Autopilot Stab Trim Authority

High altitude upsets and recoveries are covered in pilot training today, but these type accidents continue to occur, perhaps, for a different reason - Stalls. Yes, we have begun stalling transport category airplanes. And, we have a new culprit - sophisticated (or less-than-sophisticated IMHO) autoflight systems and Flight Management Computers that control all aspects of the airplane’s operation.  They give the pilot whatever he asks for, and, sometimes things he didn’t ask for.  These days they control all the flight controls - even the speed brakes / spoilers - and the engines also.  Even the brakes.  One of the things they do is move the horizontal stab.

Commercial jet flying has been going on for over 50 years and I've been around for most of it - nevertheless,  - let me say right here from the get-go --  I’ve decided I don’t like that scheme.  At least I don’t like the Command Authority the autopilot has over the stab position.  It should be much more limited.  (This is not intended to be a discussion of “cockpit automation”, although that is certainly involved.  That White Paper is still in my future.) (See Note 1 below.)

Having said that, I’d like to point out to those who have started shaking their heads, that an uncomfortable number of recent accidents have been associated with stabs at or near limits that have been driven there by the autoflight systems.  I think that’s just plain wrong.  A stab trimmed to it’s limit is in a place it shouldn’t be, for anything that wants to be described as “normal” flight.  The resultant configuration is a badly mistrimmed airplane that is ripe for an untoward event.  Almost always, in this situation, the  autopilot will wind up getting disengaged - either by its own actions or the pilot’s.  When that happens, the pilot is confronted with a confusing scenario requiring quick thinking - and actions - and an airplane that is going to do some bad things to him in a hurry.  Too many times, the final outcome is fatal.

One more thing - a severely mis-trimmed stab - usually in the nose-up direction, can result in an abrupt and severe pitch-up on disconnect.  With the majority of transport airplanes now having wing-mounted engines, the natural reaction of the pilot, indeed the approved procedures - confronted with rapidly decaying airspeed and a high body attitude and AOA is to rapidly apply power.  In airplanes so configured, this can produce a very large nose up pitching moment of its own, exacerbating an already bad situation.


They should be able to trim within a “normal” flight envelope band - beyond that, the pilot needs to be brought into the loop so as to recognize a developing situation, and not presented with a fait accompli.  My definition of “normal” is very much truncated from what is “available.”   A stab that is at its trim limits should be there ONLY due to pilot command (VERY unlikely!)  Although the trim position is annunciated with assorted dials and warning lites, etc - often the pilot is unaware of the severe mis-trim, and in most cases, fails to rectify it when the autopilot suddenly says “Your airplane!”

Boeing Ops Bulletins advise against using “too much trim.”  They advise this of the pilot community - but fail to address the built in systems that show no such restraint.

With all the smart electronics and software engineers and aero people that are out there these days, logic should be available to tell the autosystems to not try and make the airplane do what the pilot may (seem to) want, because it would place the airplane in an abnormal aerodynamic situation.  Yes!  I know, that sounds a bit like the Airbus "Protect the airplane from the Pilot Philosophy" - but that's NOT what I'm suggesting.  My idea of "abnormal" is having the Autopilot stall the airplane.  That's not abnormal - it's unsafe.  Some examples follow:

MD-82 Accident Venezuela  16 Aug 2005

In the MD-82 accident that precipitated this outburst, the stab was continuously retrimmed by the autopilot until it reached 10.8 units nose up at an AOA of +7.7.  It remained there for the rest of the flight.  The autopilot was trimming nose up to maintain FL330 while the autothrottles were trying to maintain M 0.75.  When the engines couldn’t maintain the speed, it decayed - requiring the autopilot to trim more and more nose up to maintain the altitude.  The speed decayed to M 0.60 and the airplane started descending, unable to maintain the selected altitude.  The pilot then disconnected the autopilot with the severe mistrim now in place.  He never corrected it.   40 seconds later, with the autothrottle still engaged, the stick shaker activated and the stall warning went off.  The airplane entered a deep stall from which it never recovered. 

My position is, pilot aptitude not withstanding, the auto systems were able to place the airplane in a flight regime where it should not be.  That’s wrong.

ANZ A-320 Perpignon, France  27 Nov 2008

The ANZ A320 accident at Perpignon was even more egregious.   Here, an airplane on an acceptance flight test was brought into an approach to stall at about 4000 ft.  In this case,  there were unit failures (AOA vanes) that prevented the Airbus “flight protections” from “doing their thing” - namely protecting the airplane from the pilot - Airbus’ flight control philosophy.  But, despite problems with the AOA sensors, the normal Flight Management Computer actions were available and in use.  (Note:  In a highly automated airplane using FBW systems, it is difficult to differentiate between the FMC and the autopilot.  Disconnecting the autopilot still leaves the pilot flying a computer based system, one that not only votes on the pilot’s “requests”, but actually has the “Final Vote.”  Specifically germane to this paper is the fact that the autopilot trimmed the stab to its nose up limit, where it remained for the remainder of the flight.  The Airbus stall warning activated normally and appropriately, and the pilot initiated a normal stall recovery.  TO/GA power was applied - which exacerbated the pitch up problem.  However, the FMC flight control law had passed to Direct law due to the loss of the Normal law operating conditions. The auto-trim system was thus no longer available.  (The changes of Law that followed did not allow the auto-trim system to move the stab from the nose-up position.)  Manual trim was still available, but the Captain did not react with any input on the trim wheel at any time or to reduce engine thrust in any prolonged manner.  Due to the position of the stabilizer at full pitch-up and the pitch-up moment generated by the engines at maximum thrust, the crew lost control of the aeroplane during the increase in thrust.

At one point, during this flight, the autoflight/FMC system had actually trimmed the stab full nose up, while simultaneously commanding the elevators full nose down.  (The poor pilot at this moment was commanding elevators full nose up  - but he was “out-voted.”)  I suggest, for the purposes of this paper, that somewhere in the S&C and software writers community, there must be a line-of-code lite bulb that goes off and would identify and prevent such a grossly misconfigured airplane condition from occurring.  Again, the authority of the autoflight systems is excessive and allows the airplane to be handed off to the pilot, after having created a very bad situation.

THY 737-800 Amsterdam 25 Feb 2009

The Turkish 737-800 accident at AMS  is yet another example.  The airplane was making a coupled ILS approach using autopilot and autothrottle.   Descending through 1950 ft, the LH radio altimeter malfunctioned and suddenly showed -8 ft, prompting the autothrottles to fully retard the thrust levers to Idle power - where they remained for essentially the remainder of the flight.  The airplane was configured for landing and intercepted the glide path normally at 1330 ft and 169 kts.   With Idle thrust, tracking the glide slope caused a decay in airspeed.  The autopilot  responded with continual nose up pitch commands and trimmed the stab.  At 460 ft, the stick shakers activated, the speed was now 110 kts, the pitch angle was 11 degrees nose-up, and the AOA was +20.  By 420 ft, the autothrottle was disengaged followed by the autopilot.  Although the pilot attempted to recover, there was insufficient altitude to effect  a successful recovery.  The aircraft struck the ground at low speed with a high sink rate and 22 degrees nose up attitude.  The accident report does not specify the stab setting except to say the jackscrew was found in a high nose-up trim position.

Again, pilot proficiency and malfunctioning RA notwithstanding, the autoflight systems had blithely brought the airplane into a stall situation with a grossly mistrimmed configuration.  Actually, when the stick shaker activated, the co-pilot pushed the throttles (and control column) ahead, but the autothrottle - still being engaged at that time - immediately retarded them again to Idle.  Bad business in my book.

Thompsonfly 737-300 Bournemouth 23 Sep 2007

An almost identical near-accident event occurred on 23 Sep 2007 to a Thompsonfly 737-300 which stalled on approach to Bournemouth, UK.  This airplane was also performing a coupled ILS approach.  When the Capt selected a reduced speed, the autothrottle reduced the thrust to Idle, and then disconnected.  The disconnect went unnoticed by the crew.  The autopilot tracked the glideslope accurately, and with only Idle thrust, adjusted the aircraft pitch upwards while trimming the stab accordingly.  Airspeed continued to decay down to 110 kts at 1540 ft altitude, when the stick shaker activated at an aircraft pitch attitude of +12 degrees and the stab at +7.9 units nose up.

The Capt moved the throttles full forward and pitched the nose down to +5 degrees to counteract the expected pitch-up from power.  The airspeed decayed to 101 kts.  Four seconds after applying full power, the autopilot - engaged all the while - disengaged, and the nose pitched up (because of the large amount of nose-up trim from the stab.)  Although the pilot had moved the control column fully forward, the nose continued to pitch up, reaching 22 degrees nose-up. (Stab trim over-powered the available elevator authority.)   Both engines were now producing substantially more than rated go-around power.  The co-pilot reselected flaps from 40 to 15 (which would adversely affect stalling speed and reduce the flaps’ nose-down pitching moment.)  The nose continued to pitch up passing +27 degrees, while both pilots applied full forward on their control columns.  They reported “they had no pitch control authority.”  The stick shaker was going, CAS fell below 107 kts, and a roll to the left developed.

Aircraft pitch passed 36 degrees nose up with a 22 degree left wing down roll.  Pitch finally reached 44 degrees nose up and airspeed fell to 82 kts.  At this point, and with no change in the full nose down elevator position (this not being a T-Tail aircraft), the aircraft stalled, the nose fell through (normally - unlike the A320) and pitch rate went from positive to negative.  The aircraft pitch fell to +20 degrees over 10 seconds, and fell a further 15 degrees (to +5 degrees) over the next 2 seconds, while the airspeed increased rapidly.  The Capt then stabilized the airplane at 3000 ft with a 5 degree nose-up attitude and 147 kts airspeed.  The airplane was then circled around and landed normally - there was no damage and no injuries to the 137 occupants - a happier outcome than the previous three incidents.

There were a number of “culprits” in this incident - such as failing to notice the autothrottle disconnect, or adequately monitor the decaying airspeed.  However, from the airplane side - there were no malfunctions and everything operated “normally.”  The fact is, the autoflight  systems ALLOWED this situation to develop, and possessed the authority to do so.  The stab was trimmed excessively (my opinion) by the autopilot to maintain vertical flight path on the glideslope, and resulted in a badly mistrimmed airplane that pitched up radically when the autopilot was disconnected, and go-around power applied.  The mistrimmed stab was sufficient to easily overcome full down elevator.

Actually - in common with the three accidents described above - the stab was not manually retrimmed by the crew until the entire event was over.  They were lucky in that they had more altitude to play with than the Turkish  airplane, (but less than half that of the ANZ A-320) and that the airplane had the benign stall characteristics that allowed it to go into a full stall with accompanying nose break and be recovered conventionally with minimal loss of altitude (about 600 ft.)


Well - here you have four recent events that make my case - I could add more, or you can take my word for it.  Autoflight systems have too much authority - especially over the horizontal stabilizer.  Stabs on (most) modern jetliners are large and powerful.  Trimming them to large, mostly airplane nose-up, positions creates a poorly configured airplane and sets up the flight crew for a bad “gotcha" when the autopilot is disconnected.  Very few human pilots would think of trimming the stab to such large values, and almost certainly would never wind up with a hand flown situation like the automated ANZ A-320.  (Note in the NW 727 probe icing accident - while the crew maintained up elevator throughout the flight - the stab remained in a near neutral position.) 

All four events I described above involved badly mis-trimmed stabs and autoflight systems that transferred control to the crew after having badly botched the airplane’s situation.  Yes, it can be easily argued that the crews in each case needed to have done a better job monitoring what the automation was doing.  I assert that, in addition to humans being lousy “systems monitors,” which is well-known,  the autoflight systems created the problem in a very insidious manner.  We used to be concerned about stab runaways, and installed stab trim lockout brakes.  Now we have stabs that “slowly” runaway, and trim to extreme positions, not only with our being  mostly unaware, but with our blessing.

Autoflight systems also need to have a “does this make sense” logic that will kick them off when they find they are commanding the airplane to be doing something outside its normal envelope. Autopilot needs to talk more to Autothrottle - or at least more than I think it is talking.  (If Autothrottle is disconnected - and power is at Idle - Autopilot should not be trying to maintain an altitude or track the glideslope without regard to airspeed.  Does the A/P even know what the airspeed is?  If so, how can it run the airplane down into stick shaker???  And below.)

Autopilot Stab trim should be limited, and this limit should act to reinsert the pilot in the control loop early - in time to analyze what’s going on and take corrective action.   In no case should the automation be able to stall a perfectly good airplane, which is exactly what happened in the above examples.  Automated flight path control is supposed to make flying safer; none of the above events would have occurred in a hand flown situation.  Automation, at least indirectly, caused these accidents.

Stalling large jet transport airplanes is best left to the manufacturer’s test pilots - not to line pilots - and especially not in the middle of the night.

Today’s systems do our bidding - no matter how contradictory it might be.  Like personal computers, they are more dumb than smart.  So, we have stabs at one extreme while the elevators are at the opposite - all in their electronic quest to show us how nubile they are at doing our bidding, and how wondrous are those software engineers who write the lines of code making it all happen.

NOT!.... I say.  We can do better.  We NEED to do better.

Just - My two cents.  Feel free to tell me I'm AFU.

Bob Bogash
August 2010

An interesting article I ran across while  researching this epistle speaks to stab trim position “awareness”.  I used to think the 707 / 727 /737 airplanes with their manual trim wheels made autopilot trimming obvious - unfortunately, not so.

Note:  Autopilot misconfiguration of the airplane is a recognized problem in the turboprop regional airliner world, where several accidents related to airframe icing have occurred.  The problem is the autopilot masks the handling characteristic changes that are occurring while a dangerous ice build-up occurs.  When the autopilot is disconnected, the pilot suddenly discovers he has control of an airplane with badly impaired flying qualities, pitch-up, and loss of control.  To rectify this problem, the recommendation is to hand fly the airplane during icing conditions.  Apparently, there’s still a place for the human pilot and his often superior skillset.

Note 1 - Cockpit Automation - Canadian Report on wake turbulence encounter AC A319 and UA 747-400  A08W0007

Out-of-the-loop performance problems are well documented as a potential negative effect of automation.  Under certain circumstances, such as when an automatic system is managing an unusual event but there is no direct indication of this action provided to the pilot, it is possible for the pilot to misdiagnose the emergency as an automation failure. This can result in the pilot responding to a presumed automation failure rather than to the actual emergency. In essence, the understanding that the operator has developed of the aircraft and its environment is incorrect because the pilot has not been directly interacting with the aircraft controls. The more complex the situation, the less capable the human is of entering the loop when an emergency occurs. As a result, being outside the control loop can result in adverse aircraft-pilot coupling as a pilot responds to a sudden emergency.

When startled by a sudden unexpected event, a pilot is susceptible to delayed reactions, which are based on previous training and experience. This may lead to making inappropriate control inputs for the conditions at hand.

Note 2:

About 7 months after I wrote the above, the Flight Safety Foundation held a conference, reported in Aviation Week,  stating that loss of control, many involving stalls, had replaced CFIT as the Number One cause of airline accidents.  "“The same thing happens over and over again,” Michael Coker, a senior safety pilot at Boeing, agreed."

Their proposed solution:  Better Pilot Training.  Perhaps revised procedures (" the industry has for decades been teaching the wrong recovery procedures.")

My reaction?  A big MAYBE........  Sounds like the samo-samo - "It's the pilot, stupid."  I have always felt it is poor policy to teach humans to adapt to the quirks of machines.  Rather, the machines should be designed to not entrap the dumb human.  In any event, at least there has finally been recognition that we have been regularly stalling and crashing large jetliners - a rather astounding situation IMHO.

Some Feedback

From one correspondent:

This is probably simplistic. But whenever a class of accidents punches through the noise level I would assume the company is interested. Boeing used to lead these sort of topics, even to the point of involving our competitors. In some cases the result was a "Training Aid". In some cases we revised design. In some cases we did both; for example wind shear. 

In one area Bob and I strongly agree that although crew error is most assuredly a factor in both types of accidents, the airplane's failure tolerance to crew errors of known classifications is also a factor. 

I'll mention one more factor on Bob's investigation: this type of accident is not limited to airplanes with autothrottles installed and operating. GA airplanes are susceptible to the autopilot-induced-stall as are some smaller transport category airplanes with no autothrottle installed. 

Hi, Bob

You've done a great deal of thought on this. I do agree with a number of your comments, but I'd like to add a few points at least in partial defense of the company. Many of the flight control and autoflight designs you refer to are or are derivatives of aircraft that go back 40-45 years - 737, 747, DC-9 and MD-80 families of transports. Along the way, changes were made to address many of the issues you raise. Progress is sometimes slow for sure but the intent and really good work of some very talented engineers has made significant progress starting nearly 20-30 years ago up until the current generation.

In the early 1980's the company help develop the SP-177 autopilot for the 737. This was the first step to address some of the issues you raise. Both the auto throttle and autopilot had minimum speed logic that would come into play in some of the slow speed/high AOA regimes. If in level flight or on an ILS glide slope the auto throttle would not slow the airplane below a minimum speed if it was engaged. The auto pilot would revert to flight level change (airspeed control with pitch) at minimum speed if asked to perform too high a vertical speed or flight path angle. Obviously, in this generation if neither system is in use the pilot can force the aircraft into a stall. Also, as I recall one of the concerns at that time was unnoticed departures from altitude hold or glide slope by automatic systems, so no logic was put in that would address those events. This concern may have been related to the 1972 L-1011 Eastern 401 crash into the Everglades when the disconnection of the autopilot went unnoticed while the crew worked on a gear problem. The 737NG had a few more enhancements that made it a bit harder for the pilot to force the airplane into a stall as well.

Twenty one years ago, when we started development of the 777, we set out to address more of the issues you have raised. As you have pointed out, it is necessary to have the auto throttle, flight controls and autopilot integrated. This is no small feat and would be extremely difficult on a retrofit basis, consequently such measures need to be built into the original design. When the 777 flew 17 years ago we had an airplane that could not be trimmed into the stall either by the autopilot or the human pilot. He could stall it if he wanted to by holding fairly high stick forces, which would allow him to deal with erroneous AOA vane inputs if he had to. If the auto throttle was armed, but not engaged it would engage itself and attempt to hold a minimum safe speed. If the auto throttle were off (not armed) the autopilot would leave altitude hold at a minimum speed (near stick shaker) to protect against stall (with the airplane still trimmed at an even higher maneuver speed if the pilot disengaged the A/P). The latter design decision logic was pretty simple - i.e. if the airplane really stalls it is not going to hold altitude anyway. The new 787 has these same features plus several more enhancements that make the airplane extremely hard to stall at all.

Design for being too slow on approach is very difficult beyond the basics of good auto throttle logic and alerting. Allowing the autopilot to depart the glide slope on its own analysis of speed/AOA or nose up trim limits can lead to safety issues as well. Major wind gusts and shears can cause momentary excursions into the stick shaker range and generally it is best to just hang on or go around. Added to those concerns is the fact that jet transports have very large stabilizer trim ranges due to having a large gross weight/ cg range that must be accommodated. Logic to limit the ability to "trim into a stall" or limit the the combined pitch effects of stabilizer and high thrust requires a sophisticated control system logic that would be very hard to retrofit.

I guess what I'm trying to point out is that a lot has been done over the past 45 years even though some of the initial designs are still in service. This work, by both Boeing and Airbus, does address many of the issues you've raised. Neither company has ignored these issues. Boeing has continued to lead both technical design and in development and distribution of training materials, and numerous presentations within the industry for both airline operational and flight test safety. The release of the the "Large Airplane Upset Training Aid" in 1998 (revised 2004 and 2008) was a joint effort of Boeing, Airbus and the Flight Safety Foundation. It is an excellent body of work and, if incorporated and followed as part of an airline's training program, would address many of the concerns in operating earlier models of non-protected aircraft. The Av Week article you mention is part of the effort to get the training industry and FAA Flight Standards to incorporate the lessons of the training aid. Of particular importance to your concerns are the concepts that point out the differences between "stall recovery" and "approach to stall recovery". For many years trainers and FAA standards wanted a precision flight maneuver and insisted on the minimum loss of altitude during the "approach to stall recovery". The big problem was that many pilots were thinking this was the method for a full stall recovery - and some programs,even labeled the training "stall recovery". Undoing these planted memories and misunderstandings has taken a great deal more time than any of us thought but at least it is getting wider exposure today. As I understand it the training aid has been amended and adopted for smaller business, commuter and general aviation. Finally the FAA is about to release an Advisory Circular that will change the entire focus of teaching stalls from the introduction for student pilots to jet transport pilots. Boeing played a significant role in this effort as did Airbus. While this approach is pilot-centric it is one of the few methods available to address aircraft qualities that do not have initial design envelope protection features and to address misconceptions that have gotten some pilots into trouble for many years in all types of aircraft and situations.

Thanks for listening,

Senior Boeing Test Pilot

18 Mar 2011

Some more info

Dynon Skyview Autopilot System

Dynon Avionics in Woodinville, Washington, produce a sophisticated Avionics package for Experimental General Aviation airplanes that contains a pair of capable Autpilots - a Simplified one and an Expert one - pilot selectable.

I fly that equipment in my Vans RV-12 airplane, using the sophisticated version of the A/P.  One of its features is flight envelope protection, similar to that discussed above.  Although Skyview airplanes do not have Autothrottles. they can experience commands to maintain an Altitude or Track a Glide Slope that would potentially cause the airplane to lose enough Airspeed to result in a Stall.  However, the Skyview will not permit that, and, if going too fast or too slow, will invoke pitch commands to speed up or slow down the airplane appropriately.

Related Accidents since this page was written

Asiana 777 at San Francisco

This aircraft struck a seawall at low airspeed and high nose up body angle just short of the runway at SFO.

From the NTSB Synopsis:

The flight was vectored for a visual approach to runway 28L and intercepted the final approach course about 14 nautical miles (nm) from the threshold at an altitude slightly above the desired 3° glidepath. This set the flight crew up for a straight-in visual approach; however, after the flight crew accepted an air traffic control instruction to maintain 180 knots to 5 nm from the runway, the flight crew mismanaged the airplane’s descent, which resulted in the airplane being well above the desired 3° glidepath when it reached the 5 nm point. The flight crew’s difficulty in managing the airplane’s descent continued as the approach continued. In an attempt to increase the airplane’s descent rate and capture the desired glidepath, the pilot flying (PF) selected an autopilot (A/P) mode (flight level change speed [FLCH SPD]) that instead resulted in the autoflight system initiating a climb because the airplane was below the selected altitude. The PF disconnected the A/P and moved the thrust levers to idle, which caused the autothrottle (A/T) to change to the HOLD mode, a mode in which the A/T does not control airspeed. The PF then pitched the airplane down and increased the descent rate. Neither the PF, the pilot monitoring (PM), nor the observer noted the change in A/T mode to HOLD.

As the airplane reached 500 ft above airport elevation, the point at which Asiana’s procedures dictated that the approach must be stabilized, the precision approach path indicator (PAPI) would have shown the flight crew that the airplane was slightly above the desired glidepath. Also, the airspeed, which had been decreasing rapidly, had just reached the proper approach speed of 137 knots. However, the thrust levers were still at idle, and the descent rate was about 1,200 ft per minute, well above the descent rate of about 700 fpm needed to maintain the desired glidepath; these were two indications that the approach was not stabilized. Based on these two indications, the flight crew should have determined that the approach was unstabilized and initiated a go-around, but they did not do so. As the approach continued, it became increasingly unstabilized as the airplane descended below the desired glidepath; the PAPI displayed three and then four red lights, indicating the continuing descent below the glidepath. The decreasing trend in airspeed continued, and about 200 ft, the flight crew became aware of the low airspeed and low path conditions but did not initiate a go-around until the airplane was below 100 ft, at which point the airplane did not have the performance capability to accomplish a go-around. The flight crew’s insufficient monitoring of airspeed indications during the approach resulted from expectancy, increased workload, fatigue, and automation reliance.

The safety issues discussed in the report relate to the need for the following:
  • Adherence of Asiana pilots to standard operating procedures (SOP) regarding callouts. The flight crew did not consistently adhere to Asiana’s SOPs involving selections and callouts pertaining to the autoflight system’s mode control panel. This lack of adherence is likely the reason that the PF did not call out “flight level change” when he selected FLCH SPD. As a result, and because the PM’s attention was likely on changing the flap setting at that time, the PM did not notice that FLCH SPD was engaged.
  • Reduced design complexity and enhanced training on the airplane’s autoflight system. The PF had an inaccurate understanding of how the Boeing 777 A/P and A/T systems interact to control airspeed in FLCH SPD mode, what happens when the A/T is overridden and the throttles transition to HOLD in a FLCH SPD descent, and how the A/T automatic engagement feature operates. The PF’s faulty mental model of the airplane’s automation logic led to his inadvertent deactivation of automatic airspeed control. Both reduced design complexity and improved systems training can help reduce the type of error made by the PF.
  • A context-dependent low energy alert. The airplane was equipped with a low airspeed alerting system that was designed to alert flight crews to low airspeed in the cruise phase of flight for the purpose of stall avoidance. However, this accident demonstrates that existing low-airspeed alert systems that are designed to provide pilots with redundant aural and visual warning of impending hazardous low-airspeed conditions may be ineffective when they are developed for one phase of flight (i.e., cruise) and are not adequately tailored to reflect conditions that may be important in another phase of flight (e.g., approach). During the approach phase of flight, a low airspeed alert may need to be designed so that its activation threshold takes airspeed (kinetic energy), altitude (potential energy), and engine response time into account.

Swiftair MD-83 - Mali - 24 July 2014

This aircraft crashed after an upset and descent from cruising altitude in which  the Autopilot and Autothrottle were involved in changes in power settings and altitude.    The airspeed decayed significantly  to 162 kts and the engines experienced large fluctuations in power.  However, this airplane did not enter a "deep stall" but rather upset into a spiral dive, striking the ground at high speed in a 58 degree nose-down attitude.  This investigation is in the early stages - this information was released in a preliminary Interim Report on 20 Sept 2014.

23 minutes after takeoff the aircraft levelled off at FL310 at 0.74 mach and was turning right to return onto the original track. The aircraft accelerated with engine EPR at 1.92, the aircraft reached 0.775 mach, then the airspeed began to decrease with the autothrottle still in Mach mode. About one minute later, at mach 0.752, the autothrottle changed to MACH ATL mode, pitch and EPR increased, altitude was stable, engine N1 remained stable, but speed continued to decrease.

The flight data recorder revealed that EPR and N1 fluctuations occurred on both engines for about 45 seconds, then over the next 25 seconds the EPR increased from 1.6 to 2.5 and decreased again twice, N1 varied between 70 and 91 percent, roll oscillations 4 degrees to the left and right occurred, autothrottle disengaged between the EPR oscillations 1.6 and 2.5.

About 65 seconds after the EPR/N1 oscillations began the airspeed had decayed to 203 KIAS/0.561 mach and the aircraft began to descend,  - 29 seconds later the autopilot disengaged, at that time the aircraft had lost 1150 feet of altitude, airspeed was 162 KIAS, 0.439 mach and both engines were nearly at idle.

Subsequently pitch and bank angles were subject to large changes, the aircraft's nose pitched progressively down to 80 degrees nose down, the aircraft rolled to 140 degrees left bank. About 20 seconds prior to impact both engines accelerated and reached near maximum thrust. The last flight data recorder data indicated the aircraft at 1601 feet (QNH1013), 384 KIAS, 58 degrees nose down, 10 degrees left bank.

".....the deflection of the elevators and the position of the trimmable horizontal stabiliser continued pitching up."

Revised 10 Sept 2010

Revised 10 Jun 2011

Revised 24 Sept 2014