Malaysian flight MH370 went missing on March 8, 2014 over the Southern Indian Ocean. The timing of several satellite communications signal pings provided good accuracy for the distance to the plane from the satellite. Residual Doppler frequency shift analysis suggested headings during the flight, and on the final ping may have indicated that the plane went from level flight to a steep dive within 8 seconds. The attitude of the plane in its final moments is a significant factor in the potential size of the estimated search area. The assumption of a steep dive kept the previous search areas close to the last 7th Arc ping radius. Analysis of found debris along with the doppler data has led to diverse perspectives from different investigators on the possible glide distance and speed of impact. This report presents a new analysis showing that the SATCOM signal strength is at a maximum when the pitch of the plane is near zero degrees for level flight, and that was the condition for the final hours including the last ping. Level flight at cruise altitude might imply a long glide favored by some, but could rather indicate a level stall at the end with rapid vertical drop. Such a stall is compatible with various debris reports showing both retracted flaps on impact, and trailing edge damage to multiple control surfaces from water entry.
INMARSAT engineers began keeping records of additional aircraft satellite communication (SATCOM) parameters after the tragic loss of Air France flight AF447 in the Atlantic in 2009, which fell from high altitude in a stall due to iced pitot tube airspeed sensors that confused both the autopilot and crew. For MH370, INMARSAT recruited multiple international teams to analyze the Burst Transfer Offset (BTO) time delay from a plane’s transmit timeslot to the ground station, giving a fairly accurate estimate of the distance from aircraft to satellite. They also analyzed the Burst Frequency Offset (BFO) of dozens of flights to help determine whether MH370 went north over Asia or south over the Indian Ocean.
The BFO is not a direct measure of the Doppler shift, because the SATCOM unit on the aircraft is adjusting its transmit frequency to compensate for that Doppler shift to match the assigned receive frequency at the ground station. The SATCOM unit uses live inertial navigation and GPS track info to cancel the Doppler. Its software algorithm estimates the correction from track speed and heading, but uses a set geostationary coordinate for the satellite, and ignores vertical speed. It does not take into account the orbital shift of the satellite, which is significant since it no longer maintains station due to lack of propellant. The rate of climb is not factored because in normal flight the frequency shift is relatively small. It is the residual BFO value differences that carry some information about heading and vertical climb during the flight.
Likewise, the SATCOM unit also manages the aiming of the high gain antennas (HGA) toward the satellite to normalize the received signal strength. Besides a low gain antenna available to the SATCOM for maintaining communications, there are two high gain antenna arrays on the left and right side of the aircraft behind the wing and above the third exit doors. These are phased array antennas that can be steered in both azimuth and elevation toward the satellite. The SATCOM unit utilizes the same inertial navigation data to aim the antenna toward the satellite, and also uses known characteristics of the antenna gain pattern to modify the transmit power for nominal signal strength at the receive end.
Similar to uncompensated vertical speed in correcting the BFO, it appears that the SATCOM unit was not correcting for pitch attitude when normalizing antenna gain. This may have been from the same premise that there were only brief periods during takeoff/climb and landing during normal flight where the pitch would be beyond a few degrees from level flight. The resulting variations in signal strength are the focus of this report.
The same plane 9M-MRO flew hours before as flight MH371 from Beijing to Kuala Lumpur, and was returning as flight MH370 back to Beijing. Flight track data made public for analysis includes ACARS (Aircraft Communications Addressing and Reporting System) telemetry from MH371 for comparison, plus the normal portion of flight MH370. ADSB (Automatic Dependent Surveillance–Broadcast) 1090 MHz location transmissions were picked up during the early portion of flight MH370, and may be available for MH371. Analyzed here are the SATCOM layer ping data provided by INMARSAT that normally carry ACARS, phone calls, engine reports, and other services.
All of the SATCOM data presented here was relayed via the INMARSAT Indian Ocean Region IOR-F4 satellite positioned in geostationary orbit over the equator with a sub-zenith point near longitude 64.5 degrees. Flight MH371 also relayed through a satellite over the Pacific ocean with different signal characteristics, but those data points are excluded. The uncorrected motion of the satellite is mostly north-south, reaching a maximum latitude of about 1.5 degrees N at around 1930Z during flight MH370, and 1.5 S at 0730Z during MH371. This directional error is relatively small compared to the broader antenna gain pattern lobes, but it is within the range of pitch attitudes shown to influence the signal strength, and may cause a cyclical daily shift in the optimal pitch.
Pitch Attitude and Signal Strength
The chart in Figure 1 plots signal strength over nearly 24 hours with various data known for each sample point. The online plot saved (April 13, 2022) in html file format is interactive, generated using an open source python module provided by Plotly. Hovering the cursor over a datapoint gives a popup window with additional info. Zoom and Pan options allow close examination of neighboring points that may be just seconds apart. The chart can also be decluttered by clicking on the legend to select the content.
Figure 1. MH371 and MH370 Signal Strength plot (expand 220413-0857-9MMRO-SNR-plot.html)
The signal strength values are in three groups by type of communication, slower 600 bps channel transmissions had signal strengths typically 1.1x higher than the 1200 bps channels, except for one reversed burst 15 min before takeoff from Beijing. The C-channel telecom signals were much stronger, but there were no MH371 C-channel values for comparison. Both C-channel signal bursts at 18:40Z and 23:14Z had a unique matching pattern of a linear 10 dB decline in strength over 60 seconds. Flight speed, track heading, and flight level (1/100 of altitude in ft) are plotted when known.
The signal strength value used here is derived from two separate recorded measures of signal quality, the Rx (receive) Power in dBm, and the Carrier to Noise ratio (C/No). Figure 2 shows that there is a linear relationship between the two values, with noise variation on both. (In researching the topic for this report, a similar graph was first published in a concise June 12, 2017 analysis of MH370 SATCOM by Mike Exner.) The highest noise component is occasional drops in the C/No value. This could be caused by inadvertent radio interference, possibly two planes attempting to use the same timeslot. Scaling the two measures to match and averaging for signal strength still has a large variance between neighboring values. The real interest is in the available SATCOM transmit antenna gain, and it is found that taking the maximum of each scaled sample pair as the signal strength produces much less variance in the Figure 1 time plot.
Figure 2. Rx Power has a linear relation to Carrier Noise level (expand 220413-0857-9MMRO-RXPower-vs-SNR-plot.html)
A direct measure of pitch attitude is not available for analysis, but would be part of the inertial navigation data sent to the SATCOM unit and saved on Flight Data Recorders. Pitch might also be estimated using a flight simulator, but for this analysis it can be approximately inferred from the changes in altitude. Roll attitude or bank can also be deduced from the track heading changes, and appears to have been correctly compensated when aiming the high gain antenna. The yaw heading also appears to be correctly compensated, because there are no significant signal strength shifts before/after turns, or during taxi.
It is of interest to note that signal levels were relatively low while the plane was on the ground. Photogrammetric analysis from various images of 9M-MRO and other Boeing B777-200 planes show the pitch of the fuselage is approximately -2.5 degrees when the plane is on the ground, with the nose slightly lower than the tail. The lowest signal strength values came at the time when the reverse thrusters were activated on MH371 landing at KL, and after the plane had just leveled off from a climb to cruise altitude, where higher thrust may have induced a negative pitch.
The highest MH370 1200-ch signal level was at the 22:41Z 5th Arc, (when the plane may have been flying between Cocos Island and Christmas Island), and most important is that the final pings were the next highest strength of the entire flight. If signal strength is related to pitch, this reveals that MH370 was in level fight every time it checked in, including the final pings. Those signal strength levels are comparable to the highest values from the plane when it was flying level as MH371 at cruise altitude from Beijing to KL.
Refinements to this analysis might come from confirmation that the Honeywell software does not incorporate pitch attitude into the phased array antenna aiming or transmit power calculation. Validation could also come from comparisons of statistical flight recorder pitch values vs SATCOM signal strength for any B777 dataset.
The BFO value of the 00:19:37Z final ping has been variously interpreted to be part of an un-piloted phugoid (climb/descent cycle), a rapid spiral descent (per simulations), or a hypersonic (faster than Mach 1) vertical dive with flutter that first destroys trailing edges of control surfaces then rips them off the plane. During portions of a vertical or spiral dive, it is likely that the belly of the plane might be facing toward the satellite, where all of the SATCOM antennas would then be entirely hidden with zero signal strength. The BFO was nominal just 8 seconds before at 00:19:29Z. If the relation between pitch and signal strength can be validated, it will mean that 9M-MRO could not have suddenly gone into a high-G steep dive, and the BFO shift had some other cause. Even a track change that quickly seems unlikely.
The BFO of the 18:40Z unanswered phone call pings has also been interpreted in the past as either a heading change south or a rapid vertical descent. The signal strength and pattern is nearly identical to the unanswered 23:14Z call, which might also imply that the plane was flying at the same level pitch both times, with the 18:40Z BFO indicating a new track heading.
Given this new evidence that slight deviations from level pitch appear to reduce signal strength, it is unlikely that under more extreme attitudes the SATCOM software would be attempting to compensate by correctly aiming the High Gain Antennas. There would have instead been a switch over to the Low Gain Antenna (LGA) with higher transmit power.
The plane may have been piloted if it remained level at 00:11Z after each engine ran out of fuel, which would have caused asymmetric thrust and cancellation of an autopilot mode. For candidate sites where the plane was assumed to be in Long Range Cruise at high altitude until fuel exhaustion, the consequence is that the plane may have continued a piloted glide as far as 240 km (over 2 degrees of latitude) after the last ping, potentially leading to a very large search area. Candidate sites that assumed low altitude flight would have a smaller search radius. A finding of level flight at the end does not change the search area for the Java acoustic candidate site, which is based on the comparatively precise seismic epicenter of a heavy section of the plane hitting the seabed.
There have been different attempts to explain the trailing edge damage found on recovered control surfaces like the flaperon, yet lack of damage on the leading edges and upper/lower surfaces. A theory of high speed flutter was presented within a day of the flaperon being found. This fit with the assumption of a high speed dive while suggesting that aerodynamic flutter at near supersonic speed caused the trailing edge damage and also tore the flaperon from the plane to give it a soft impact, since a high speed impact would crush or shatter all components of the aircraft. Similar trailing edge damage to multiple control surfaces found later makes that scenario less likely, but the high speed dive remains a well supported theory, partly based on evidence that the flaps were retracted. Other investigators have modeled the damage as from a low speed ditching with control surfaces still attached to the plane, but that scenario implies a piloted entry which would presumably have flaps deployed.
Combined with past findings that the flaps were retracted, level pitch at the end introduces the possibility of the plane falling in a low speed level stall. Air France AF447 crashed with nose-up pitch in a relatively slow stall, with track speed near landing speed, but a high descent rate nearly as fast as the forward speed. Such an upward force angle on a level MH370 water entry could account for the trailing edge damage, undamaged upper surfaces, and breakup of the fuselage to account for pieces of the interior cabin being found. Such a low speed impact may also account for why the 7th Arc acoustic event near Java was 55 minutes after the last ping. Some portion of the fuselage may have either sunk slowly at first, or stayed afloat for up to 45 minutes if it then sunk rapidly. Also matching trailing edge evidence would be a scenario where MH370 was level at the time of the last ping before piloted ditching, and shortly afterward hit the water tail first in a steep nose-up stall to minimize speed and cockpit damage.
We know that MH370 appears to have navigated between waypoints before it left radar, then kept on flying for hours. We can only speculate whether this was due to sensationalized malicious intent, or actual damage to the plane. A complex electrical fault may have occurred, with redundancy allowing continued flight, but damaging comms and any of various complex systems required to safely land. A prolonged flight away from populated areas and heading towards daylight may have given the crew their best chance to attempt repairs and land. It could explain the partial restoration of SATCOM an hour after comms were lost, which has given us key evidence. If the flight was at low altitude holding speed for maximum duration and (as acoustics indicate) flew by the only two island airports in the Indian Ocean on a path towards daylight, it may remove the assumption of an un-piloted flight to oblivion that underlies accusations without evidence against the crew.
A finding that MH370 was in level flight pitch at key moments does not directly tell us the cause of the disappearance, but it does help determine the accuracy of proposed candidate sites for a continued search. Only by finding the crash site of the plane and the cause of the tragedy can speculation come to an end, helping us all toward finding closure.