not well indexed in English-language databases
The specific coordinates don’t match any named incident I could find in shipwreck logs, MarineTraffic incident reports, or news archives
Google Earth’s historical imagery sometimes shows wrecks that were never formally reported
My recommendation: The most productive next steps would be to:
Check the Indonesian KNKT (Komite Nasional Keselamatan Transportasi) accident database
Contact KSOP (Kantor Kesyahbandaran dan Otoritas Pelabuhan) Panjang, the port authority for that area
Try the Global Maritime Wrecks Database or MarineTraffic’s incident history directly with those coordinates
If you can share a screenshot of what the wreck looks like on Google Earth Pro (size, vessel type), that might help narrow down the search significantly.”
@Niels,
The problem with the 18:40 phone call BFOs is not that it is hard to find a suitable course, it’s how you connect that course to (a) the right lateral offset and (b) a waypoint-driven method of lateral navigation, or a plausible alternative, and (c) an acceptable SIO Route.
It’s not good enough just to fit some arbitrary lat/lons. Sure, you can just turn off the parallel track to N571, but to which waypoint?
Until now, I have never seen a proposed path that turns south from N571 to a specific waypoint and then connects to an acceptable SIO Route with a plausible method of lateral navigation.
If you have one, I would like to see it.
@DrB
What are your objections to my recent paper?
@Jean-Marc Garot:
I understand the YapFF spreadsheet was one of the first to predict BFO, and was refined along with others over time. I don’t know how accurate the 2 sigma BFO cutoff is for validating a single ping measurement. My only contribution to the BFO was that there has been some consensus that the plane was flying tangent and closest to the satellite around 19:41, so subtracting the known orbital movement should give a stable reference frequency bias after the first reboot.
The Cost Index appears to only apply to a normal cruise level flight to shorten crew time. I expect that after a major electrical outage, a piloted path would either maximize distance (LRC) or endurance (holding speed). Endurance makes more sense to me, and fits with a low and slow flight path toward daylight while attempting repairs.
My path is working backwards from a specific spot on the 7th Arc. The distant hydrophones all point to an anomalous event in the general area, but it is the seismic epicenter that provides accuracy.
I’ve tried to avoid speculation about where the plane was heading, or why. Unique seismic detections at the island airport are consistent with the path, but we can’t know why the plane didn’t attempt to land. Not knowing the destination shouldn’t invalidate evidence. We also can’t know that the noise event wasn’t some extremely rare geological quirk that happens to match with a seabed depth impact just as MH370 debris was sinking. (I had set out to look for an implosion as suggested by experts, but later realized that the seismics match simple P and S wave arrivals in the crust).
Some have tried to rule out the Java site because they are sure that any competent pilot must have attempted a landing at Christmas Island before running out of fuel. That might be a fine assumption under normal flight conditions, but we can’t know what the plane was capable of – only that it continued flying.
I believe that MH370 theories about pilot suicide all stem from an assumption that the plane was unpiloted on a flight to oblivion. That probably comes from an apparent fit of the BTO to a simple curve over time. However, we know that the first pings are not on that curve, but a mirror of it. Consider that later pings were also mirrored points, leading back east toward Cocos, Christmas Island, and Java airports.
@John
You AI dump is similar to the “thinking” I saw, but the LM Studio software I’m using hides all that and then shows just the results.
The Krakatoa tsunami on Dec 22, 2018 took hundreds of lives and was devastataing. However, the earliest image showing the wreck I mentioned was on May 24, 2018. There was another wreck in April 2014 on the east side of the Sunda Straight, after the Swiss reported spotting debris.
I think that because the currents were SW out of the channel and more debris was spotted to the NE in the Java Sea, that the debris was probably unrelated to MH370.
No report on the floating wallet contents probably means it wasn’t consistent with an MH370 origin. I figure most investigations don’t get mentioned either because they turned up contradictory results, or they weren’t done at all. For example, the RMP must have explored Zahari’s web browsing history, but nothing suspicious was reported. Only the copilot’s cell phone was detected in flight, but it’s likely that only the crew was investigated due to privacy issues and corporate telecom constraints. Passenger phones may have been connecting all along the flight path.
@DrB
Ok, I see your point regarding the 18:40 BFO. But I don’t see how you could prove a “complete” path proposal that fits the data and is connected by waypoints is “the” unique solution.
As you have noticed I’m not that enthusiastic about your recent proposals. I don’t understand why the authors of the “UGIB” milestone paper don’t stand for their main results. It may have been a mistake to focus too much on the exact 180deg path, but other than that the paper is still rock solid. And actually the 2-5% probability that the debris was missed in a previous search in the “plan A” area was already known when “UGIB” was published. So, if you agree on the “no glide” scenario being most likely, then it is hard for me to see what has changed, and why then you changed your mind before the S33-S36 area search being fully completed.
@370Location: Some corrections:
ECON mode with a Cost Index (CI) is the most common way that operators balance fuel and time costs. The data suggests that CI=52 was selected for the expected route of MH370
LRC is a legacy mode that is similar to ECON except that CI can’t be changed (equivalent to around CI=150) and there is no compensation for wind speed.
To maximize range, ECON mode is selected with CI=0, which forces the speed to MRC.
If a crew needed time to make repairs (or to run checklists), it would be odd to fly away from a suitable airport. Rather, they would stay close and enter a hold or fly a box pattern until ready to land.
@Niels: UGIB 2020 checks all the boxes except the plane was not found close to the 7th arc at the predicted crossing latitude. Before the recent search, there were some areas close to the Last Expected Position (LEP) that were missed by previous searches. Those areas have now been searched with null result.
As you’ve heard me say, there are no good options left, which is why it’s hard to be too critical about presentation of new areas.
@Victor Iannello
“As you’ve heard me say, there are no good options left, which is why it’s hard to be too critical about presentation of new areas.”
I don’t think you are right about this statement. ATSB search area close (up to 25NM, give or take) to the 7th arc north from 36°S was noticeably less covered than the area south of 36°S. I showed at least a dozen of spots which contain no or low-quality data that are at least kilometer in size, and that are well charted in ATSB maps. I don’t know why anyone would choose to stay blind to that fact, especially now when the plane is still not found.
@Marijan: Ideally, we would have 100% coverage of the areas we considered to be most likely. Practically, those areas represent a small fraction of what was searched, and Ocean Infinity has shown no willingness to do a complete fill-in of unsearched areas. There is also an unwillingness to consider that the debris field might have been missed, or that isolated contacts identified as manmade were aircraft parts.
Considering the financial risk that OI assumed for doing the search, it is hard to find fault with OI’s pursuing whatever they consider the mostly likely areas to find the debris field.
@Niels,
I have not “given up” on the UGIB LEP. It would be nice if it were possible to search a zone to 100% coverage before deciding what to do next. In the real world, once you get close to 100%, other areas may then become more probable, considering their relative probabilities. The question is, are there any routes to unsearched areas which offer a better chance than those small, missed areas south of 25S?
I don’t know the answer to that question yet. I have proposed a candidate route to 23S, and it has some attractive features, but it needs more scrutiny to understand the BFORs at Arcs 5 and 6.
I have been studying the Arc 6 BFO in detail, including how the bearing at Arc 6 also affects the Arc 7 BTO. The angle from perpendicular as the path crosses Arcs 6 and 7 is bounded by the observed difference in the Arc 6 and 7 BTOs (i.e., by the distance between those arcs). Thus, you have two constraints on the Arc 6 bearing – you have to match the BFO at Arc 6 and your path has to be in the ballpark of matching the distance between Arcs 6 and 7 calculated using the predicted INOP time and deceleration and the post-MEFE time and deceleration.
It appears that an Arc 6 latitude south of 16S is required in order to match that BFO within +/- 7 Hz. To match it exactly requires a latitude south of 21S. I don’t see how latitudes north of 16S can be consistent with the Arc 6 and 7 satellite data.
I have never said that a candidate path that matches the satellite data and MEFE must be unique and therefore be correct. I do think those two conditions must be met (i.e., they are necessary but not sufficient) before one can recommend mounting a seabed search to that predicted crash location. I have always said the only way we will ever be 100% certain a route is the correct one is to find the debris field.
In the meantime, it seems clear OI will not return to complete the missed areas. The only way I can provide motivation for a renewed seabed search is to identify a plausible new route ending in an unsearched area. That’s what I am searching for currently.
@Marijan
I share your doubts.
Remember the ARA San Juan ?
OI “missed it” the first time.
An imploded submarine leaves huge items of debris, large hull sections etc.
They are easier to see on side scan sonar than two engine cores and two MLG’s.
I think we have been way too confident in the ‘efficacy’ of the search.
I particularly worry about the ‘efficacy’ of the early towed sonar searches.
We are at a pivot point.
We are either certain of the Inmarsat analysis – or we are not.
If we are certain, we missed it, we have to start again, “from square one”.
If we are not certain, consider other analysis, and search their areas.
@Sid Bennett,
You asked: “What are your objections to my recent paper?”
It predicts a crash in a zone already searched with a high-90’s probability of a negative result.
@Dr.B
Any other issues?
@Victor:
Thanks for the correction of LRC to ECON/MRC.
@DrB: wrote “I don’t see how latitudes north of 16S can be consistent with the Arc 6 and 7 satellite data.”
Is that again limited to level/descending flight at Arc 6?
@Ventus45:
I think we can be confident in the INMARSAT data, but question the assumptions used in interpreting it.
@DrB
You seem to be 100% sure that no routes ending north of 16S are possible based on the two BFO values from arcs 6 and 7. I disagree with that conclusion.
Initially there are possibilities for an active pilot to fly in special ways to produce matching values. I do not believe that option is likely, since he would have needed highly specialized knowledge about BFO to achieve this.
On the other hand, there is also the thermal issue I pointed out long time ago. If he had turned the heating off and opened a door, window or hatch, the temperature in the cabin would have been extremely low near arc 6, and low but rapidly increasing near arc 7. That could easily explain the BFO values. Actually it fits best with endings north of 16S.
This scenario also explains the very strange power levels received. Until arc 6 the received power (at the INMARSAT satellite) increased with distance of the airplane from the ground projection of the satellite. All 3 terms contributing directly to the received power predict decreasing power. The simplest term (naive Fraunhofer diffraction) predicts only an extremely small power decrease, but the two other terms are significant. One of these is the effective area projection of the antenna area perpendicular to the satellite direction. This area should be used in the Fraunhofer calculation to get better agreement. In addition there is a more advanced Fresnel contribution related to polarization. It gives a comparable negative contribution to the received power.
Therefore it is totally impossible to understand that the received power is increasing with distance. The only possibility I can come up with is that the oscillator inside the SDU (perhaps assisted by a pre-amplifier inside the same box) gives more power at lower temperature. Most microwave equipment from the time when the SDU box was developed behaved like that. However, as some people pointed out the power amplifier beside the SDU box is so carefully temperature stabilized that it would effectively be insensitive to cabin temperature.
The received power level at arc 7 is actually falling towards expected values in agreement with this model.
A small correction:
Fresnel should be replaced by vector in my last contribution.
@All
Please note that the behavior at arc 7 is complicated by a channel change plus a rapid downward acceleration, so the power drop at the end may have more than one reason.
@DrB, VictorI
Victor wrote: “Before the recent search, there were some areas close to the Last Expected Position (LEP) that were missed by previous searches. Those areas have now been searched with null result.”
Imo the main issue with UGIB 2020 and with the statement above is the precise concluded LEP and 180 deg track. The data does not support such accurate conclusion. For me the main result has always been your fig. 5 (the combined probability density function).
@Viking,
You said: “You seem to be 100% sure that no routes ending north of 16S are possible based on the two BFO values from arcs 6 and 7. I disagree with that conclusion.”
I am not 100% sure. However, there is strong evidence for this. I’ll provide a link today to a new plot illustrating that conclusion. To answer 370Location’s question, it assumes level or descending flight for Arcs 4-6.
If you allow unknown and unprecedented variations to occur in the BFOs at Arcs 6 and 7, you are simply assuming the BFO data are unpredictable and are therefore unreliable.
I submit that ignoring those BFOs expands the Arc 7 range of allowable routes to the point where there is insufficient discrimination to identify a particular area with enough specificity to motivate an affordable search in the future.
My assumption about BFOs is different than yours. I assume the BFO Bias may drift with time, but that drift should not exceed the drift rate seen over the past 24 hours. Thus, I allow for “normal” drift, but not for instantaneous “steps”. I also use ALL BFOs.
Inmarsat’s approach is different in theory but not in practice than mine. They used one average value of BFO Bias for a flight and expected the BFORs to be within +/- 7 Hz. I allow a linear drift but use a smaller value for sigma than DSTG’s Figure 5.3 (because that BFOR distribution is broadened by the uncompensated intra-flight drift). Actually, in my route fitter I use all three methods, and the results are virtually identical. So, my results are not affected in any material way by which BFO acceptance criteria I use. It only matters (a lot) that I use those BFOs in assessing a candidate route.
By far the most effective discriminator is my assumed requirement for waypoint navigation, at least until a single route discontinuity might have occurred. So far, I have not been able to achieve an acceptable route using a transition from geodesics to constant magnetic heading, although I have searched quite a bit.
For crash locations north of 25S it is a guessing game (not a number-crunching game). I try to find a combination of waypoints such that the satellite data are satisfied using plausible altitudes and air speeds and the predicted MEFE time is within the margin of error of the fuel model.
@Viking: Re:
“Therefore it is totally impossible to understand that the received power is increasing with distance. ”
As I have explained many times on this blog and in various papers, there is nothing unusual about the MH370 GES receive power time series. Yes, the receive power will decrease with distance, all other link terms being held constant. (path loss=-37-20*logD-20*logF). But all the other link terms are not constant. For example, the AES power output changes according to SDU control to maintain a nominal link margin. There are about 10 other variables in the link analysis (antenna gains, transponder loads, etc.), all of which can vary over several hours. Thus, it is actually easy to understand the slight increase in receive power.
If you never search where the wreckage is likely to be you will never find the wreckage.
In July 2015 Geoff Hyman, Barry Martin and I posted a paper to Duncan’s blog espousing the hypothesis that MH370 followed a gliding descent path after fuel exhaustion. I believe there were other posts on the subject. But, perhaps it was too soon after the event and all of us were still trying to come to grips with the data known at that time.
The Boeing simulations and interpretation of the post-fuel-exhaustion BFO has led to the limitation of the underwater search to a narrow stripe along the 7th arc.
We all seem now to agree that this search area is unlikely to prove fruitful. I was motivated to revisit the glide hypothesis in light of the analysis and discussions of this blog group and the slow leakage of information or posting of manuals on the web. Late last fall I started the study reported in the paper that can be found as a draft at:
https://www.dropbox.com/scl/fi/b5k4ug7filydizherpqvm/MH370-Post-Flameout-Descent-DRAFT-032326.pdf?rlkey=y26m478llnskumk4j9nf68mr2&st=gyw84zjo&dl=0
Some of the references may be hard to find and I have posted those I have copies of to:
https://www.dropbox.com/scl/fo/01d6bci78sddcaiy9ufrd/ALzPuV3ZzgVk0FZFPfd0JqY?rlkey=hwyachkp2lfvf9b56j1gva195&st=cewk3jqw&dl=0
If you have other references you would like to cite either to support or refute the hypothesis please let me know where they can be found.
There is a lot of material to cover and this is a DRAFT, so please comment or ask questions if you find the paper interesting or confusing.
I think this fresh look at the final minutes of the flight might motivate a change in the search concept.
Let me know what you think.
All,
Here is a link to my research note on the Arc 6 latitude range for two cases, with no turn and with a major turn:
https://drive.google.com/file/d/1YGffvYi7oMeqoom-0Yit7nIVpY3TnbHn/view?usp=sharing
With no turns, the Arc 6 latitude range compatible with the 00:11 BFO is from 40S to 28S.
With a major turn before 00:11, the Arc 6 latitude range is from 40S to 11S.
With a major turn and also considering (a) the distance traveled between Arcs 6 and 7, and (b) the negative seabed search near Arc 7 south of 24.8S, the Arc 6 latitude range is from 25S to 16S.
I believe this is the best alternative for identifying a new candidate route which might be recommended as a new search location.
@Sid Bennett,
I think your long, slow glide theory is incompatible with the physical evidence of the debris. Large trailing edge components (flap and flaperon) appear to have departed from the right wing prior to sea-surface impact, as would be expected in an uncontrolled descent with overspeed and flutter occurring prior to impact.
@DrB.
Can you put the IG best estimate and possibly UGIB on this plot to hep me visualze it?
@DrB
You are a quick reader. The parameters at touchdown are TAS 200kt, rate of descent 1200fpm, pitch -3.5deg. Someone might want to comment on what that would do in a moderate sea state.
@Sid Bennett: Your central thesis is the flight control mode remained in Normal and did not revert to Secondary. However, based on our research and information that Andrew has, we believe that the reversion to Secondary occurred due to the loss of pitot heat after the second engine flameout, as the air data is then considered to be “unreliable”.
In any event, are you proposing that the Boeing simulations were erroneous in reverting to Secondary mode? As well as the Level D simulations conducted by Mike?
@VictorI
The simulator results are highly dependent on the initial conditions, and the only information I have about the Boeing simulation came from your report. The Boeing report case 2 represents a pathological setting of the state of the AC busses causing autopilot disconnect at 1FO.
The FAA certification requirements for ETOPS would imply that “normal” mode operation has to continue for some time.
A pitot tube icing event would result in loss of proper air data, but the ADIRU is on the hot battery bus. I do not have access to documentation from Boeing to determine if the pitot tube heater is operable. Do you>
@Victor, you said: “UGIB 2020 checks all the boxes..” Let’s stress-test that statement a little.
1. UGIB 2020 relies upon starting the “automated flight” from an Arc 2 latitude that is incompatible with last known position unless you insert a loiter of ~40 minutes.
2. UGIB 2020 BFO at 1840 is incompatible with the path model direction unless a ROD is posited, while we also know that the 1840 BFOs are too consistent to be explainable by non-level flight.
3. UGIB 2020 BFO at 1941 also relies on a non-zero ROD to eliminate an unacceptable error. The explanation is not only that a manoeuvre just happened to be occurring at the time, but also exactly the right end stage of that manoeuvre to produce the required non-standard ROD.
4. UGIB 2020 is incompatible with expected MEFE timing except by explaining the shortfall by recourse to packs-off, with a commensurate fuel saving that is indirectly inferred from B777 documentation
5. UGIB 2020 offers a poor match to the best-fit drift origin for the flaperon (+/-38S) once CSIRO had incorporated revised wind factor (% and directional offset).
6. UGIB 2020 BTO fit – compared to the spectrum of LNAV path models from Arc 2 – has marginal relative probability. Not impossible, but *much* lower probability than best-fit BTO path models.
7. Other than the data holiday holes, the end zone had already been searched out to ~30NM either side of the arc. Impact beyond 30NM is counter-indicated by various aspects of the end-of-flight evidence and was entirely incompatible with the “packs off” requirement of the UGIB path hypothesis.
Boxes were indeed ticked. But only by stretching credulity in several directions at once.
@VictorI
The pitot tubes are on the AC bus and would lose power when the APU finally quits. But at FL40 the pitot tubes should not ice up if the air is clear. Do we know much about the weather in the area at the time?
@Sid Bennett: I assume you mean FL400, not FL40.
I don’t understand how ETOPS certification is relevant. If you lose an engine, you still have the other engine and the APU to generate electrical power.
The logic is that after dual engine flameout and without AC power and pitot heat, the air data is deemed unreliable, regardless of whether there is icing, and the flight control mode reverts to Secondary.
I also don’t see a reason to doubt Boeing’s simulations.
Perhaps @Andrew can weigh in, who I believe has specific knowledge.
@Sid Bennett
@Victor
The primary flight control system (PFCS) automatically reverts to secondary mode when all AC power is lost after the second engine failure. The reversion to secondary mode is not explicitly mentioned in the manuals, but that behaviour is well known by those who have witnessed dual engine failure events in the simulator. Indeed, the B777 dual engine failure checklist directs pilots to reset the primary flight computers to restore normal mode once AC power is restored.
Autopilot disconnection upon the loss of AC power is also mentioned in the ATSB report ‘MH370 – Definition of Underwater Search Areas’ (3 Dec 2015, p.9).
For some time it was thought that the mode reversion is caused by the loss of pitot heat, because the QRH prompt for the pilots to reset the primary flight computers is linked to the blanking of the HEAT PITOT L+C+R EICAS message. However, it turns out that the loss of pitot heat in itself does not cause the mode reversion. Boeing technical specialists confirmed that the mode reversion is caused by the loss of electrical power to a large number of position sensors and the TAT probe heat.
@Sid Bennett,
All you need to do is look on the plot at the bearing lines of 180 degrees (UGIB) and 186 degrees for the route you suggest.
At 0 Hz BFOR the expected latitude is 34S (like UGIB’s LEP) for both bearings (180 and 186). At +/- 7 Hz BFOR, the latitude range for both bearings is 29S to 40S.
@VictorI
I see this as a serious question that needs further thought.
The published literature on the 777 PFC architecture — specifically the Avionics Handbook Chapter 11 and ARINC 629 bus architecture papers — establishes two relevant points. First, the ADIRU is powered from the hot battery bus and is never interrupted at 2FO. Second, the ARINC 629 Flight Control Buses are passive media whose operation depends entirely on the terminal controllers within each connected LRU, not on any active bus infrastructure. The AIMS Data Conversion Gateway (DCG), which is AC-powered and would reset at 2FO, functions as a cross-bus relay for LRUs on different buses — affecting displays, FMS, and maintenance functions — but would not affect communication between two LRUs co-resident on the same Flight Control Bus.
The published identified trigger for Secondary mode is total loss of air data from the ADIRU and SAARU — not a transient interruption. In any case, if the ADIRU and PFC are co-resident on the Flight Control Bus, as the ARINC 629 architecture literature suggests, the DCG reset at 2FO would produce a display glitch visible to the simulator operator (on a different bus which needs the DCG operation) but would not constitute a total air data loss from the PFC’s perspective.
I am not aware of any published analysis — outside of MH370-specific discussion — that identifies the 2FO power transient as a Secondary mode trigger in a normally configured 777.
Specifically, for the Group 1 cases, was a Secondary mode transition observed from PFC status word output, or was it inferred from display behavior or trajectory characteristics? And was Secondary mode an initial condition in any of the runs rather than a consequence of the simulated 2FO event?
I raise this not to challenge the simulator results but because the answer has a direct bearing on whether the Group 1 trajectories correctly model and the PFC authority during the critical first 60-120 seconds after 2FO. If Normal mode was maintained throughout in the Group 1 cases — as the architecture suggests it should be — that would be an important point of agreement between the simulator and my analysis.
Again, I appreciate that this is a significant question and its significance might not have been appreciated in the early days.
@Paul Smithson,
Regarding UGIB you said: “Boxes were indeed ticked. But only by stretching credulity in several directions at once.”
Although you addressed your comments on the UGIB Route as a response to Victor’s statement, I will respond. I think those issues with UGIB are fair criticisms, except for your #5, which Victor and I addressed in our 2023 debris drift paper. I have previously raised all of them in previous papers and comments.
Monday-morning quarterbacking, six years after UGIB, might be fun for some people, but I look at the process differently. I say that when nothing is ventured, nothing is gained. There are a number of shortcomings in UGIB, some of which we realized at the time, and some of which have gained greater significance more recently, but nothing in UGIB was clearly ridiculous or impossible. It’s only now in hindsight, after we know from the seabed search that almost certainly UGIB wasn’t the correct route, that its flaws have become the rationale for its failure, rather than just being concerning issues.
It’s also useful to assess alternative routes which were published in detail by or before 2020 and which can now be demonstrated to be a still-viable candidate. I’m not aware of any candidate routes which are credible today, regardless of when they were published. In my opinion, they all have fatal flaws.
@DrB
The only flaw I discern in my model is that the 7th arc location has not resulted in find the wreckage. That is a separate issue.
@Sid Bennett: Nobody doubts that the ADIRU is powered by the DC bus, and the power is not interrupted by the dual engine fuel exhaustion. However, when data from sensors is deemed unreliable, the control mode reverts from Normal to Secondary.
If you read @Andrew’s comment above, he plainly states that Boeing technical specialists confirmed that the reversion to Secondary mode is caused by the loss of electrical power to a large number of position sensors and TAT probe heat, with the loss of pitot heat displayed as an EICAS message (which the PMDG simulation also demonstrates).
I recommend you re-examine your assumption that Normal mode is maintained with a dual engine failure.
@All,
With all the talk of flight controls degrading to secondary, I would like to revisit why I think this happened all the way back at IGARI.
There is little doubt that in the rupturing oxygen cylinder scenario there would damage the pitot/static feeds to the flight control computers. This explains why the autopilot appears ‘off’ as the aircraft turns around and flies over the Peninsula. The autothrottle will also disconnect in this damage scenario, essential for onward stable phugoid motion.
I believe the aircraft is stable enough, also laterally, whilst in secondary mode to meander on until fuel exhaustion with no one in control.
@Tim: Yet, in the history of aviation, we’ve never seen a plane fly for hours without the autopilot engaged or without some form of augmented control.
@Sid Bennett,
Your model shares a common flaw with most others, including DSTG and UGIB. It makes a simplifying assumption of no major turns after 19:41. That assumption is most likely wrong.
@Dr.B
It needs none. It does not need unnecessary turns to get to its location at 19:41 either, just one at 18:40. It is “marginal” on fuel. Fuel is either too little, too much or just right. Marginal = as close to just right as all the approximations take us. Exactly right sounds like a fit.
@Airlandsman
The power increase (or more precisely S/N increase) happened smoothly without any sudden jumps. All the mechanisms I know to give a power increase when the basic parameters predict a gradual decrease involve abrupt issues. That goes in particular for the 2-dB stepwise increase to maintain good signal quality. There was no need for any power increase, and no reason to expect any of the other abrupt mechanisms to occur, and the data show no sudden jumps (except at arc 7 where the system shifted to another channel, and many other things happened rapidly).
Therefore I can only see thermal reasons for the power increase.
@DrB
I do not agree that BFO becomes unpredictable if a thermal issue lies behind. Basically there are only 3 likely ways it could have happened:
1) Heating off
2) Open door (or similar)
3) Both open door and heating off
Number 3 fits nicely with a flight to somewhere north of 16S, but not for any flights south of there (too much cooling). I guess it might be possible to get a fit somewhat further south with number 2, and the rest of the way with number 1, but the BFO values would no longer fit with any of these possibilities.
That means a solution of the S/N issue together with a nice BFO fit can only be achieved with a flight ending north of 16S, or with an advanced active pilot scenario.
@All
There is one totally independent way to discriminate, but I will wait until after Easter to present it.
@Andrew, VictorI
If one were to encapsulate the difference between Airbus designs and Boeing FBW designs, it is that the former is flown like a fighter jet, with a force stick, and the latter is flown with a yoke like a DC-3. This is a difference in philosophy, but what goes on behind the curtain is very similar. It is the interface layer and the look and feel of the cockpit that is different. Navigation is very similar in modern designs.
If I understand the purpose of a Class D simulator correctly, it is intended to give the pilot being trained the feeling of being in the cockpit and having to perform the prescribed procedures to deal with contingencies. As such, in the case of a dual-engine failure, the simulator produces a dramatic and visible and tactile degradation of cockpit function: the control column feel goes slack as the AC-powered backdrive actuators lose their command source; the EICAS floods with alerts as the display system recovers from the AIMS (and DSG) reset; the autopilot disconnects as the AFDCs lose AC power; and the pitot and TAT probe heaters are shown as failed. The QRH dual engine failure checklist directs the crew to reset the Primary Flight Computers to restore normal cockpit functionality. That should be the experience the pilot experiences. But the PFC continues in the Normal mode (unsupervised by the autopilot).
Boeing 777 training documentation explicitly states that Secondary mode requires total loss of ADIRU and SAARU data — not loss of pitot heat, backdrive actuators, or any other AC-powered system. Absent triggering conditions the PFC operates in the “Normal” one of three control modes. The Normal mode provides the full C* pitch control law, active yaw damper, and command authority to all available hydraulic circuits. The conditions under which the PFC autonomously reverts to Secondary or Direct modes are specific and limited: Secondary mode is triggered solely by the PFC’s own assessment of the quality of the data it receives from the Air Data Modules (ADMs) on the Flight Control Bus. The PFC has no visibility into the state of the pitot heaters, the TAT probe heater, the backdrive actuators, the AFDC feel system, or the EICAS display system. None of those systems are inputs to the PFC’s mode decision logic.
The ADMs are powered from the DC bus and survive the 2FO AC bus loss without interruption. They continue to provide valid pressure data to the ADIRU, which continues to broadcast valid inertial and air data on the Flight Control Bus throughout the 2FO transient. The PFC, monitoring that data continuously, sees no degradation in data quality and has no basis to leave Normal mode.
Data is transmitted on the ARINC 629 bus is unaffected by the loss of one or more LRUs on a bus. The PFC, the ADIRU, and the ADMs are all on the Flight Control Bus (FCB). Since each one of these LRUs is on the hot battery bus nothing changes. The only possible degradation of the air data would be if the pitot tube ices up. The pitot tube heaters are powered from an AC bus and would be off during the short time gap between the 2FO and the RAT supplying essential AC power.
None of the cockpit degradation affects the PFC’s mode decision. The “reset the PFCs” checklist action does not restore the PFC from Secondary to Normal mode — the PFC has remained in Normal mode throughout. The action restores the autopilot interface to the PFC: the AFDCs, having recovered on APU power restoration, require a deliberate crew action via the MCP engage switch to re-engage the autopilot. In an unpiloted aircraft that action is never taken. The autopilot therefore does not re-engage following APU power restoration — consistent with the analysis presented in §4.2 — but the PFC has been executing the C* control law with valid ADIRU inputs continuously from 2FO to impact.
Secondary mode cannot be commanded from the cockpit. There is no switch or crew action that places the PFC directly into Secondary mode. It is exclusively an autonomous PFC decision based on ADM data quality. Since that data quality is never compromised in the 2FO scenario, Normal mode is maintained throughout the post-2FO descent. The Boeing 777 Primary Flight Computer did precisely what it was designed to do.
So, I can easily accept your and Andrew’s report of the experience with the Class D simulator as to the manifestations of a two-engine failure, but the actual effect on the navigation of the plane does not seem consistent with the design. Perhaps, the question is whether the initial conditions used for the simulation resulted in the PFC being placed in the Secondary mode without the actual preconditions for doing so being satisfied.
@DrB
I am not big supporter of 23s but perhaps starts to be consistent with the Curtin sound if you allow for some distance after Arc7. I would say as soon as you have turns after Arc5 you may have descent too.
If, instead of wasting extra distance/fuel on a right hand turn after Arc5 to conform to the assumption of fuel exhaustion at Arc6/7, you can descend, go slanted path to Arc7, now you have fuel remains at Arc7 due to less distance flown and descent fuel savings (no need for two engines BTW) Which is what I think we probably actually witnessed.
Another way I look at it, I appreciate all the open-mindedness and creativity between Arc1 and Arc2, for example I find UGIB useful. But we have been more constrained after Arc5 which is also an area where more open thought and creativity may be needed.
@Sid Bennett said: Boeing 777 training documentation explicitly states that Secondary mode requires total loss of ADIRU and SAARU data.
The 777 Training Manual that I have says: The PFCS selects the secondary mode when it detects the loss of important air and attitude sensor data from both the ADIRU and the SAARU. The PFCS also selects the secondary mode when two ACEs are in direct mode.
In addition, the FCOM states: When the PFCs can no longer support the normal mode due to internal faults or lack of required information from other airplane systems, they automatically revert to the secondary mode.
The statements I have provided from the Training Manual and FCOM are consistent. However, they differ the statement you provided which says the Secondary mode requires total loss of ADIRU and SAARU data.
What is your source, and what are the exact words?
Also, as @Andrew said, the QRH instructions for dual engine failure implicitly use the “HEAT PITOT L+C+R” as a proxy for reversion to Secondary mode due to loss of AC power. Therefore, the instruction is to disconnect and reconnect the PFCs when the pitot heat message appears and disappears. Again, this is consistent with the FCOM and Training Manual excerpts I have provided.
Finally, if you are claiming that the behavior of the Level D simulator is not consistent with the design of the aircraft, AND the Boeing specialist was incorrect in stating the control mode reverts to Secondary due to the loss of TAT heat and position sensors, it would be helpful to understand why you believe you have more accurate information.
@Sid Bennett
I am quite familiar with Boeing and Airbus FBW designs, having amassed over 10,000 hours operating wide-bodied FBW aircraft made by both manufacturers. I was also involved with training and checking B777 pilots for a number of years.
The core issue seems to be your claim that “Boeing 777 training documentation explicitly states that Secondary mode requires total loss of ADIRU and SAARU data — not loss of pitot heat, backdrive actuators, or any other AC-powered system.“
As Victor already highlighted, the manuals do not say that a TOTAL loss of ADIRU and SAARU data is required for a reversion to secondary mode, nor do they say that ONLY a loss of ADIRU and SAARU data will cause such a reversion. The loss of any parameter of critical air data or inertial data will certainly cause a reversion to secondary mode. However, the PFCS also relies on inputs from a number of other aircraft systems, all filtered through the L & R AIMS cabinets. As Boeing tech specialists confirmed, much of that data is lost or unreliable if all AC power is lost, which will cause a reversion to secondary mode.
Do you really believe that Boeing specialists don’t know their own aircraft and that this aspect of B777 pilot training has been conducted incorrectly for the last 30 years?
Many of your other claims are also wrong, but they are largely irrelevant to the core issue. I have responded to them here: Sid’s Furphies
@Andrew: “Furphy” was a new one for many of us.
@Victor
Always happy to add something new to the lexicon! 😎
@Viking: Re: “…Therefore I can only see thermal reasons for the power increase.” The GES CU receive power has nothing to do with AES “thermal reasons”. The AES HPA power output is controlled by the SDU and maintained to produce the required EIRP, regardless of the ambient temperature. There is absolutely nothing in the Inmarsat data to indicate the cabin temperature.
@Andrew
@Victor Iannello
World War One era water-tank manufacturers aside, the reversion to secondary mode with the loss of power consequent to fuel exhaustion is uncontroversial. It does raise the issue though, of where the SDU is pulling air data from for Doppler compensation when it fires up?
By the time the SDU is re-powered, presumably when the APU auto-starts, the aircraft’s electrical systems have gone through a number of configuration changes – first, when the right engine quits; then when the left engine fails; presumably again when the RAT deploys; and then again when the APU fires up. What have those changes done to the source(s) and quality of the air data that the SDU is going to use for the log-on?
@Mick Gilbert
The SDU receives GPS-augmented inertial data from the ADIRU via AIMS. The ADIRU and AIMS remain powered throughout the electrical configuration changes you mentioned. MMR R (which provides GPS R data) also remains powered, so the quality of the data needed for Doppler compensation should not be affected.
Others here are more knowledgeable about the SDU Doppler compensation and may have more information.
The SDU receives 7 IRS items:
Latitude
Longitude
Ground speed
Track angle
True heading
Pitch
Roll
The ADIRU can supply all of the above items. The MMR can supply the first 4. From what I can tell, the GPS data are used by the ADIRU only for initialization, not normal in-flight operation, and the IRS data sent to the SDU may be sourced entirely by the ADIRU’s internal air and navigation functions alone.
The Doppler compensation calculation makes use of latitude, longitude, ground speed, and track angle. The calculation is particularly sensitive to any error in the track angle. I estimate that a track angle error of -28 degrees at the end of flight would cause the BFO to change from 255 to 0 Hz. (As an aside, spoofing the track angle is the easiest way to make the plane appear to go S when it was actually going N.) While it seems unlikely that such errors could be introduced, the possibility should at least be noted.
@sk999
The B777 Maintenance Training Manual (34-20-00, p.38) says that GPS data is used by the ADIRU for initialisation and calibration of the gyros and accelerometers. The manual states:
“The GPS position and velocity goes to the sensor calibration logic. The sensor calibration logic supplies calibration data to the gyros and accelerometers. The calibration does not correct the ADIRU errors, it only reduces ADIRU drift errors as the airplane flies.”
Andrew,
Thanks for that clarification. I was thinking of newer IRS units where GPS and inertial data are “blended” to produce much more tightly coupled outputs.
@sk999: On the B777, the blending occurs in the AIMS, and the blended position is used by the FMS for navigation. It’s not clear whether the ADIRU, MMR, or AIMS position is used by the SDU for pre-compensation of Doppler shift, nor does it matter unless there is a failure.
Victor,
Yes, I saw that the FMC utilizes both GPS and ADIRU data for navigation but couldn’t find anywhere that it was sent other than to the displays. What I was most interested in was the source of the ground track data, and I’m pretty sure everything is derived from ADIRU data on the ARINC 629 Flight Controls bus. So the question then becomes what is the accuracy (both random and systematic) of said data.
@sk999: It is VERY unlikely that the displayed data is different than the data used for navigation. That would be extremely confusing to a pilot.
The position data from the MH370 ACARS position reports is laterally offset from the Terengganu ADSB data, which has essentially no error. Likely, the ACARS position data is from the ADIRU, so the accuracy of the ADIRU data can be gleaned. Since the data is available only at the start of the flight, we don’t whether the ADIRU further drifts.
Previously, I looked at the MH371 data to see if the position reports from the ACARS data had the plane offset from known airways. At one point, there was a significant lateral offset, but I attributed that to an ATC instruction to fly with an offset, as later in the flight, the ACARS position reports had the plane directly on the airway.
@Victor
My money is on the SDU using ADIRU data, sent via the flight controls ARINC 629 buses to the DCGF within AIMS, then out via an ARINC 429 bus to the SDU. The SATCOM chapter in the Maintenance Training Manual states (23-15-00, p.6): “The DCGF changes inertial reference system data into ARINC 429 data for high gain antenna (HGA) beam steering.” The SATCOM interface diagram on p.41 of the same chapter shows that both AIMS cabinets send IRS data to the SDU.
If the SDU used GPS data from the MMRs for Doppler pre-compensation, then inoperative GPS would presumably cause a downgrade in SATCOM function. The MEL, however, shows the aircraft may be dispatched with both GPS inoperative and no concurrent downgrade of the SATCOM.
I haven’t found evidence of the AIMS FMCF sending blended ADIRU/GPS data to the SDU.
Happy to be convinced otherwise!
FWIW
@All,
Here is a link to a new paper describing a candidate route ending at 20S in the unsearched area:
https://drive.google.com/file/d/1FnueqTwbXFlJ0aj0lsG3srXMO5X7UNhf/view?usp=sharing
It employs waypoints only and has major turns. It leads to one specific location, not a range in crash latitudes.
@Andrew: I doubt pure MMR data is used for anything, as it would be dependent on the availability of reliable GNSS data, which we know is not always the case.
[Other than operations that specifically call for GNSS and/or navaid data, such as an RNAV and ILS approaches.]
@All: Happy Easter!
Happy Easter!
@sk999 (April 4, 2026 at 12:00 pm)
You said:
“I estimate that a track angle error of -28 degrees at the end of flight would cause the BFO to change from 255 to 0 Hz.”
I am wondering. If it took 8 seconds to go from 255Hz to 0Hz, and if the track angle DID actually change by 28/30 degrees in that 8 seconds, (and it was not diving as has always been said) then the question becomes, what maneouver could explain that ?
My first thought was when breaking out of low cloud base, a pull-out from the descent, then a savage boot full of rudder in the last seconds prior to ditching to kick off drift to align with a swell crest.
@sk999:
Re: “I estimate that a track angle error of -28 degrees at the end of flight would cause the BFO to change from 255 to 0 Hz. ”
I guess I don’t understand what you mean by “track angle error of -28 degrees”. Please explain. The BFO value changes relatively little in horizontal flight as you make a 360 degree circle. See: https://bit.ly/4kkLy2z
As those charts show, the BFO value only changes by +/- 10 Hz during a 360 degree turn.
@airlandseaman: @sk999 is referring to a hypothetical deviation between the actual track and the track used in the Doppler pre-compensation.
@DrB
Very interesting paper, of which I confess to understand just a little.
Are you confident that this solution to a flight path that matches all known data is unique, and is not just one fit of a new group of paths?
In regard to the ability of the flight path to have been preset at or before 19:18 UTC the path appears, at least superficially to me, to convoluted perhaps indicating indecision on the part of a live PIC.
I’m looking forward to a robust discussion from others in the group.
@John,
You said: “Are you confident that this solution to a flight path that matches all known data is unique, and is not just one fit of a new group of paths?’
I don’t know if another route using waypoints-only navigation exists which matches the satellite data comparably well. In my experience this route is in a different class altogether, and so far I have been unable to identify any comparable alternative at ANY crash latitude. It turns out that requiring waypoints be used is a powerful route discriminator in the SIO because of the scarcity of waypoints there. Not finding another comparably good route strengthens the case for searching there. Finding several comparable alternatives would weaken that case. We’ll see if I or anyone else can demonstrate a second example with fully compliant BTO and BFO residuals which are not “tuned” by fitting arbitrary non-integer bearings or adjusting ROC/ROD at Arcs 2-6.
You also said: “In regard to the ability of the flight path to have been preset at or before 19:18 UTC the path appears, at least superficially to me, to convoluted perhaps indicating indecision on the part of a live PIC.’
Maybe so. My interpretation is that the route was pre-planned by ZS for execution during MH150. That didn’t happen. For some unknown reason, he decided to do it on MH370. The only major adjustment to the Simulator Route was in the destination which had to be farther north than 45S because of the reduced fuel. In my opinion, the final turn at IKASA to EGAVI may have been to put the fuel exhaustion location in the deeper basin to the ENE of IKASA. Similarly, in my opinion the zig-zags were pre-planned and intended to obfuscate the crash location in the event the aircraft was detected at any point along the SIO Route. After all, wasn’t that the purpose of flying to the NW on N571 until out of radar range, and then doing a 180 to the SE, to establish a false trail? There are half a dozen major turns altogether, which may have been used to prevent the crash location from being predicted.
You said: “I’m looking forward to a robust discussion from others in the group.”
Me too. Comments, criticisms, and suggestions are welcomed. I understand this is quite a change from the routes we have been trying to refine over the past 12 years. That approach didn’t work out, so a re-think is needed. I did that from November to now, and this is what I came up with. It’s interesting in hindsight to try to figure out why it took this long, because the basic information needed has been there all along, except for the motivation provided by the negative seabed search results. On my part, 12 years ago I knew almost nothing about how a B777 was controlled in flight, so I had a great dela to learn. Thanks to people like Andrew and Victor, I slowly became more knowledgeable about how to model the flight path. Recently I was especially focused on understanding why OI seemed to give up suddenly, because they are not known to be prone to making rash judgments. After much thought, what they did now seems quite reasonable to me. Understanding why OI decided not to spend more money searching circa 34S is not the same, of course, as figuring out where to look “next”. OI’s cessation motivated two activities: (a) an intense review of the satellite data, especially the 18:25 – 18:28 BTOs and BFOs, and also the Arc 7 BTOs and (b) thinking “out of the box” by relaxing the assumptions, both recognized and hidden, which constrained the unsuccessful predictions.
I have posted a revised version on my paper at the same link:
https://drive.google.com/file/d/1FnueqTwbXFlJ0aj0lsG3srXMO5X7UNhf/view?usp=sharing
It contains new material on debris drift predictions and ATSB’s NW point, fixes a few small errors, and presents the detailed route information in a more readable format.
@DrB:
When I proposed a waypoint path to the Java anomaly site years ago, it was called “contrived”, presumably because I added one additional waypoint near the end to hold a constant altitude. BFO was considered somewhat unreliable at the time. You are now forcing BFO matches onto so many waypoints that it has turned into a meandering path. This seems even more unlikely than the previous UGIB proposal where an evasive pilot made right angle turns to match FIR boundaries.
IIRC, CAPTIO proposed a BFO and BTO compliant path ending at around 12S just beyond the 7th Arc. Their path was far simpler than yours, and had the same waypoint at Cocos Island airport as mine, though my flyby then headed to Christmas Island airport. A complex meandering waypoint path can’t be unique when we know simple ones will fit.
If there is enough time for the plane to zig-zag around yet meet BTO arcs, then it is probably flying too fast. You still have the plane at FL370 in cruise, when an oxygen altitude flight near holding speed fits just fine (regardless of pilot intent).
Optimizing paths for a tight BFO match is not the same as optimizing for BTO. The BFO is sensitive to instantaneous fluctuations. You appear to be ruling out a momentary ascent at Arc 6 just before fuel exhaustion. Consider that even moderate turbulence can cause excursions beyond 1000 fpm.
You are optimizing a path using narrow sigma tolerances where you dismiss anything outside that as impossible. This was the same strategy used in previous papers, which later needed to relax constraints on glide etc by expanding the search zone. I suggest it may be better to combine all of your optimizations into proposed path with final latitude, and then allow those conducting the search to decide how accurate it is. I’d like to further point out that you are dealing in probabilities using very small sample sizes, trying to find the most likely place of impact. In reality, the plane is unlikely to be right at the peak of your curve, just as any one coin flip is an outlier.
@All:
I suggest that there are more assumptions that might be set aside. If MH370 continued following waypoints instead of an being an unpiloted flight to oblivion, that pretty much removes the basis for most hypoxia (innocent crew) or malicious pilot scenarios. Speculative intentions to evade radar or hide the plane in deep holes make little sense.
A supersonic descent from high altitude with multiple control surfaces detaching from the plane before impact is still controversial when other experts have shown that the trailing edge damage without the leading edge being crushed is consistent with water impact. There must have been a pilot to continue flying waypoints, and flaps being retracted does not preclude a failed ditching. AF447 entered the water in a nose-up stall with similar damage. If a pilot had no flaps because the RAT doesn’t power them, he might attempt to stall and drag the tail in the water. We do know the fuselage was shattered from cabin debris, but that has happened to airliners at very low speeds off the end of a runway.
With proposals now narrowing on the unsearched 7th Arc N of 25S, there is other new evidence to consider. I am not alone in showing that the growth of the flaperon barnacles happened in a matter of weeks, not 15 months. I can’t accept it as evidence for conspiracy claims that all the debris was planted. Earlier La Reunion arrival mainly affects MH370 drift modeling by favoring sites N of 25S, but multiple analyses also reveal the temperature profile of barnacle growth.
It has always been a paradox that growth began at the warm end of the lepas anatifera reproductive range at around 25C, with the surrounding waters cooling then warming again before discovery. Another conundrum is the lack of barnacle growth on all but a few pieces of found MH370 debris. An appendix to the ATSB Final Report said gooseneck barnacles found on the closet fragment were likely 45-50 days old. Snails and algae were found on several other pieces with no signs of cold water species. A plausible explanation is that most of the debris never left tropical waters above 25C.
Using the same CSIRO drift model particle data as UGIB, I tracked them all to see which never went into waters cooler than 25C. The starting crash latitude threshold is around 21S. That is very consistent with a search between 8S and 25S, and might be used to narrow it for general optimization.
I of course invite anyone to try matching a fully optimized BFO plus fuel exhaustion path to the 8.32S 7th Arc Java anomaly epicenter site, that also crosses the two island airports where doppler infrasound noises were detected at the correct flyby times.
@370Location
Ed, when you say “other experts have shown that the trailing edge damage without the leading edge being crushed is consistent with water impact“, who did you have in mind by way of “experts”, and by what mechanism has trailing edge damage without the leading edge being crushed being consistent with a water impact been demonstrated?
@DrB
Below link is my 2021 essay “New Interpretations of the Pilot’s Home Flight Sim Studies” which was based on the new ATSB guidance at the time.
It does not appear that MH370 followed N571 (or waypoints). The data suggests the aircraft started on B466 and a set heading for VAMPI before TASEK (before getting into Indonesian airspace). At approx VAMPI the icon is dragged to N10 where it headed south. We have no flight path in the complete data, and I strongly suggest it may represent 180 CMH with Fair Weather winds (which was Victor’s idea years ago as an alternate explanation).
https://docs.google.com/document/d/1cpaWvW861hpKoy8IHK3NACuw-NyeW9qIIQ6S4Fd8GOA/edit?usp=sharing
Keep in mind, the original 2016 RMP sim data leak was incomplete compared to the “complete” data secretly available to ATSB. The complete sim data is still secret, but I believe with ATSB guidance a few of us now understand most of the dataset. There are mixed opinions, with some feeling the sim data is useless. I disagree, I say it is probably the actual flight path after a straight flight ~ISBIX down to Arc5. At approx Arc5, MH370 merges onto the 180S flight path.
@John said: I’m looking forward to a robust discussion from others in the group.
The measured BTO values at the handshake times from 19:41z through 24:11z can be accurately fit with a 3rd order polynomial with a single local minimum. Preferred route reconstructions would have BTO values between handshake times that would also match that polynomial. Otherwise, a more complicated BTO curve would be required to coincidentally match the simpler BTO curve at the handshake times.
That doesn’t mean a route following waypoints that zig-zags didn’t occur. I would, however, classify it as unlikely. I would say that other routes with maneuvers that produce BTO curves that match the 3rd order polynomial, whether or not those routes follow waypoints, would be more likely. Unfortunately, it becomes difficult to define a unique route.
One possible recommendation is to continue the subsea search starting north of 24.8S and continuing until past Ed’s hotspot at 8.4S latitude.
@TBill,
One could infer that the Simulator Roue probably followed N571 because between 5N and 10N it passes by VAMPI (which is on N571) and at 10N it is between LAGOG (which is also on N571) and DOTEN. However, you are correct that it’s just those two waypoints and we don’t have any additional information on the exact path between 5N and 10N. The MH150 flight plan follows N571 from VAMPI to LAGOG before turning to DOTEN on N877, which is why that flight has been associated with the Simulator Route.
@Victor Iannello,
You said: “I would say that other routes with maneuvers that produce BTO curves that match the 3rd order polynomial, whether or not those routes follow waypoints, would be more likely.”
I wonder if any airline pilots would agree with you that B777 flight paths are more likely to match 3rd order polynomials in BTO instead of waypoints.
I would suggest because this flight is unique in the annals of civil aviation, we should not expect or assume or require normality and simplicity. We already know it was quite abnormal in its complexity from 17:24 through 19:41. Assuming simplicity and allowing poor or artificial BFO matches has led to unsuccessful searches. That train has left the station without us. It’s time to find a new ride.
@370Location,
Thank you for commenting on my Candidate Route. It’s always helpful to hear other opinions.
You said: “You are now forcing BFO matches onto so many waypoints that it has turned into a meandering path. This seems even more unlikely than the previous UGIB proposal where an evasive pilot made right angle turns to match FIR boundaries.”
Yes, the True Route must match ALL the BFOs. Period. That is indeed a hard and fast requirement. Any route which doesn’t is incorrect, at least in part, and therefore the predicted crash location is also likely to be erroneous, especially if the BFO errors occur later in the flight. The FIR boundary tracks in UGIB were not at times when we had BFO/BTO data, so they were only a means to create delay, not to match either BTOs or BFOs. One could just as well employ a Hold.
You said: “CAPTIO proposed a BFO and BTO compliant path ending at around 12S just beyond the 7th Arc. Their path was far simpler than yours, and had the same waypoint at Cocos Island airport as mine, though my flyby then headed to Christmas Island airport. A complex meandering waypoint path can’t be unique when we know simple ones will fit.”
First, you don’t know simple ones will fit.
Second, you don’t know that a meandering waypoint route can’t be unique. I think it may be.
If you or anyone else think you have a route which does fit properly, I invite you to provide the specifics for it using the same information as contained in my BTO and BFO tables (now called Table 1 and Table 2). Then I and other readers can compute the BTOs and BFOs for the whole route and we’ll see whether or not it is flyable, has MEFE at the right time, and is a viable candidate.
Nor do you know that a simpler path is more likely to be correct. That is an example of the sort of thinking (mine, too, in the past) that has been unsuccessful to date.
You said: “I suggest it may be better to combine all of your optimizations into proposed path with final latitude, and then allow those conducting the search to decide how accurate it is.”
I have already done that in my paper. I only predict one location, at 19.72S 104.62E.
You said: “If MH370 continued following waypoints instead of an being an unpiloted flight to oblivion, that pretty much removes the basis for most hypoxia (innocent crew) or malicious pilot scenarios. Speculative intentions to evade radar or hide the plane in deep holes make little sense.”
I disagree. I am not aware of any evidence which is inconsistent with a malicious pilot scenario. Please provide this if you have it. As I wrote, all the inputs to complete my Candidate Route can be completed circa 19:18, so you don’t need pilot inputs after that, although that could have occurred.
Fortunately, we don’t need to consider the scenario or motive in determining the route followed. We have the satellite data, and AFAIK there is only one route which properly matches all of it without climb/descent contrivances to match BFOs. All are welcome to demonstrate a second example, whether the route is waypoint-driven or not, and no matter the Arc 7 latitude.
@DrB:
Independent of pilot intent, if sparse data can be fit to a low-order polynomial, it’s likely that between the data points, there is also a match. The BTO fit to a low-order polynomial does not exist prior to 19:41, so I think the comparison to that part of the flight is irrelevant.
However, if you want to look at how a pilot might fly to a location, if there was an attempt to not reveal the location, a pilot wouldn’t follow waypoints, which would in itself be revealing. Rather, with the destination entered in the flight plan, a pilot might periodically select new headings, always aware of the required course to the destination from the present position and the cross-track deviation from the most direct path, much the same way a pilot deviates around bad weather.
@Victor
” … cross-track deviation from the most direct path, much the same way a pilot deviates around bad weather.”
Another possibility is that any weather is probably a significant risk to the pilot.
He would want to remain visual, to be able to see any airliner strobes in good time.
What might look like a random meandering path may be avoiding clouds to remain visual.
Comparing candidate paths to the satellite pictures and cruise level met may be useful.
@Mick Gilbert:
Thank you for investigating precedents for ditching damage with a focus on the vertical stabilizer. That fragment of MH370 found could have been due to impact from another piece of debris, or even cartwheeling of the plane.
Only the French DGA physically examined stress deformations on the flaperon and concluded, “… taking into account the results of the examinations, it appears that the flaperon impacted the water while still attached to the aeroplane and that at the time of the impact it was deflected.”
They also began that section of the report with, “First of all, it appears possible to exclude in-flight loss of the flaperon since its weight is concentrated forwards, which would a priori lead to a fall with the leading edge forwards and the probable destruction of the latter. The damage to the trailing edge would also likely be different. A simulation of a flaperon fall with an initial speed corresponding to that of an aeroplane in flight would enable this to be definitively eliminated.”
Dr Argiris Kamoulakos with the CAPTIO team did just that. He conducted FEM modeling of water impact damage to a detached falling flaperon compared to water impact on the trailing edge of the flaperon, with animations. He showed that the flaperon would fall leading edge first and show compression damage there.
Seasoned air crash investigators Larry Vance and John Cox independently concluded that the flaperon trailing edge damage was due to water impact.
All of the above made the assumption that due to the damage being from water impact on the trailing edge, that the flaps were deployed. It seems clear from witness marks that they were retracted. None mentioned any consideration of a flaps-up, nose-up stall like AF447 that would reconcile the issue.
Simon Hardy also supported a controlled ditching. The ATSB said it did not support a *successful* controlled ditching, though Pete Foley was quoted by the BBC via 60 Minutes Oz as being open to a controlled glide and ditching.
There are few clear photos of AF447 control surface damage. One underwater photo of the left wing ailerons shows them completely sheared off leaving only struts and the leading edge intact within the wing.
The flaps and flaperon were impacted by other debris, but had no crushing damage to the leading edge (which would have been mostly protected inside the wing during a ditching). It seems very unlikely that the flaps and flaperon acquired the same trailing edge damage, then spontaneously left the wing, and somehow collided on the way down or hit debris on the ocean surface (as the body of the plane would continue forward).
I think this is sufficient to show that there is at least some controversy about whether high speed flutter or a ditching attempt caused the trailing edge damage. The case for high speed flutter appeared here a day after the flaperon was found, based on the final BFO value indicating dive up to 23,000 fpm just 8 seconds after indicating level flight.
Would it even be possible for a B777 with engines out to achieve trans-sonic flutter in a dive from 15K ft? If so, how could the leading edges of the control surfaces not be damaged on impact with the water?
@Victor, @DrB:
I agree that a waypoint path doesn’t fit with a malicious pilot trying to evade detection and make the plane disappear. Pilot Juanda Ismail in a recent Youtube interview laid out other reasons not to blame Zahari. As others have, he points out that the best way to hide the plane would be to continue flying east at IGARI, away from radar centers. He attributes the manual steep turn there to the plane being kicked out of autopilot due to an electrical malfunction that also took out comms. I would add that a pilot intent on evading detection would not fly right past Butterworth AFB where jets might be scrambled. You may not consider this “evidence” against a malicious pilot, but it is sensible.
I speculate that the crew was attempting to deal with multiple problems while heading for potential landing sites, then avoided populated areas, with the major turn south intended to buy time while still attempting repairs and later head toward daylight and ditching or landing at the Java coast.
I’ve made the case that the BFO still is not as reliable as the BTO arcs, which I think are well established. A report on my website shows that even a simpler 2nd order polynomial fits the BTO values well. I believe that curve is in some part the basis for the past assumption of an unpiloted steady path south. However the first arc BTO is not irrelevant. The curve implies a somewhat linear path with a closest approach tangent to the satellite around the 2nd Arc at minimum of the curve. We do know for a fact that the plane had been flying NW at the 1st Arc, not south, yet the point still lands on the curve. It is relevant to show that later points after Cocos Island may be to the east and NE yet still land right on the curve. The curve gives the illusion of a smooth straight path. I do agree that it implies there isn’t likely any meandering or holding patterns for delays.
If strict adherence to BFO and waypoints generates complex random paths without any other evidence, then I suggest one of them is wrong. If the BFO could be wrong about a supersonic unpowered dive, then it may be inaccurate for determining the correct path.
@370Location: Without new evidence, it’s as pointless for you to persuade me that the evidence suggests malfunction caused the disappearance as it is for me to persuade you that the evidence suggests a deliberate diversion on the part of the pilot. That’s why I prefer to reconstruct paths without attributing an intention, which many might argue is impossible.
At this point, I think the best option is to progressively proceed up the 7th arc starting where the search ended at 24.8S latitude. Then, we don’t have to guess about intentions, navigational modes, flight inputs, fuel models, barnacle sizes, or anything else that people might use to suggest one area might have a higher probability than another.
@Victor:
I realize that the point counterpoint may seem pointless and repetitive. I should not have combined my response to you with the questions Bobby posed.
Many thanks for providing this forum, and especially for highlighting my candidate that can be searched quickly and soon.
I fully agree that searching near the 7th Arc north of 25S is the way to go. If OI is willing but time is limited, I would hope they search from north to south.
@370Location
“I speculate that the crew was attempting to deal with multiple problems while heading for potential landing sites, then avoided populated areas, with the major turn south intended to buy time while still attempting repairs and later head toward daylight and ditching or landing at the Java coast.”
As Victor has already said, that scenario is not how pilots deal with serious failures. So-called “potential landing sites” at Christmas Island and Cocos Islands don’t offer any advantage over Penang or Kuala Lumpur. In fact, they are far inferior airports, with shorter runways, very limited facilities and no rescue/fire fighting services. In addition, the crew of MH370 very likely had no information on the weather conditions at either Christmas Island or Cocos.
If the crew needed to deal with a serious in-flight emergency but needed time to prepare for a landing, they would have loitered near an airport such as Penang or Kuala Lumpur, not continued towards remote Indian Ocean airports. From a pilot’s perspective, it makes no sense that they would do so.
Hello everybody. I am a Chinese. I recenty notice a video about mh370 by a medium. He said that he can communicate with Zaharie and mh370 was ended in the sea area of Madagascar cross 7th arc.
https://www.youtube.com/watch?v=17javdd2SBQ&t=1s
What do you think? Is he a liar?
@Cui Shineng: Yes, he is a liar.
@DrB
I agree the Simulator Route starts out as if MH150 to Jeddah is being studied in that case, however that is analogous to the actual MH370 flight using normal waypoints to IGARI. The open question of course is route strategy after the point of diversion from normal.
@Mick @Ed
Also regarding the Flaperon damage, Boeing representative inspected the flaperon (requested by France) and judged it was probably water damage on the trailing edge. The Boeing rep was attempting to argue for a piloted water landing of some form, and of course faced heavy criticism from the Official Narrative believers of uncontrolled nose-dive-on-Arc7. Not to mention Voldemort (he who must not be named) who I always speculated might have contacts in Boeing, of course I refer to Larry Vance.
@TBill,
Those who argue that the flaperon damage along the TE was the result of water damage seem to ignore that immediately in front of the flaperon is this bloody huge engine.
What do they imagine happens when the engine hits the water?
@370Location,
You said: “I would add that a pilot intent on evading detection would not fly right past Butterworth AFB where jets might be scrambled. You may not consider this “evidence” against a malicious pilot, but it is sensible.”
I disagree. It is quite sensible for a malicious pilot to lay out a false trail by flying near the Butterworth radar, continue flying to the NW until out of radar range, and then making a 180 degree turn to the SIO heading SE.
Again, I say there is no evidence which is inconsistent with a malicious pilot scenario, if you allow the pilot to be intelligent and to have multiple strategies for obscuring the crash location.
You also said: “If strict adherence to BFO and waypoints generates complex random paths without any other evidence, then I suggest one of them is wrong. If the BFO could be wrong about a supersonic unpowered dive, then it may be inaccurate for determining the correct path.”
You are the only person I have heard speculate the descent after MEFE was supersonic. It does not have to reach Mach 1 for flaps and flaperons to depart the wing. The descent rates implied by the BFOs are not at supersonic speeds.
No one has demonstrated that “strict adherence to BFO and waypoints generates complex random paths”. I have demonstrated that it produces ONE path which is certainly not random. It is just the opposite – it is fully deterministic. I also think this path is unique and therefore singular. It is certainly the opposite of random. Maybe you meant “unexpected” rather than random. Maybe my proposed route is “unexpected” to you, but it is certainly not random, and there may not be more than one.
@Andrew:
I appreciate your perspective as an experienced pilot. I recall several instances where planes reporting Pan-Pan have gone into a long holding pattern near a major airport before successful landing. The key differences for MH370 appear to be the complete loss of comms (creating a hazard for a holding pattern) and night operation when most of the diversion airports were closed and dark.
As Victor noted, focus on the intent of the pilot isn’t very helpful. My candidate site is instead founded on acoustic evidence. I was responding with speculation to Dr Ulich inviting “any evidence which is inconsistent with a malicious pilot scenario”.
My forte is in tweaking algorithms that elucidate signal from noise. That may not be convincing in the real world logistics of optimally getting from point A to B, but there is a consistency in the underlying data.
I’m currently looking at using NMF (Non-negative Matrix Factorization) methods on seismic data to isolate the unique infrasound patterns of a flyby at Cocos Island from the repetitive waves crashing on nearby beaches.
@DrB:
My Java site is not dependent upon BFO. Non-adherence to strict BFO can only be used to insert doubt for preventing a quick search before the OI contract runs out.
When the MH370 crash site is eventually found, the intentions of the pilot should become clear.
@370Location
I’m not disputing your candidate site. My beef is with speculation that the crew ended up near some Indian Ocean airfield because they were looking for somewhere to land. That’s not how it works in the real world, lost comms or not.
@Andrew:
Thanks for that. I will follow my own advice and try to stick with facts, not speculation.
May I recommand you read the work done by my friend
Dr Argiris Kamoulakos, retired Scientific Director of ESI Group, 40 years industrial experience in Numerical Analysis and Simulation in Aerospace
https://www.linkedin.com/in/argiriskamoulakos/
MH370: Impact with the sea (ditching)
https://comp-as-s.com/mh370/impact-sea-ditching/
@Jean-Marc Garot
The report refers to Rib 25 and states that it “appears to be in the vicinity between the inner part of the outer flap and the right engine support pylon, close to the flaperon.” Rib 25 is actually located further out along the wing, about one-third of the way along the outer flap. The position of Rib 25 is marked on the following drawing of the left wing (right wing is the same): Rib 25 Position
That puts the failure point that was demonstrated during fatigue testing further out along the wing than suggested.
@Jean-Marc Garot,
Unfortunately Dr Argiris Kamoulakos ignored that the flap and flaperon were in the retracted position. This is a hard fact and cannot be dismissed. Therefore his ditching dynamics won’t work.
Victor, you say: “At this point, I think the best option is to progressively proceed up the 7th arc starting where the search ended at 24.8S latitude.”
In our previous discussions, I was under the impression that you regarded my 39.6S proposal as a next-best candidate. Perhaps I misunderstood. But, if not, what has changed?
I don’t want to belabour the point, but I believe that I have put forward plausible and empirical means by which this path can be reconciled with fuel, BFO and drift. Which of those do you think is not satisfactorily addressed?
@everyone
There is ib=n this blog recurrent speculation that a PIC orchestrated the demise of a planeload of passengers and crew, flew the aeroplane to oblivion so it would never be found, and then attempted a controlled ditching – but whatever for? The PIC had no escape, and by design no one was going to find them or the plane, so why the intricacies of deception?
There seems to me to be little point in trying to apply logic to interpolate illogical behaviour. Better to stick to objective known facts.
@paul smithson: If I had to limit the search to very small areas, I would probably recommend your candidate site near 39.6S and Ed’s near Java. However, if I had to make a search recommendation, I would progressively continue up the arc from where OI last searched, which I think maximizes the chances for success. I still question whether your candidate site is consistent with drift and fuel models (not withstanding your strong disagreement).
These are preliminary thoughts, subject to change.
@Brian Anderson
Good question. I try to be open on crash dynamics. Malaysia excluded NTSB/FBI/Boeing from the analysis of debris. Except for France, nobody asked Boeing for forensic analysis. Malaysia’s main goal was to ID and rank if the debris came from MH370. I try to focus on flight path/crash location, and I do strongly feel we witnessed a major descent and slow down before and after Arc6, and probably fuel remains after Arc7. That means I believe flaps are possible, but I have a hard time believing a Sully-type slow ditch scenario. I more aligned with CAPTIO/CAPTION fast-ditch crash ideas, which includes a number of possible scenarios including some mention of inverted landing into the water. I have some other ideas (Holding pattern) which could have the flaperon upwards. I do not know, if the flaps were all retracted, if a fast ditch could still erode the trailing edge. Ethopian 961 semi-fatal fast ditch debris could be instructive but in that case I do not believe the debris was documented.
@All
Here is the basic conundrum about the MH370 crash:
For a ghost flight, we are not expecting a nose dive at Arc7 on fuel exhaustion. Flight sims show that we are expecting a fast glide with up-and-down phugoids and lazy circles.
IG has proposed various theories, assumptions, why a nose-dive might have unexpectedly developed at Arc7. Most recently IG are favoring engine-re-light as the proposed cause of the sudden dive. UGIB report speculates the deceased pilot may have slumped over the steering column. Other ideas have been suggested, and Victor points out B777 fuel exhaustion at altitude is an unusual sim case, for which we really do not have much if any real-life experience.
Meanwhile accident theorists (pilot-is-innocent crowd) have wrongly adopted IG’s nose dive theory as proof of ghost flight, but paradoxically, a ghost flight could actually be a fast glide down. The accident theorists fight tooth-and-nail in support of nose dive. The decent at Arc7 IMHO is probably much more likely to be deliberate, and if you see that Arc7 BFO as the end of the flight (I do not, but if you do) the nose dive was most likely deliberate. Which was something the late William Langwiesche opined in The Atlantic. In other words, the dogged support of the nose dive scenario by accident theorists is perhaps misguided.
Acknowledgement- former contributor @Gysbreght pointed out here years ago the flaw in the Official Narrative’s Nose Dive scenario is that we are not expecting the aircraft to fly level for 2-minutes and then suddenly adopt a spontaneous nose dive upon SATCOM restart. That looks more like an active pilot’s hand on the wheel, and we have de-valued that interpretation…but can you imagine? Inmarsat miraculously captured the last recorded maneuver, but it is not appreciated as such.
@TBill,
You said: “UGIB report speculates the deceased pilot may have slumped over the steering column.”
Yes, but that was just one of five potential causes listed for inducing a lateral asymmetry leading to a high rate of descent.
@370Location,
You said: “My Java site is not dependent upon BFO. Non-adherence to strict BFO can only be used to insert doubt for preventing a quick search before the OI contract runs out.”
That is untrue. It can also do the opposite of inserting doubt – it can establish confidence.
Strict BFO adherence is necessary to identify the True Route. If BFO non-compliance “inserts doubt”, that’s because the route is in error and a search there is highly likely to be a waste of time and money.
@DrB
OK I did say “other ideas”.
In MH370 past discussion, we’ve talked about the “remarkable” and complex accident that would be needed to cause the otherwise apparent deliberate diversion, and we’re basically saying the crash needs to be “remarkable” too.
Imho, Inmarsat incorrectly concluded the 18:25:27 log on originated from a Honeywell MCS 6000 SDU.
@TBill,
You said “. . . and we’re basically saying the crash needs to be “remarkable” too.
I’m not saying that there is anything remarkable about the crash. All high-fidelity simulations without pilot inputs after MEFE show an impact speed and ROD sufficient to disintegrate the aircraft. That’s what the debris show, too.
@John Gilbert: Please tell us more about your theory.
A broader explanation can be found on Facebook MH370 In Search Of The Truth under, MH370 JOINING the DOTS.
@John Gilbert
I asssembled your “Joining the Dots” on Facebook into a PDF.
https://drive.google.com/file/d/1tCc6LJJql88CQwLPiKOIlc3z5BGw0wPD/view?usp=sharing
@John Gilbert: Fiction.
Thankyou Ventus, I really appreciate that.
@ Victor Iannello. If the Inmarsat data was worth a damn you would have found MH370 years ago.
@John Gilbert to be taken seriously you at least need to reference valid and searchable sources for the many claims you have made.
For example, I’d like to know and be able to validate the source of the information about the time and position at which the purported Chinese patrol plane switched to autopilot.
Dear all this is the first time I interact with your blog, it is a new experience for me.
I have received the following message from Jean-Marc Garot:
Tim says:
April 9, 2026 at 7:26 am
@Jean-Marc Garot,
Unfortunately Dr Argiris Kamoulakos ignored that the flap and flaperon were in the retracted position. This is a hard fact and cannot be dismissed. Therefore his ditching dynamics won’t work.
So this is a reply to Tim:
My work on ditching dynamics that I published in the AIAA does not depend on whether the flaperon was extended or not, it depends on the relative velocity of the flaperon to the water true surface and its angle of impact.
The Finite Element simulation of a flaperon having its trailing edge breaking off was made in 2019 to validate the FE model for a case where the flaperon trailing edge should fail according to the theory published in the AIAA, and it corresponds to an extended flaperon hitting a perfectly flat water surface. But this simulation was not to imply that this is what happened to the MH370, it was only a validation type simulation for an eventual study for different scenarios that unfortunately never happened, due to the fact that there was in 2021 a takeover process of my company and I had to “take” retirement… since then I never managed to continue the simulation work for obvious reasons, however I hope to be able to continue the theoretical work in the future plus re-approach the simulation work using different software.
I would be more than happy to collaborate with the IG for scenarios they have identified and I could be of value, as this is an open subject and since nothing has been found yet, all credible scenarios could be candidates.
Thank you very much for reading my work at comp-as-s.com
By the way, I set up the comp-as-s (COMPutational ASsessment Suite) one-man-show company after my retirement, so that I can continue my simulation “hobbies” with my personal proprietary software and continue my MH370 riddle work.
@Argiris: Welcome to the blog! I look forward to more of your comments and analyses.
@John Gilbert
The Inmarsat data appears very solid based on many flight sim runs through the Arcs.
You can see here in this current discussion, Victor is suggesting searching Arc7 in spots all the way from 40south to Java. That’s thousands of miles. I would personally say we may be misinterpreting distance from Arc7, could be 200 miles or more. Finding an aircraft crash in the deep ocean without GPS coordinates may well be impossible even if we made all of the correct flight path assumptions, and I am pretty sure we are not making the correct assumptions.
@DrB @John Gilbert (with reference to @sk999)
RE: Possible Max Crash Distance from Arc7
In response to John Gilbert, who is voicing portions of Jeff Wise’s argument, and also DrB, who voiced above a belief that MH370 lies less than 20nm from Arc7, I would like to show the other side of those arguments.
Let’s say MH370 took a version of IG’s flight path straight 180 south until Arc5, but at Arc5, the heading was changed southeast to fly to 30s & Arc7 (and beyond). In SkyVector, the proposed flight path to Arc7 is simply:
ISBIX-2130S09345E-3098S
(which assumes Arc2 starts at ISBIX and covers 1864nm to Arc7).
From the SIR report, Boeing quotes max MRC range = 2806 at FL400 add +125 glide = 2931nm (possible total flight distance from Arc1 at MRC). Assuming MH370 could fly at MRC to ISBIX from 18:25 (Arc1) to 19:41 (ISBIX/Arc2) I estimate about 610nm of the range was used up to get to ISBIX. That leaves 2321nm available range from ISBIX.
So 2321-1864 distance to Arc7=+457nm max calculated distance away from Arc7!!
Now then, those skilled in the art will quickly realize that this flight path requires substantial descent and slow down before Arc7, such that the max 457nm distance cannot be achieved. We have to discount, and say maybe 70% would be max distance. That is +320nm from Arc7 max. I have not considered this case, but @sk999 above suggests an outside chance that the descent at Arc7 was a false reading.
More likely I suspect the descent at Arc7 is real, and I also suspect the pilot descended as low as 5000-ft at Arc7 to fly in the thick cloud layer. So now, due to the low fuel efficiency at FL050, we are down to 50% of max distance, so that is about 230-nm max from Arc7. Which I personally consider realistic, quite possibly something close to what we actually witnessed.
Hi again, furphy here. @Andrew, my silence should not be misinterpreted. I have taken your critique seriously and and researching and revising the draft paper you have commented on. But, I was distracted by a need to look more closely at the flight simulator data. Thanks to Victor’s efforts the obtain the tabulated data from ATSB with permission to publish, I was able to download the data from this blog and do some analysis. I was amazed by what I found and it is summarized in the paper which is accessible by this link. Read it!
https://www.dropbox.com/scl/fi/k6jkdt5u4pl8mcia59mtc/MH370_SimulatorAnalysis_041026.pdf?rlkey=lkeq5lh84bwphf75ulc430aef&st=dtgg60ez&dl=0
@TBill,
You might be interested in a quick measurement of the distance from ISBIX to my result (which is close enough to the IG original location..and I shouldn’t have had to work so hard to replicate it) is about 2301 nm, +/- 1 furlong..
@Sid Bennet
From your linked paper Analysis of Boeing 777 Engineering Simulator Data for MH370 Table 1. Sample of raw simulator output data (Case C01, near 7th arc crossing) I noticed there is a step change discontinuity in the data series between time 599 and time 600 in all three parameters X (NM), Y (NM) & Alt (ft). This occurs before the T₀ − T₂FO (s) at 602. Is this expected, unexpected or irrelevant?
@John
Good catch! Good eye….The table caption says “raw simulator output data” but what was actually shown were the geolocated (rotated) coordinates, rounded to 2–3 decimal places. The rotation to 187.5°T changes the X and Y values, and the rounding then produced an apparent step discontinuity at t=599→600 that does not exist in the actual simulator output. Here is the corrected table with the true raw simulator values:
Table 1 (corrected). Sample of raw simulator output data (Case C01, near 7th arc crossing). X and Y are in the original simulator flat-earth frame, prior to rotation and geolocation.
Time (s)X (NM)Y (NM)Alt (ft)5961.1611−64.315837,9955971.1640−64.401937,9895981.1665−64.487937,9825991.1693−64.573737,9766001.1718−64.659637,9706011.1747−64.745237,9636021.1772−64.830837,9576031.1797−64.916437,9506041.1822−65.001837,9436051.1847−65.087137,938
The corrected table shows perfectly smooth increments throughout: ΔY ≈ −0.086 NM/s (monotonically decreasing as the aircraft decelerates slightly during the single-engine phase) and ΔAlt = −6 to −7 ft/s. There is no actual discontinuity at t=599→600 or anywhere else. The gust onset T₀ occurs later at τ=+602s (sim time 1081s), well outside this window.
I will correct the table and re-post the pdf tomorrow.
Thanks for your observation.
@370Location
G’day Ed,
Thank you for those references.
I think that we need to be very careful when quoting experts or ostensibly expert reviews, particularly when they lead to unconditionally declarative statements on matters that are otherwise conjectural.
In this instance I’m referring to the assertion that the damage to the trailing edge of the flaperon was possibly caused by water impacting it in the course of a ditching, and the related contention that the flaperon could not have separated inflight as the damage is not consistent with being dropped into the ocean from altitude.
The former possibility – damage in the course of a ditching – was first advanced in the French Direction Générale de l’Armement – techniques aéronautiques (DGA TA) examination report on the flaperon from back in 2015-2016.
In section 10.3. CONCLUSION – SEPARATION AND DAMAGE SCENARIO, of that examination report, the DGA TA “hypothesize [their bolding] a scenario that could have led to the separation and to the damage noted on the flaperon.”
Section 10.3 concludes, with the following,
“Beyond the possible in-flight loss, which is not entirely excluded, this scenario can only be considered as a hypothesis …”.
They go on to note that “the flaperon is aft of the engine which leaves some doubt as to its loading during aeroplane impact with the water“.
Here it is important to note that no specific research was conducted into the practicalities of the hypothesised sequence of events, nor apparently was any evidence gathered that might lend weight (or otherwise) to the hypothesis. No simulations were run, nor were any structural calculations undertaken.
While it may sound unfairly uncharitable, the DGA TA hypothesis was a guess. But it was a guess that, being generally poorly reported, and therefore generally poorly understood, fuelled the confirmation bias of a good many contributors on the matter such that, in some circles, it has effectively assumed the status of confirmed fact.
Subsequent to the DGA TA hypothesis, in the early 2020s Dr Argiris Kamoulakos undertook some modelling for the then MH370-CAPTIO Team. In that work, he looked at both relatively low angle of impact ditching style events, as well as higher angle of impact, higher rate of descent “slamming” (his description) events.
An overwhelmingly important caveat to all of Dr Kamoulakos’s work on this topic is that he did not have access to any Boeing structural data. This shortcoming meant that much of his work was based on (and again for fear of sounding unfairly uncharitable) guesses.
I would argue that you can get a good sense of the validity of the DGA TA hypothesis, and the accuracy of Dr Kamoulakos’s modelling by comparing some of the predictions with actual outcomes by looking at the damage/wreckage data for ditching/controlled-flight-into-water (CFIW) events.
Notably, the damage/wreckage data for ditching/CFIW events does not support the DGA TA hypothesis, nor does it comport with the predictions of Dr Kamoulakos’s modelling.
While there are no B777s in the ditching/CFIW event database, there are a number of somewhat contemporary Boeing commercial aircraft, notably a B767 (Ethiopian Airlines flight 961, Comoros Islands attempted ditching) and four B737s (Garuda Indonesia flight 421, Bengawan Solo River ditching; Lion Air flight JT904, Bali Sea CFIW; Air Niugini flight PX73, Chuuk Lagoon CFIW; and Transair flight 810, Mamala Bay, Honolulu ditching).
Here it is worth noting that the inboard flap on the B737 is of a not dissimilar construction to the B777 flaperon, where a composite (or lightweight) trailing edge component is attached to the rear of the main structural assembly of the device to complete the requisite aerodynamic profile. Unlike the B777 flaperon, the B737 inboard flap is not protected from the hydrodynamic forces of a water impact by the engine, although there would likely be some protection afforded by the main undercarriage if it was extended.
In the handful of ditching/CFIW events involving those aircraft types, there has never been an instance where hydrodynamic impact forces broke the trailing edge component off the main structure, even when the flaps had been deployed for landing.
In fact, there is no evidence that the trailing edge component of any trailing edge device has ever been damaged in the manner of the MH370 flaperon in any ditching/CFIW event on record.
And with regards to Dr Kamoulakos’s work, if you look at Fig. 18 Envelope of ditching scenarios for possible wing fracture of his 2021 paper Aspects of analysis and simulation of a wing ditching scenario, his modelling strongly suggests that in the cases of the two B737 CFIW events, with impact groundspeeds of around 150 knots, and with 1,000+ fpm rates of descent, not only should the inboard flaps have been damaged, but the wing should have fractured. In the real world events, neither of those occurred despite the flaps having been extended fully.
Bottomline, I do not know how the flaperon managed to be separated from the wing. What I can say is that the fact that it did separate is largely inconsistent with what might be expected from a ditching. And I say that based on physical evidence rather than a guess.
@ Jack
That didn’t render very well. Let’s see how this does.
Time(sec),X(nm),Y(nm),Alt(ft)
596,1.161107634693986,-64.3157958984375,37995
597,1.1639677582860326,-64.40185546875,37989
598,1.1664674136125008,-64.4879150390625,37982
599,1.1693273776636597,-64.57374572753906,37976
600,1.1718268878400893,-64.65957641601562,37970
601,1.1746866932059756,-64.74517822265625,37963
602,1.1771860590513692,-64.83078002929688,37957
603,1.1796853257430069,-64.9163818359375,37950
604,1.1821845549548173,-65.00175476074219,37943
605,1.1846836852706222,-65.08712768554688,37938
@Mick Gilbert. Your, “….Bottomline, I do not know how the flaperon managed to be separated from the wing. What I can say is that the fact that it did separate is largely inconsistent with what might be expected from a ditching. And I say that based on physical evidence rather than a guess.”
Below is a wing break hypothesis of some years back, based on physical evidence.
Bloody lengthy but it is a complex issue.
https://docs.google.com/document/d/1ihS5uxdY_s08H1QN-upbCcDVEJYp4q6_/edit?usp=drive_link&ouid=105452605762640315384&rtpof=true&sd=true
@Mick Gilbert:
For your amusement, a friend advises me that I write with a pedantic Gunning Fog Index around 13, and that you are mocking me by responding with a breathtakingly uncharacteristic index of 16.5 😉
It certainly makes sense to compare the damage seen on MH370 debris with the results of past ditching/CFIW events. What we don’t have for comparison is any situation where multiple control surfaces of a plane were ripped apart and detached due to flutter, to then avoid damage from impact with the sea. Whatever happened with MH370 was unprecedented.
The epicenter of my candidate site is within the accuracy of the 7th Arc BTO and seismic estimates for speed of sound of P and S waves in the crust. That proximity would seem to imply either a steep dive, or very low altitude before impact.
I hope that Ocean Infinity will see fit to search my candidate site before their contract expires. If the plane isn’t there, I too will be able to move on.
G’day Ed,
Definitely not mocking.
And yes, guilty as charged. I’m not in the habit of writing for teens, and I’m an advocate for using one 3+ syllable word in lieu of three or more one syllable words.
I hope that OI give your site a go.
@370Location said: What we don’t have for comparison is any situation where multiple control surfaces of a plane were ripped apart and detached due to flutter, to then avoid damage from impact with the sea. Whatever happened with MH370 was unprecedented.
SilkAir 185 and China Air 006 both had flight control surfaces separate from a high speed descent.
@David F
G’day David,
Thanks for sharing that paper (I have a inkling that I’ve read it before).
Again, for fear of sounding unfairly uncharitable, it is a series of guesses about the what may have given rise to the physical evidence that we have.
The wing break hypothesis would carry significantly more weight if someone could model the sort of unpowered descent profile, and uncommanded attitude changes that would cause the overstress required to actually break a wing. Similarly, a physical drop-test with a suitably damaged flaperon to demonstrate that it could impact the water without sustaining damage significantly different to the recovered flaperon would be handy.
Given where we are at presently, more than a decade post-event, and with waning public interest, I’m certainly not holding my breath for the latter.
@Victor Iannello
G’day Victor,
I don’t think that it is correct to attribute the damage sustained by China Air flight 006 to a high speed descent. The investigation report states that,
“Examination of the reliable recorded airspeed data points showed that the Vmo limitation was not exceeded during the descent.” (p.25, 2. ANALYSIS)
Given that the damage sustained by that flight included the APU being torn from its mounts, and the wings being left permanently bent upwards by a few inches at the wingtips, excessive G loading from flight crew control inputs, rather than excessive airspeed, was almost certainly the cause.
The inflight damage sustained by SilkAir flight 185, where the aircraft was estimated to have well and truly exceeded the maximum dive speed Mmd of 0.89M, was most assuredly due to excessive airspeed occasioning anti-symmetrical aerodynamic flutter.
The SilkAir flight, however, had to have been physically held in its dive by manual control inputs, and the excessive airspeed was only achievable because the aircraft was being accelerated by full engine thrust.
@Mick Gilbert: I assure you that a B777 in a steep banked descent would exceed Mmo without engine thrust. I wrote about it in this post:
https://mh370.radiantphysics.com/2017/06/04/mh370-end-of-flight-with-banked-descent-and-no-pilot/
@Mick Gilbert. Thanks Mick.
Ethiopian Flight 961, Comoros 767, ditched left wing down:
https://www.reddit.com/r/CatastrophicFailure/comments/16ebly3/1996_the_crash_of_ethiopian_airlines_flight_961/
Left wing broken:
https://i.imgur.com/8pk6CtR.png
To me, looking at flaperons and right aileron in the video, he had right stick, flaps little deployed if any.
Both engines separated, the left’s possibly hitting the bottom.
All the same the Asiana Flight 214 777 crash, short of the runway, did not break a wing though both engines separated.
As to MH370, aside from a high ‘g’ separation in pitching up as per the Silk Air 737 but unmanned, a manned MH370 ditching parallel to seas in a seaway but to the left of the trough, might have a similarity, particularly if at no-flaps speed. So could an unmanned impact towards the bottom of a phugoid.
Your thoughts?
@All
MH370 book author Christine Negroni on recent Sit Down with Sid, one of her main points is the MH370 O2 tanks, she seems to claim, were not simply refilled, but maintenance actually had to remove the O2 tanks for servicing before the flight. Is this true? I assumed there was some way to routinely refill the B777 O2 tanks in-situ. I have not previously heard anything like what she is saying.
@David F
G’day David,
Thanks for that reference and photo.
I am somewhat familiar with the ET961 ditching. It was a flapless (flaps has not been extended prior to fuel exhaustion, and then could not be extended on the RAT), deadstick attempt that would serve as a salutary lesson for anyone who ever had the notion to plan on ditching a large jet in the open ocean; in short, do not attempt this without thrust!
And yes, it also serves to illustrate that if you stick one wing of a 100+ tonne jet into the water at 180-odd knots, you are very likely to break that wing (as well as the rest of the airplane). Notably, the flip-side of that was that the opposite (starboard) wing was largely undamaged. Apart from some damage to the outer tip of the aileron, all the trailing edge and secondary devices on the starboard wing, together with the flap track canoes, were left intact and in place.
All that said, doesn’t the right wing break hypothesis contemplate the wing being broken while the aircraft is still inflight? I think that that would help explain the preponderance of adjacent trailing edge wreckage from that wing. I’m unclear on what uncommanded manoeuvre would cause said wing to break inflight though.
@TBill
G’day Bill,
They’re serviced via an external port – https://www.instagram.com/reel/CsmMq8Wtjsj/
@TBill
The MH370 SIR states that the crew oxygen system was replenished and says “the system was topped up to 1800 psi”. There is no mention of any oxygen bottles being replaced. If that had occurred, the serial numbers of all the bottles (ie new and old) would have been recorded in the Aircraft Technical Log. There is no such record in the relevant pages of 9M-MRO’s Technical Log.
Did Ms Negroni provide any evidence to support that claim, or did she make it up?
@Mick, Andrew, Bill.
See page 59 of the following Boeing 777-200/300 document for the remote fill panel mounted on the nosewheel bulkhead as seen in Mick’s Canadian Airlines video above.
https://lae.life/wp-content/uploads/2024/03/35___084.pdf
And see page 99 – much more detail.
@TBill
@Mick Gilbert
@ventus45
FWIW, some airlines prefer to replenish oxygen systems by changing out the cylinder(s) instead of using a servicing cart. It’s not a major job and reduces the risks that servicing carts can present. Mind you, that doesn’t appear to have been the case with MH370.
https://www.youtube.com/watch?v=SjkgqGDugJE
Airbus Safety First – Safe Oxygen Servicing
@Andrew
Christine Negroni considers it a “pink flag” re: O2 tank servicing. She infers the SIR report indicates the mechanics requested the O2 tanks, normally filled in-situ, had to be removed for servicing. She infers that there may have been reports of problems from prior flights. She also cites a problem with the pilot’s pre-flight O2 mask test procedure, she claims the test frequently is superficial such that during the test, the pilots are only seeing stale O2 (that is simply O2 hold-up already in the piping) such that the pilot’s are not seeing the O2 system may have a supply issue. Many apparent technical flaws in her arguments, but of course she has the usual ad hominem attacks in her defense.
Her basic argument is ZS, as a middle aged man, probably needed to go to the bathroom. A pressurization problem suddenly came up which a less experienced co-pilot Fariq responded incorrectly which caused a drunken hypoxia. Christine does not know why pressure may have failed, but we get a detailed example about the construction difficulty joining aircraft fuselage sections together properly in the manufacturing process.
Not sure yet how well this podcast matches her MH370 book. Re: TWA800, I find Negroni’s work to be excellent, but in that TWA800 case she is basically reporting/in full agreement with NTSB’s extensive investigation.
@TBill
I can’t see anything in the SIR to support Christine Negroni’s claims: There is no indication the mechanics requested the oxygen cylinders be removed for servicing, nor is there any mention of reports about problems with the oxygen system.
Her comment about the pre-flight oxygen mask test doesn’t make sense. The procedure checks for the continuous flow of oxygen to ensure the cylinder shutoff valve is open. If the valve is closed, there will be a large drop in the indicated oxygen pressure as oxygen in the lines is depleted, and the pressure will not be restored when the test is completed. Pilots are required to check the indicated oxygen pressure before and after the test. If they failed to do so and there were an issue with the supply, an EICAS advisory message would alert them to the problem.
@Andrew @MickG @Ventus
Thank you.
@TBill
No problem. I forgot to mention there’s also an oxygen bleed test on engine start. The system bleeds oxygen for 25 seconds from a valve located in the line between the cylinders and the pilots’ oxygen masks. If there’s an issue with the oxygen supply (eg shutoff valve closed), the line will not re-pressurise when the bleed valve closes and an EICAS advisory will be generated to alert the crew.
Slightly revised (for Pressurized O2 Mask failure)
Christine Negroni’s basic argument is:
ZS, as a middle aged man, probably needed to go to the bathroom. An aircraft pressurization problem suddenly came up which a less experienced co-pilot Fariq responded to incorrectly (which Christine attributes in part to a presumed failure of the cockpit pressurized O2 mask system) which in turn caused a drunken state of hypoxia. Christine does not know why aircraft pressure may have leaked/failed, but she gives a detailed example about the construction difficulty joining aircraft fuselage sections together properly in the manufacturing process.
Further in her website write-up, Christine attributes the Transponder turn-off was probably a mistake by the the hypoxic co-pilot who she feels had intended to dial-in the emergency Transponder code.
@Mick Gilbert. Thanks again.
“…..doesn’t the right wing break hypothesis contemplate the wing being broken while the aircraft is still inflight?” Yes, if the steep descent of the final transmission continued; that is if one sees the asymmetry between the recovered right (‘starboard’ yes indeed!) wing wreckage when compared with the left’s is statistically acceptable.
Mainly the differences are the recovery of the right’s flaperon as against the closing panels from above the left’s; the aileron and canoe/fairing parts coming just from the right, the inner part of the right outboard flap having a jagged trailing edge (torsion?) whereas the left’s piece of outboard flap trailing edge separated along its rear spar (bending?).
If on the other hand one sees the 767 ditching outcome similarity as compelling statistically then the steep descent of the last transmission was not evident at impact. Take your pick.
If one then supposes the presence of a suicidal pilot then to me that makes the continuing steep descent the more likely. Though with APU auto-start there are those two minutes or so delay from fuel exhaustion to the low descent rate evident at the final log-on request (the penultimate transmission), that would not apply were he to have shut the engines down shortly after selecting APU start, fuel remaining.
Then again if it is supposed that there was no active pilot then a high speed spiral dive post fuel exhaustion becomes more likely. To me, based on the Silk Air 737 experience, that might result in excessive ‘g’ in a pitch up.
As to whether, the aircraft could reach wing flutter divergent speed and thence disintegrate in mid-air, there is no evidence of that in the Silk Air 737 powered descent, though it did reach rudder and elevator divergent flutter speed, shedding parts of these. No MH370 debris of rudder or elevator origin has been recovered. While not proof that this did not happen it cannot be disregarded entirely.
I leave aside issues to do with RAT deployment, other aspects than the above of APU start and the availability of hydraulic power from the windmilling engines as being secondary.
Considering all that conjecture, while the decisions about search area priorities to date have been reasonable, all the same it is understandable that no wreckage has been found. It may not be until underwater searching becomes yet more efficient or more evidence is disclosed.
As for search thoroughness to date, while the original bottom search aimed for that, in OI’s that followed the success incentive in the Malaysian contract would tend to encourage calibration and identification approaches with an eye on the best prospects in a given search time, supposing search costs/opportunity costs are closely related to that. The effect, if any, on find probability is unclear.
Hi,
After posting my study of some of the Boeing777 engineering simulator data, I started to think about whether the remaining cases could provide some useful information. So, I did the same data processing for all of the 10 cases. In the process I found that some of my parameter tables had been botched in editing. So here is the updated paper. I can come to no other conclusion than the Boeing simulator data has been significantly misunderstood.
Perhaps it is because we have both accepted the ATSB interpretation as authoritative and not done a detailed analysis of the CSV files that @VictorI had obtained from the ATSB in 2018 and published with ATSB permission. But they were shown as track plots and the actual implied aircraft dynamics were not extracted from the data.
When the 10 case are separated into three groups: Group1 (sequential flameout, high altitude, high speed at 1FO); Group 2 (sequential flameout, -5000ft from Group 1 and slower speed at 1FO); and, Group 3, (simultaneous flameout (no 1FO) high speed at 2FO the pattern of evolution of the flight path becomes evident. Group 3, which is not favored by any researcher results in the immediate spiral descent that has been cited as the basis for ALL of the underwater search efforts. Group 2 is interesting and does result in a spiral descent. But Group 1 (4 cases) has a long straight path after 2FO, before suddenly exhibiting a strong spiral descent. In each case, there was a strong external perturbation at that that time that cannot be attributed to stochastic representation of atmospheric turbulence! It is just short of the magnitude that would have created a reportable incident by itself. There were several other cases that had such singular impulses and they provide further insight into the artifacts.
Absent the spurious impulsive forces, it seems likely that all of the Group 3 simulations would have continued in a stable descent. If that stable descent had continued until impact, the path length would have been 135NM.
**This area has never been searched, and will not be unless these results are taken seriously by the IG**
https://www.dropbox.com/scl/fi/h57wgzajnbslkklnmbrjb/MH370_SimulatorAnalysis_FINAL041326.pdf?rlkey=yycptwap25jaz7gsomcjdkkoh&st=59hmslo6&dl=0
For context, recall that this information was not publicly available until provided to @VictorI in 2018. Prior to that, the official “conclusion” that the aircraft must have gone into a immediate spiral dive was in one form or another the basis for the definition of the underwater search areas.
There is probably more detail that can be extracted from the simulator data, but what I have found is sufficient to assert that the simulator data has been improperly assessed, at least for a flight at about FL400 and about M0.80 just prior to the first engine flameout.
@Sid Bennett: It simply comes down to this: Is it more probable that the simulator data did not capture one or more phenomena that actually occurred (such as an engine restart, a known shortcoming of the simulation) or that our interpretation of final BFO values is incorrect. If you believe the downward acceleration of 0.67g implied by the final two BFO values did occur, then either the plane crashed relatively close to the 7th arc or there were pilot inputs at the end of flight.
@VictorI
The Boeing simulator values were used by ATSB and others to rationalize searching close to the 7th arc. It was also the only apparent economically feasible search strategy resulting from the Bayesian analysis performed by DSTG which did not lead to a specific location on the 7th arc immediately after 2FO. But they missed some of the implications of the simulation and did not identify the apparent problems that my study uncovered.
We should start by trying to better understand the Boeing simulator data. My paper provides a link to the csv data that I produced by coordinate rotating and time syncing the path data, so that others who may wish to process the data have a starting point.
I do not accept the log on data at 00:19:29 and 00:19:37 as definitive as there is no IFE log-on following it to confirm the second BTO value. Also if parts of the 18:25 log-on are suspect, some of the later log-on data should be similarly suspect.
There is nothing in the simulation data that suggests the BTO data interpretation by Holland is plausible. I believe that we questioned in in 2015 and still do so.
Read the paper, study the plots and see for yourself. Only simultaneous two engine flameouts consistently result in strong phugoids.
@Sid Bennett: First, I think you are confusing the BFO and BTO data.
The Holland paper demonstrated that the BFO values increase (not decrease) due to the thermal transient after a long shutdown. In the case of the log-on at 24:19z, the shutdown was not long, and the values decreased. There is no reason to believe the BFO values were anomalous.
The Boeing simulations did not include the effect of an engine restart. The simulations that Mike Exner coordinated showed that an engine restart forces a steep bank. There might be other missing phenomena in the Boeing simulations.
If you believe that the steep downward acceleration implied by the final two BFO values did not occur AND you believe a long glide is possible with no pilot inputs, then the search should area should have been wider.
@Victor,
Does the data that Mike Exner etc collected in their SIM tests [much earlier, I think, than the ATSB sponsored tests], provide any more clarity.
My superficial reading of the analysis that Sid has none suggests that some of the initial preset conditions of the ATSB Sim tests were unrealistic. i.e. some initial speed and altitude settings.
. . correction “done”, not none.
@VictorI
Yes, I menant BFO. The BTO anomalies were explained and the BTOs corrected as I recall. There was a proposed equipment issue in selecting a time slot. But whether this corrupted the BFO data has not been explored, and perhaps cannot. But if there are other log-ins in flight for other instances that can be subject to the same analysis, this issue can be better understood. Offhand, I recall that one of the 18:25 BFO data points was excluded from most analyses.
@Brian Anderson
You provided me with a spread sheet of one of Mike Exner’s simulations. It was at 35kft, I believe, and is not typical of the FL400 data. There are two 35kft examples in the Boeing data and they shows spiral descent. Whether the spiral descent happens appears to be dependent on the starting FL and M, and whether the event was a sequential flameout or a simultaneous flameout. So each of the simulations has to be considered in context.
All of the results so far may be part of the mosaic of initial conditions that may actually be consistent. So the more detailed information that is assembled, the better.
@Sid Bennett
RE: Analysis of Boeing 777 Engineering Simulator Data for MH370 (revised 13 April)
If I understand correctly your conceptual search area Figure 9 is bounded by a series of right angled triangles being where the sum of the adjacent (D2) and opposite (D1) sides equals 135 NM. Wouldn’t possible 135 NM paths be limited by the hypotenuse of the same triangle, namely the line that connects the non-coincident ends of each D1 and D2 pair? That would that give a half (or perhaps even full) circle search area with a radius of 135 NM centred at the arc crossing point.
@Sid Bennett
I think your paper is misidentifying the timing point of 00:17:29 in the raw simulator data (or at least it’s different from what ATSB presented?).
Take Case 2 as an example. In figures 2 and 3, what is labeled “gust onset” (the sudden loss of XY-stability and change in altitude/speed) should actually be the 00:17:29 point. In the raw data, it happens at around t=1070s. 2 mins later (1190s) the Y-value is -109.6. If you create an XY-graph (like linked below), and look at the point in the graph that has this Y-value, the location corresponds to what ATSB presented as the 7th arc location in the 2016 figure (and the equivalent figure in the 2018 article where this spot is marked “2 mins”). Your paper puts this 7th arc location much higher in the graph (around Y=-63).
https://www.dropbox.com/scl/fi/mhlguy1d2yzcjemjorhkt/346464.png?rlkey=7iip8qp5bjzopakailkrpl9xd&st=35dquh0w&dl=0
So, I think the apparent XY-stability prior to the “gust onset” is simply the autopilot maintaining this stability prior to 00:17:29.
@Sid Bennett: You want to throw out the downward acceleration implied by the final two BFO values, but I fail to understand why. The final BTO values are corrected in a manner consistent with previous logins, and transient BFO values after a long power down have also been explained as repeatable behavior of the OCXO which did not occur at 24:19z due to the short duration of the power down. So what evidence is there to dismiss the final BFO values as evidence of a steep descent?
@eukaryote
Thank you for your question. Your linked plot is not annotated, so it a little hard to correlate with Fig. 2 of the paper.
But, my interpretation of Fig. 2 (which actually shows 4 cases, three of which are so similar that their traces almost completely overlap) is that 1FO occurs at just after 00:11 where there is a abrupt small decrease in ground speed. The altitude is held constant by the autopilot as the speed decays. (One engine provides just enough thrust to do it as the plane is much lighter..it is almost totally out of fuel). All of the traces show an abrupt nose down and and recovery at 00:17:29, which is interpreted as 2FO. The crossing of the 7th arc, 2 minutes later is in agreement with the ATSB statements.
Now, it gets interesting. Look at Fig.3, for example, where the vertical component of acceleration,nz,is plotted. Three of the four traces show a strong impulse shortly after 00:27 (the C02 case is already in a spiral mode and shows a strong disturbance at the same time in bank and yaw (see Fig.5). The spiral mode for C02 starts at 00:22:30, coincident with an impulse in nz that does not appear in the other cases.
I believe that this is a consistent interpretation of the data.
@John
My stated assumption for the search area is that the plane continues fly along the azimuth that it had at 2FO until such time as it encounters a turbulence effect that causes a spiral mode to develop. It then makes a 90 degree turn (other angles will result in a end of path within the search area) and flies until the sum of distances D1 and D2 equals the total estimated length of the glide. The length of the glide has been estimated at 135nm based on the along path length of Group 1 cases (that is, straightening out the spiral).
Years when Geoff, Barry and I proposed the glide, we suggested a semicircular search ares. At that time we had no notion that the plane would continue to fly straight until the spiral mode (if any) developed.
The previously searched area can be omitted from the new search, but if there are substantial gaps in the earlier search they should be filled in.
@Sid (or @Victor)
Did you provide somewhere recently the CSV data of the simulator descents? I’ve been wanting to re-review that data. I am especially curious to see speed/etc at the water impact, if they give that.
@TBill: In August 2018, I posted an article with the numerical results from the 10 simulation runs from Boeing.
https://mh370.radiantphysics.com/2018/08/19/end-of-flight-simulations-of-mh370/
@Victor
This issue goes back to almost the very beginning, so a bit of this is from memory.
The Inmarsat Classic Aero system uses Slotted ALOHA on the R-channel (aircraft-to-ground). In this protocol, the aircraft transmits in a randomly chosen time slot. When the SDU cold-starts after a power interruption, it needs to acquire synchronization from the P-channel pilot signal before it can transmit in a correct slot. The details of the synchronization process that occurs between the log-on request and the log-on acknowledge are not described in any public documentation that I have been able to find. If you have a reference I would appreciate it.
You were involved in determining that the log-on synchronization had not fully completed when the log-on acknowledge was transmitted — at both the 18:25 and 00:19 events — and you determined how to correct the BTO of both signals accordingly.
Holland took the position that the BFO is not affected by the same synchronization problem and therefore requires no correction. I do not believe this position has a sound technical justification. The BTO and BFO are both products of the same transmission, and if the slot timing was not yet fully resolved when the log-on acknowledge was sent, there is no a priori reason to assume that there might not also be an anomaly in the BFO as well.
Ashton et al. explicitly recommended discounting the BFO at 00:19:37, along with the values from 18:25:34 to 18:28:15. Holland subsequently reversed this judgment using the OCXO thermal settling argument. I recall many discussions of this on the blog. He does not appear to have consulted with Ashton et al. before doing so. His position is consistent with the ATSB position that the Boeing simulator data showed a spiral mode immediately after 2FO, but that is a weak justification for the conclusion.
I used the BFO at 00:19:29 — the log-on request, which both Ashton and Holland regard as reliable — in my kinemetric simulation of the path to the 7th arc I find this value to be consistent with a descent rate of approximately 4,500 fpm at the arc crossing. This is broadly consistent with the Boeing simulator Group 1 results, which show descent rates in a similar range during the stable glide phase prior to the anomalous impulsive perturbation events in the simulator results.
Finally, there are no further transmissions after 00:19:37 with which to validate whether the OCXO had settled by that point. Holland’s steep-descent conclusion rests entirely on two data points from a truncated initialization sequence, one of which the original Inmarsat authors said to discount.
I am not sure this will resolve the issue, but it does not detract from the information that can be gleaned from the Boeing simulator data.
@Sid Bennett
I’ve highlighted in red the 2-minute part from 0:17:29 to 0:19:29:
https://www.dropbox.com/scl/fi/eywat8audz67c4l3nemlv/465454.png?rlkey=zugckncckq6fto1lemuxix8ux&st=3rlmd3o4&dl=0
The 0:19:29 is at the same point of Case 2 as the “2 mins” marking in the 2018 figure (which is also the same in the 2016 ATSB figure):
https://mh370.radiantphysics.com/wp-content/uploads/2018/08/EOF_Paths.png
For ground speed and altitude, these same points are (red is 0:17:29 and green is 0:19:29):
https://www.dropbox.com/scl/fi/nx71j8icjj0lb411eugqs/154554.png?rlkey=qzz6u385bnxior0lh6tsg2i9i&st=ehztf3kj&dl=0
As you may see, the red line (0:17:29) is around the same spot that you’ve labeled as “gust onset” in figure 2, while you use a separate spot about 10 minutes earlier (yellow line) for the 0:17:29 point.
So if 0:17:29 and the corresponding loss of autopilot happened once in the simulation, it can’t be in to parts of the same data depending who’s interpreting it. Either it’s incorrect in your paper or the 2018 article (and 2016 report).
It’s the same issue with many of the other cases (compare the 7th arc position in your figure 1 with the 2018 figure).
@Sid Bennett: A log-on would not occur without synchronization to the P-channel. What we see is synchronization to a different slot than for other events. The correction to the BTO for the log-on request and log-on acknowledgement was determined from the log-in events for MH371 and applied to MH370.
The BFO is not affected by this time shift at log-on, so I don’t understand why you believe the BFO values are invalid.
Of course Holland was working with Ashton! Holland’s work on the final BFO values supersedes Ashton’s approach. Do you really think the DSTG and Inmarsat were not strongly collaborating?
If the OCXO had not settled, the descent rates would be even higher, as the thermal transient introduces a shift upwards in value.
@eukaryote
Please look:
https://www.dropbox.com/scl/fi/es32wqqj7bt73zslj7ec3/Markup-of-Fig.-2.pdf?rlkey=oytnvnbyrjck3uk4nbmhny52c&st=1f7lnopd&dl=0
I have marked up your version of the figure with my interpretation of the data. The difference between the figures is you use of the absolute time of the simulation, whereas I presented the data oriented to an interpretation that is synchronized to what I have identified as 2FO.
Focusing on the speed, the 1FO clearly observed as the speed beginning to decrease, so as to maintain altitude under autopilot and possibly VNAV control. The speed continues to decrease steadily until 2FO when the altitude drops sharply and then continues a steady descent till the impulse occurs and the phugoid starts. The 7th arc is a fictive isopleth determined as being 2 minutes along the path (at some speed) of constant range to the satellite. The green line has no relationship to a specific event that I have identified.
Try to look at the other plots also, it takes time to fit all of the pieces together. On the other had my eyes became crossed at some point and I will try and further explain the plots of the Boeing data if I can….
@Sid Bennett, @eukaryote: My interpretation of the data is similar to @eukaryote’s interpretation: The initial change in groundspeed is due to the first engine flameout, and the initial change in track is due to the loss of the autopilot for either the “normal” or “alternate” electrical configuration, although the loss of autopilot occurs after the second engine flames out for the “normal” configuration and after the first engine flames out for the “alternate” configuration.
For the case of the “normal” electrical configuration, after the first flameout, the plane decelerates while initially holding altitude at full thrust from the remaining engine. However, the single-engine thrust is not sufficient to maintain cruise altitude, so when the airspeed approaches stall, the plane begins to descend with the autopilot still engaged. After the second engine flames out, the autopilot disengages, and the plane begins to bank. Wings level flight only occurs with the autopilot engaged.
Also, for clarity, what you refer to as “heading” is actually “track” and what you refer to as “yaw rate” is actually “turn rate”.
https://www.dropbox.com/scl/fi/h57wgzajnbslkklnmbrjb/MH370_SimulatorAnalysis_FINAL041326.pdf?rlkey=yycptwap25jaz7gsomcjdkkoh&st=nhrrjv9g&dl=0
I realize that the pdf cut off the bottom of Fig.7 which is the only set of cases that has a high descent rate early in the simulation. But those cases are for simultaneous two-engine failure.
@ Victor
What is the stall speed at that altitude?
@Sid Bennett: With no flaps and at 173 MT, I would estimate it to be around 150 KIAS. (@Andrew would know better than me.) At standard conditions and FL400, this corresponds to a true airspeed of around 295 knots. In your figures, the descent begins at a groundspeed of around 300 knots, so that is roughly consistent.
Also, envelope protection would start a descent before stall speed is reached. There is also a stick-shaker speed and minimum maneuvering speed, so there would be speed margin above the stall speed.
@VictorI
If one engine was still running then, would you expect an abrupt nose down and and a short period of increased speed when the autopilot disconnected?
@Sid Bennett: I again looked at the Exner simulations. It looks like the MMS was 200 KIAS, which is the top of the amber band on the PFD. With the autopilot engaged, I would expect the descent to begin around that speed with one engine.
As for what the speed does when the second engine is lost and the autopilot disengages, I suppose it would depend on what the trim speed was prior to the disconnect. Again, @Andrew would understand the dynamics.
I should add that in the Exner simulations, the second engine flamed out before the engine out descent began.
@Victor
@Sid Bennett
In the simulator, at 174.3 MT and FL360, the minimum manoeuvring speed was 208 KCAS and the stick-shaker speed was 168 KCAS. According to the regulations (14 CFR 25.207c), the stall warning must begin at a speed at least 5 knots or 5% CAS before the stall, whichever is greater. That puts the stall speed in this case at about 160 KCAS.
> “If one engine was still running then, would you expect an abrupt nose down and and [sic] a short period of increased speed when the autopilot disconnected?”
When I tested this case in the simulator (aircraft decelerating in level flight with one engine still running), the AFDS stall protection pitched the aircraft down by about 5 degrees when the speed was just above stick-shaker speed. The speed increased due to the lower attitude and, after a bit of hunting around, stabilised at about 8 knots above the stick shaker.
The autopilot did not disengage until the second engine failed. At that point the aircraft pitched down by about 2 degrees, the speed and rate of descent increased and the aircraft settled into a gentle phugoid with a period of about 45 seconds.
@Sid Bennett
BTW, the draft of your paper “MH370 Post-Flameout Descent” claims the B777 PFCS uses a C* pitch control law. The pitch control law implemented on the B777 is actually C*U, not C*. The “U” term adds speed stability to the mix, so the system commands pitch adjustments that will return the aircraft to a reference trim airspeed if the speed deviates. That behaviour is quite different to the C* control law implemented by Airbus on its FBW aircraft, which may have implications for your analysis.
Background information on the C* and C*U control laws and their differences may be found in the following paper:
FLY-BY-WIRE AUGMENTED MANUAL CONTROL – BASIC DESIGN CONSIDERATIONS
@Andrew
Thank you for your report on the simulator run. I am trying to further analyze the Boeing simulator data in order to determine the stability boundary conditions. Knowing the FL and M and whether the or how long the plane flew straight after 2FO would be very helpful.
Thank you for the copy of the paper on FBW. I am retired and aeronautics was not my field of work, so my library and knowledge of the subject is limited, as you must have recognized by now. But, I try to find the most relevant technical and practical information possible to continue to participate in the work of this blog. Your tolerance for my errors and corrections are sincerely appreciated.
I have not yet read the paper, but would I be close to correct if I concluded that in the secondary mode the Boeing C*U became C*, since air data is might not be used due to loss of pitot tube heating. On the other hand, since the air data itself is not lost, the PFC determines whether the sources of air data agree, and it not, it would be rejected. The FBW approach seems to provide an enormous amount of scope in the design to optimize the flying of the plane.
@All:
There were some previous comments about flight constraints along the 6th arc imposed by the BTO and BFO data. I was able to introduce another constraint which made it possible to determine the groundspeed and track at the 6th arc. This track could then be extrapolated forward to determine the latitude of the crossing of the 7th arc.
The additional constraint is the rate of change of the BTO at the time of the 6th arc (24:11). By fitting the BTO at handshakes 2-6 to a 3rd order polynomial, the rate at 24:11 was determined to be 43.3 us/min. The BTO rate is composed of twice the two components: the range rate between the GES and the satellite, which is -16.6 kn (negative because the satellite is moving towards Perth, located in the southern hemisphere); and the range rate between the satellite and the aircraft. This second range rate can be further subdivided into the range rate due to the satellite velocity (which varies long the 6th arc) and the range rate due to aircraft velocity (which varies along the 6th arc and also the groundspeed and track of the aircraft). So for a given latitude, the range rate due to aircraft velocity can be calculated, which is the radial velocity of the airplane away from the satellite, Vrad.
We also know that the BFO is a very weak function of the groundspeed along the e-w direction, so only the n-s component GSns determines the BFO. So, at a given latitude, the groundspeed GSns can be calculated that matches the measured value of 252 Hz.
Here’s where things get interesting. Knowing Vrad and GSns, we can write the equations:
(1) Vrad = GS*cos(e)cos(a-T)
(2) GSns = GS*cos(T)
where GS is the groundspeed, T is the track, a is the azimuth radial outward from the satellite to the aircraft, and e is the satellite elevation angle relative to the airplane. The elevation e is a function of the latitude, but only weakly varies due to the oblateness of the earth, and is about 39.7 deg. The two unknowns, the groundspeed and track at the 6th arc, can then be calculated as a function of latitude and extrapolated forward to the 7th arc to determine the latitude of the crossing of the 7th arc. By then varying the BFO error by +/- 7 Hz, the allowable range of speed and track can be determined at each 6th arc crossing.
The results are shown in the figure below. If we set the upper limit of groundspeed to something like 520 kn, then the possible range of 7th arc latitudes can be estimated. For an upper limit of 520 kn, the range of possible 7th arc latitudes is -37.6 to -13.5.
https://www.dropbox.com/scl/fi/gok7zcegoepwskas1xn4k/2026-04-16-7th-arc-latitude-from-6th-arc-constraints.png?rlkey=ciocwkgrwm7f0xs6sfvcotah4&dl=0
As a reminder, the assumptions for this analysis are a smoothly varying BTO versus time curve so that the slope at 24:11 can be estimated, and a permissible BFO error range of +/- 7 Hz.
@Victor
Interesting! If there was a means to test the assumption of level flight…
@Sid Bennett
RE: “…would I be close to correct if I concluded that in the secondary mode the Boeing C*U became C*, since air data is might not be used due to loss of pitot tube heating. On the other hand, since the air data itself is not lost, the PFC determines whether the sources of air data agree, and it not, it would be rejected.”
No, secondary mode does not use C* in place of C*U if the air data is unreliable. C* also needs speed data for the blending of pitch-rate and load-factor. As mentioned in your paper, that blend is scheduled by airspeed, with pitch-rate dominant at low speed and load-factor dominant at high speed. Airbus FBW, which uses C*, also reverts to an alternate mode (ALT2B) if airspeed data becomes unreliable.
In the B777, secondary mode uses “simplified computations to generate flight control surfaces commands” [B777 FCOM 9.20.7]. The FCOM also states “Airplane pitch control is different in the secondary and direct flight control modes. The control columns now command a proportional elevator deflection instead of a maneuver command” [B777 FCOM 9.20.12].
The secondary mode reversion occurs when the PFCs detect an actual loss of necessary data. The reversion is not caused by the loss of pitot heat alone, which may or may not cause a loss of data.
RE: “Knowing the FL and M and whether the or how long the plane flew straight after 2FO would be very helpful.”
I don’t have detailed data, but the simulation began at FL360 and the aircraft was initially descending (1FO) with the speed at about 176 KCAS. My notes shows that 2FO occurred about 4 minutes after the aircraft started descending, by which time the altitude was FL330 and the speed was still 176 KCAS (M0.51).
The aircraft initially remained wings level, but the rate of descent increased to 2,500 ft/min and the speed increased to 235 KCAS. It then gradually rolled left to a bank angle of 5 degrees and developed a phugoid motion, with the speed varying from 212 to 235 KCAS and the rate of descent varying from 500 ft/min to 3,500 ft/min. The bank angle increased to about 7 degrees during this motion and remained constant. I had to stop the simulation shortly after that point due to time constraints.
@VictorI
Interesting estimates based 6th arc BTO rate and BFO.
Could you do the same for arc 5 and then solve for groundspeed, assuming this stayed more or less constant after arc 5?
@Niels: Yes, the same analysis can be done for Arc 5. However, my goal was to provide a range of possible 7th arc crossing latitudes with the minimum of assumptions of what occurred prior to Arc 6 other than the shape of the BTO curve which only affects the rate of the BTO at Arc 6.
@VictorI
If you would have a reasonable estimate for the groundspeed, the indicated latitude range would be rather confined. For example, groundspeeds just under 500 kts seem to indicate roughly S33-S37 deg.
With the same BTO curve also applied to arc 5 (plus the arc 5 BFO) it would probably be possible to plot arc5->arc 6 distance vs groundspeed, and we know the delta time, hence probably groundspeed can be resolved for the leg leading to arc 6. Assumptions added are level flight and constant groundspeed arc 5 -> arc 6.
@Niels said: For example, groundspeeds just under 500 kts seem to indicate roughly S33-S37 deg.
But also S22 to S14.
Let me think about what additional information we can draw by looking at other arcs without just recreating the LNAV paths that satisfy the BFO criteria.
@Vor. Your deductions on Arc 6 BFO are elegant. But do they tell us anything we didn’t already know?
We knew that the LNAV solutions exhibit rotational symmetry so that there is a straightforward dependency on initial azimuth (and therefore track at Arc 6) and final latitude. We also knew that the further south your solution, the larger your BFO error at Arc 6. By focusing exclusively on Arc 6, you have demonstrated the range of latitudes compatible with BFO residual <7Hz.
Previous path analyses would show the same result for Arc 6, but also have the merit of including BFO residuals across the other 4 datapoints (6 if you include the phone calls). When you include all BFO datapoints and assume no manoeuvre, you find best BFO fit is obtained for paths ending near 37S,
I understand that your 3rd order polynomial doesn't explicitly assume LNAV path with no manoeuvre. In practice, we know that the function/curve of BTO against time is reproduced by great circle paths with constant M and altitude. Alternative nav modes produce (greatly) inferior fit.
Conclusion: latitude discrimination using inferred BFO error at Arc 6 by new method doesn't add insight. Methods that take account of combined residuals across all arcs provide greater statistical discrimination.
Extending the discussion to path compliance to BFO and BTO across all arcs.
The three plots linked here show the relative probability of terminus latitude using BTO, BFO and BTO & BFO combined. Also shown are the effect of relaxing the assumed BFO sigma and/or altering the respective weightings of BTO and BFO. Conclusion: judging by the satellite data only, and assuming LNAV no-manoeuvre path from Arc 2 onwards, termini north of 32S are vanishingly unlikely, while unsearched termini south of 39.5S have substantial probability.
@Andrew
Thank you for the additional information. The simulation you reported seems to be somewhere between Group 1 and Group 2 in its characteristics.
Your comments on Secondary mode are interesting and at least have the authority of a Boeing publication. But I have been confused, and still am confused, by the actual criteria for switching to secondary mode. I understand that the loss of critical data forces the switch, but at 2FO, only the autopilot is lost. Its the same thing as disengaging the autopilot and flying the plane manually. (Without a pilot, this means that the cockpit controls provide a neutral input, I believe.) If and when the air data becomes corrupt, then the Secondary mode would be used. Since this depends on the actual atmospheric conditions, would a simulator be configured to act as if it had occurred (either promptly or after a fixed time)?
This is somewhat less important for the analysis of the simulator data as the paths are what the simulator did, than the earlier paper where I was trying to deduce the configuration of the PFC by reading about the design details in published literature.
Since the FCOM is intended to guide users, rather than to expound on design philosophy I can see that there might be some uncertainty in confidently determining the actual state. It is why it is so difficult for people to agree on a common understanding of a technical phrase when their point-of-view of the information differes (e.g., pilot vs aeronautical engineer).
Again, I appreciate your reasoned approach to the discussion.
https://www.dropbox.com/scl/fi/xsnk3hv304amrhhv4p19q/Simulator-Analysis-041826.pdf?rlkey=30786ip9gmnej3ovkvdpnotrg&st=3vba6uwl&dl=0
This is a link to an updated version of my Boeing simulator data analysis. I have added sections describing the “turbulence” events and used the data that @Andrew provided as an informal simulator result in the context of the events analyzed in the paper.
I believe that the results indicate that a significant revision to the underwater search strategy is warranted, so long as you agree with the general premise that the flight was at a high altitude and a high speed to reach the 7th arc.
@Victor
A “curved” or “dogleg” path southeasterly requires slow down probably descent and optionally heading change, due to the upward swing of the Arcs. Such changes probably start around Arc5 but could start a little before or after Arc5. So I will be looking at the proposed rule in that light.
https://www.dropbox.com/scl/fi/yxzod3odf30nm903yfh9w/kmz-for-10-cases.pdf?rlkey=52qv42gkq0bbeo86zywttkugt&st=jlqul5d0&dl=0
You might like to see a GE Plot of the Geolocated tracks of the 10 Boeing simulations. You can guess which 4 correspond to the high-altitude high-speed initial conditions.
@Sid Bennett
RE: “But I have been confused, and still am confused, by the actual criteria for switching to secondary mode. I understand that the loss of critical data forces the switch, but at 2FO, only the autopilot is lost. Its [sic] the same thing as disengaging the autopilot and flying the plane manually.”
The autopilot disengages at 2FO as a consequence of the PFCS reversion to secondary mode, as Victor and I have tried to explain. The autopilot is not available with the PFCS in secondary mode, as stated in the Boeing manuals [B777 Maintenance Training Manual, 27-02-00, p.8; B777 FCOM, p.9.20.7].
Your claim that “The autopilot disconnect at 2FO is caused by loss of power to the three Autopilot Flight Director Computers (AFDCs), which are powered from the AC transfer buses and de-energise with them at 2FO” is wrong, as I have already explained. The L and R AFDCs are powered by the 28V DC L & R buses respectively, while the C AFDC is powered by the 28V DC Battery bus [B777 Maintenance Training Manual, 22-11-00, p.13].
The L & R AFDCs lose power at 2FO, but the C AFDC remains powered by the Battery bus. Only one AFDC is needed for the autopilot to remain operative, albeit in a degraded state (ie no auto-land). In short, disengagement of the autopilot at 2FO is NOT caused by a loss of electrical power to the AFDCs.
The following might help explain why the PFCS reverts to secondary mode at 2FO:
With the PFC disconnect switch in AUTO, the PFCS only operates in normal mode if both the following conditions are satisfied:
• The needed input data is available.
• At least one PFC and three ACEs are in normal mode.
[B777 Maintenance Training Manual, 27-02-00, p.40]
Some input data is lost at 2FO, as already explained. Furthermore, two of the four ACEs lose power and are inoperative.
The L2 ACE is powered by the 28V L DC bus. It has no other source of power. The 28V L DC bus loses power at 2FO and the L2 ACE therefore becomes inoperative [B777 Maintenance Training Manual, 27-02-00, p.24-25].
The R ACE is powered by the R PSA, which is powered by the R engine PMG or the 28V R DC bus. There is also a 28V FCDC battery that provides power to the PSA for up to one minute during power supply switching [B777 Maintenance Training Manual, 27-02-00, p.24]. The battery is sized specifically for momentary bridging during power transfers of the order of seconds; it is not intended for sustained operation.
The R PMG and the 28V R DC bus stop supplying power at 2FO, leaving the 28V FCDC battery as the only power source for the R PSA. Once the battery is depleted, the R PSA loses power and the R ACE becomes inoperative.
In summary, the L2 and R ACEs are unpowered and inoperative shortly after 2FO. The conditions required for the PFCS to operate in normal mode are therefore not satisfied and the PFCS reverts to secondary mode. Also note that a loss of the L2 and R ACEs causes control of the right elevator and several pairs of spoilers to be lost. Contrary to the claim that “only the autopilot is lost”, there is significant degradation to the PFCS at 2FO, which causes the system to revert to secondary mode.
Glossary:
ACE = Actuator Control Electronics
FCDC = Flight Controls DC
PFC = Primary Flight Computer
PFCS = Primary Flight Control System
PMG = Permanent Magnet Generator
PSA = Power Supply Assembly
RE: “If and when the air data becomes corrupt, then the Secondary mode would be used. Since this depends on the actual atmospheric conditions, would a simulator be configured to act as if it had occurred (either promptly or after a fixed time)?”
The simulator is configured to accurately replicate the systems degradation that occurs at 2FO. It is not configured to assume the loss of pitot heat causes the air data to become unreliable. If an instructor wanted to add unreliable airspeed to the mix, they could activate a separate function to replicate that fault in addition to the faults that occur at 2FO.
@Andrew
Please keep going. Many years ago I remember getting to this point in discussions and thinking (and I once said it): “Let me explain this to you again in words of one syllable or LESS!”
Your detailed description set me thinking. There are 4 ACEs and three of them are needed for Normal operation. Once the degredation occurs and the autopilot is lost and the PCE goes into secondary mode, there is no automatic recovery. And this eventually happens after 2FO.
But you also pointed out that the loss of the L2 and R ACEs is probably what triggered it. Yet the R ACE is sustained for “a period of about 1 minute” by the FCDC battery. During this time, the R AVE is active, RAT is deployed and the APU starts up. The APU comes on line in about 1 minute, so that if the battery lasts that long, things are OK for as long as the APU runs. If not, for as long as the battery backup lasts.
The posted figure is the data derived from the Boeing files that Victor obtained. (The green trace shows that I misidentified the 2FO time for that case by about 25 seconds). I do believe that the Boeing simulator data is and should be trusted.
You can see the abrupt transient within the first minute that damps out. I also have studied the heading drift for the relevant cases, and in the period before the abrupt start of the spirals, 3 of the cases had no drift at all and the 4th case showed less than 1 degree.
I am stubborn about wanting to understand the underlying mechanisms to better interpret the simulator data.
Now for the questions of turbulence simulation….I believe there are standard models for this. Are they generated as statistical numbers for each event separately or are they pre-generated and selected to give the pilot a controlled experience? If it is the latter, I can imagine that the impulse might very well occur at the same time with respect to the start time of the simulation. Are the simulated turbulence effects chosen for various phase of flight such as take off, climb, cruse etc and are the specifically set up as part of the simulator initial conditions?
https://www.dropbox.com/scl/fi/i6upbj5hke7jorp7hn2dk/Fig_2FO_transient_ACE_window_1.png?rlkey=c1rntb7h2zdf1jifr1mxmy0eu&st=pp7yh01k&dl=0
@All,
With all this discussion of ‘secondary mode’ I’d like to highlight how I think events unfolded…I believe that the point of reversion to secondary needs to be moved all the way back to IGARI. Yes I know that’s controversial on this site!
Rupturing oxygen cylinders WOULD definitely disrupt the pitot/static feeds sufficiently to cause the switch from normal to secondary. As Andrew highlighted there would be a “loss of input data” to the flight control computers.
So it’s just a case of just how stable the aircraft is in this condition with no active pilots. Could it prescribe the flight profiles many have postulated , most have needed much imagination, or could a random meandering flight fit too.
I believe a meandering flight can fit, the primary radar data we have seems to imply a certain amount of wandering.
I have yet to be persuaded we have enough evidence not to ignore this accident scenario.
Dear Victor,
Thank you very much for the welcome, I shall try to participate as I can.
I don’t know much how blogs work, it is the first time, and I thought that if someone writes something about me I shall be informed so that I could answer.
I was in Greece for Easter and now back in France I just looked to see if somebody wrote anything and I find an important contribution from Mick Gilbert.
In fact he points out a serious misconception my figure 18 gives, and I thank him very much for this, which was not intended when I wrote the article. The original figure has always been in my website (https://comp-as-s.com/mh370/flaperon/) and it shows the locus where we have THEORETICALLY flaperon TRAILING edge fracture (nothing to do with the flaps). This was the state of my research up to 2020 after which I stopped (hoping to restart again … let’s see). In the article I added a vast circle to say (but improperly presented) that somewhere within this circle COULD lie an area where the wing might break. So IF we identify it, then being also in the area where trailing edge fracture THEORETICALLY could happen, it could qualify for a wing breaking and releasing the flaperon from its supports after it has lost its trailing edge. In reality, if this area is identifiable, then it will give information about the type of missed approach that might had happened … IF it had happened !
The curves shown correspond only on THEORETICAL fracture of the trailing edge of the flaperon (nothing to do with the flaps!) and the theoretical work about the wing fracture was only initiated but never finished and was not part of the publication, only the comment that if the sea impact deformed the wing by about 24 ft tip deflection, wing fracture will most probably happen. The wording in the publication was very unfortunate, I had a lot of pressure at that time as I was departing from my company.
There was a comment, quite correct, that the finite element models of the 777 are with Boeing and what I did was a guess work.
But this was the whole point! It was ENGINEERING guess work, not fiction. My main motivation was the comment from the DGAC that they couldn’t perform simulations due to inadequate data. But if you have enough structural simulation know-how, as I had in my life, there is ALWAYS something you can do that has some engineering value. There is another publication that Mick Gilbert can read to see the construction process of my model and the type of assumptions that were made.
One last comment, the B767 (Ethiopian Airlines flight 961, Comoros Islands attempted ditching) event should not be taken into consideration as a ditching event, as the aircraft hit a coral reef (not just the left engine) which precipitated its disintegration. In aviation ditching research, this attempted ditching event is left out for this reason, so it should not be in the discussion for the MH370.
@Sid Bennett
RE: “But you also pointed out that the loss of the L2 and R ACEs is probably what triggered it.”
I didn’t mean to imply the loss of the L2 and R ACEs is “probably what triggered [secondary mode]”. There are two conditions that must be satisfied for normal mode to operate; the necessary input data must be available AND at least one PFC and three ACEs must be in normal mode. NEITHER condition is satisfied in the 2FO case.
To understand why the first condition is not satisfied, consider the following. The L2 ACE has a single power source; the 28V L DC bus. That bus loses power immediately when the AC transfer buses lose power at 2FO. When that happens, the L2 ACE, its associated flight control components and their sensors become inoperative. That affects the following (Note: all references below are from the B777 Maintenance Training Manual):
• Left outboard aileron redundancy
• Spoiler pairs 4/11 (speedbrake function)
• Spoiler pairs 5/10
• Right flaperon redundancy
• Right stabiliser trim redundancy
• Right elevator redundancy
• Left elevator feel actuator
[27-02-00, p.27; 27-30-00, p.268]
My former airline colleagues and I were told by Boeing that the loss of these components and their associated sensor inputs is enough to cause the PFCS to revert to secondary mode. I see no reason to doubt that advice.
RE: “Yet the R ACE is sustained for ‘a period of about 1 minute’ by the FCDC battery. During this time, the R ACE is active, RAT is deployed and the APU starts up. The APU comes on line in about 1 minute, so that if the battery lasts that long, things are OK for as long as the APU runs. If not, for as long as the battery backup lasts.”
The claim that it takes about one minute for the APU to come on line seems to be based on the time it takes for the APU generator loading to become available after the APU controller engages the electric starter motor during the start sequence. The following video shows the indications on the APU Maintenance Page during an APU start on the ground:
https://youtu.be/U1iXyYyXMko?t=65
The APU RPM is displayed on the SPEED SENSOR 1 & 2 lines in the left column, and the APU generator load is displayed on the APU GEN LOAD line in the right column. The APU RPM starts to increase at 1:14 and the APU generator picks up the electrical load at 2:09, some 55 seconds later (approximately one minute).
What is not shown is that the ELMS must open the APU air inlet door before the start sequence commences. The APU air inlet door must be fully open [49-00-00, p.124], which takes about 30-40 seconds [49-10-00, p.44]. In the APU autostart case, the ELMS energises the control relay to open the air inlet door when the AC transfer buses lose power. Accordingly, the total time from the loss of AC transfer bus power until the APU generator comes on line is about 90-100 seconds.
Given the FCDC batteries only provide power for one minute and that it takes 90-100 seconds for the APU generator to provide electrical power, it’s unlikely the second condition for the operation of the PFCS normal mode would be satisfied under any circumstances.
@All:
I previously asked for assistance on estimating the probability of whether the Java anomaly was of natural origin or related to MH370. Hearing none and not being a statistician, I tried using AI chat a few months ago and shared the result. It’s clear that AI has rapidly improved since then, so I collaborated on the same question with the latest Claude Sonnet 4.6 LLM. I asked for a probability comparison of the Java event being from MH370 vs a natural geological event. I provided the catalog of all cataloged quakes near the Java coast. I presented the seismic then hydrophone evidence factually. I added that the Java Anomaly site is consistent with the other factual evidence (plus drift and previous search areas), but asked for a focus on the historical seismicity. The result was compelling enough that I prompted for an unbiased report in scientific format for peer review (that’s you guys, or arxiv prepub), inviting Claude AI to be credited. Here’s the unedited product with only a minor revision of the email address.
https://drive.google.com/file/d/1QGlFXlwL7IIwOQM7ZkvCeujtJZQU0L-I
A key quote is: “The probability that natural seismicity reproduces this specific pattern by chance is approximately 1 in 10,000 per day.”
I’m being open about the assistance of an AI, because I’m perceiving that others like Godfrey/Thomas are simply reading the results of their biased AI prompting in their daily vlogs.
If anyone is concerned about bias, I can provide the actual prompts that generated the report.
@Andrew
This is excellent data. I believe that most analyses have used the one minute time for the APU to supply AC power and the one minute SDU warm boot time to place 2FO log-in request at at total of 2 minutes later. While this information changes the timings a bit (and probably 8NM or so in range), it does not meaningfully affect the overall scenario. Amazing that this has not been settled before this.
So purposes of our search, may I take it that you believe that the autopilot loss and the switch to secondary mode takes place immediately on the loss of the second engine? In and “ordinary” flame out incident, there would still be fuel and the APU could run for some time. The crew could reset the autopilot etc. and use whatever procedures were needed to try and start the plane, with the plane in a controlled glide. Not true in this case.
Since the Boeing simulator is the best “ground truth” we have, the analysis and conclusions of my paper are not affected. The simulator data is the basis for asserting that, at least in the sequential flame out situation, the plane was under control of the autopilot until the second engine failed and was therefore wings level and not descending (at least in cases 01, 02, 07 and 09, which are representative of the conditions of a flight that had the fuel to reach a location on the 7th arc).
The uncertainty then is the nature of the impulsive disturbances implied by the transients in the simulated flight data after the 2FO. The data show that the plane was under control and in a stable glide before the strong impulses. Do you have any information on the programming of the turbulence for the simulator?
The “simultaneous” flame out case, based on everyone’s assumption, is not being considered, so I won’t waste time trying to clarify anything except to note that it has a much more chaotic signature at that time.
There was a comment from Andrew that Jean-Marc communicated to me:
Andrew says:
April 9, 2026 at 4:38 am
@Jean-Marc Garot
The report refers to Rib 25 and states that it “appears to be in the vicinity between the inner part of the outer flap and the right engine support pylon, close to the flaperon.”
Rib 25 is actually located further out along the wing, about one-third of the way along the outer flap.
The position of Rib 25 is marked on the following drawing of the left wing (right wing is the same): Rib 25 Position
That puts the failure point that was demonstrated during fatigue testing further out along the wing than suggested.
In fact we do not talk about fatigue failure of the wing but the Ultimate Load fracture of the wing, the most important certification test that happens to all new aircraft.
In my analysis I have used in fact the images of the Boeing test who show reasonably well where the main fracture happened.(https://comp-as-s.com/mh370/impact-sea-ditching/). Even if my choice of where the Rib 25 was incorrect, this is not important because the main fracture in the vicinity of the inner flap and flaperon interface is obvious. In fact the wing showed two fracture initiation areas, which are visible in one of the photos in my website, and the Rib 25 of the drawing of Andrew actually seems to be to the outer fracture zone. However the main fracture occurred further down towards the inner flap / flaperon interface. The actual report text does not say exactly the location of the main fracture, but instead its says “the right wing failed beginning in the area of rib 25 …” The images and video support the final fracture to be not in the area of Andrew’s Rib 25 location but further inwards.
@Sid Bennett said” The simulator data is the basis for asserting that, at least in the sequential flame out situation, the plane was under control of the autopilot until the second engine failed and was therefore wings level and not descending (at least in cases 01, 02, 07 and 09, which are representative of the conditions of a flight that had the fuel to reach a location on the 7th arc).
We’ve already established that with autopilot engaged, the plane descends when the decaying speed reaches just above the stick shaker speed. The departure from wings level flight later occurs when the second engine flames out, the autopilot disengages, and the flight control mode reverts to Secondary.
@Andrew has been very patient and polite about correcting your misconceptions, but I fear that even he has limits.
@Paul Smithson said: Conclusion: latitude discrimination using inferred BFO error at Arc 6 by new method doesn’t add insight. Methods that take account of combined residuals across all arcs provide greater statistical discrimination.
Of course constraining the path with no maneuvers and requiring a statistical match at arcs 2-6 will provide better latitude discrimination. The purpose was to relax constraints, such as the no maneuver requirement, so see if there is a justifiable range of latitudes for searching.
@VictorI
I appreciate the effort that Andrew has put into correcting my lack of detailed knowledge. I will ask no more of him. The paper stands.
I will agree with the statement that the autopilot is present between 1FO and 2FO and is effectively lost at 2FO.
The Boeing 777 engineering simulator data which you obtained from the ATSB and that I analyzed and presented in detail establishes that:
a) In the cases cited, the plane was at its initial cruise altitude at 2FO with the speed having decreased from about M0.8 so as to maintain altitude of FL400.
b) The plane continued on a wings level descent after 2FO for at least 5-10 minutes after 2FO until a strong vertical force was applied to the aircraft by the simulator, when the plane immediately went into a spiral descent.
c) The reason for this upset force is not known but it occurred at a fixed time into the simulation, not with respect to the 2FO. It is believed to be an artifact.
d) Absent the upset force, the plane could have glided up to 135NM along its initial azimuth (Boeing agrees with the glide distance (see Malaysian SIR, Appendix 1.6E, p8) but believe there would have needed to be a pilot. I do not.)
I will say no more on the subject on this blog unless you or @Andrew specifically request a comment.
If anyone asks for my email address, please provide it to them privately.
@Sid Bennett
RE: “So purposes of our search, may I take it that you believe that the autopilot loss and the switch to secondary mode takes place immediately on the loss of the second engine?”
Yes.
RE: “Do you have any information on the programming of the turbulence for the simulator?”
I don’t know much about Boeing’s engineering simulator, except that it is a fixed-base device used for testing and development. It uses the same aerodynamic model as the Level D full-flight simulators used for airline training, but unlike those simulators it does not move to replicate aircraft motion. In the Level D simulators that I have used, the turbulence may be selected at pre-set levels by the instructor; the default setting is ‘off’.
I can’t explain the perturbation you observed in your analysis. If it were turbulence, I would have expected there to be a series of ‘impulsive disturbances’ instead of a single event.
@Andrew
Well, take a look at the plots. I can also provide a package of information for the study to be replicated by others. I am only reporting the information that I have plotted. The amplitude of the impulse is close to an incident of a very severe downdraft.
I have no ideas at the moment how to proceed.
@Sid Bennett
If I’m not mistaken, the change of g-loading during these events was about -0.3 g. Contrary to the claim in your paper, that is categorically NOT a significant event. A change in g-loading of that magnitude would be associated with light turbulence, not a severe downdraft.
See the following for a description of the commonly accepted categories of aviation turbulence:
Australian Bureau of Meteorology Aviation Services – Turbulence
@Sid Bennett
Here’s how I think you should proceed:
Revisit the conclusions in Sect. 9 (Distinguishing 2FO from the Turbulence Perturbation), which affect your entire analysis. What you refer to as ‘2FO’ for the C01, C02, C07 and C09 cases is actually the point where the aircraft reaches minimum speed, where the stick-shaker activates to provide stall warning. At that point the aircraft is still 1FO and the autopilot is still engaged. The autopilot speed protection kicks in and lowers the nose to prevent the speed reducing below minimum speed. The aircraft then descends under autopilot control, with the speed maintained just above minimum speed. The TAS and ground speed decrease as the aircraft descends.
The point where the so-called ‘perturbation’ occurs is actually 2FO. The loss of electrical power causes the PFCS to revert to secondary mode, the autopilot disconnects and the aircraft pitches down due to:
– The loss of thrust.
– The aircraft is out of trim for the current speed. The PFCS envelope protection causes trimming to cease when the speed reaches the minimum manoeuvring speed (MMS) as the aircraft decelerates under autopilot control during the 1FO phase. MMS is above the minimum speed. The out of trim condition causes the aircraft to pitch down as it tries to regain its trimmed speed when the autopilot is lost at 2FO.
2FO appears to have occurred earlier in the C09 case, perhaps because the fuel imbalance at 1FO was less than the other cases in that group. I don’t think it’s possible to say with certainty without knowing the initial conditions for each case.
With thanks to Victor for pointing out the sequence of events.
@Andrew
Thank you and Victor for suggesting further things to think about, and I will do so. I welcome your criticisms and suggestions.
@370location
Thank you for the link to your revised paper co-authored by Claude, which I found very intriguing. My professional career has meandered around various aspects of electroacoustics. Although a couple of decades ago I did provide some advice on suitable equipment for aeroacoustic demonstration of submarine propulsion screw acoustic signatures, none of my experience relates directly to hydroacoustics.
Firstly I don’t understand the numbers used for airframe weights in Table 3. Surely a swamped airframe would weigh in the order of 100,000 kg. Am I missing something?
Secondly in my own experience, like other AI assistants, I have observed that Claude Sonnet 4.6 suffers from sycophantic tendencies and hallucinations which make it’s responses more compelling than they ought to be. I find Claude an incredibly useful tool as long as the bounds of possibility are known beforehand. I am simply suggesting proceed with caution.