Turkish Airliner Crash

[empiredude]

here is what I (think I) know: an airplane uses the air pressure to measure the altitude - above 10000feet the crew sets the altimeter to a default pressure (QNH) (1013hPa in europe I think) - below 10000feet the altimeter should be set to the local air pressure (information being received from ATC during approach). could be that the crew set their altimeter incorrectly - causing the aircraft to display a wrong altitude (ref. sea level).
however, if the airplane (at least airbus do that) gets below something like 2000feet - you also get the distance between ground and plane measured by a radar instrument. that one only displays wrong information if the system itself is broken...
if the aircraft gets below something like 100feet - there is either a hearble countdown (100fett / 50feet / 20feet / 10feet) or a "terrain" warning (depending on whether the airplance is "supposed" to be landing or not)
either way,
it's still strange three pilots in the cockpit did not notice the wrong altitude despite all the warnings, etc...

(if I just wrote a total incorrect nonsense, please correct my statements...)

[MathFox]

What I gathered from news reports is that the autothrottle system reacted as if the plane already had landed:  -8 ft radio altitude -> wheels on the ground -> no danger in stalling.

[andreblt]

The737-800 can do an "Automatic Flare/single channel flare" due to it's autopilot design.
Even with A/P B engaged, the A/T recieves its inputs from FCC A and respective "A" radio altimeter, even if "B" channel is engaged. So if the A/T thinks it landing due to the erroneus RA input and retards the P/L's.
I find it quite believable that any or all of the crew could have missed RETARD / FLARE or even WAKE-UP annunciated on the FMA as this is a very common omission even among highly experienced crew.
What I find unbelievable is that 3 pairs of eyes could have failed to notice a loss of 40kts speed from 1950 - 450 ft along with 3 pairs of ears failing to notice the reduction in airflow noise, along with 3 asses that didn't feel the change in body angle required to stay on the glidepath.
The final error "appears" to have been the other all too common fault of commencing a G/A but not pushing TOGA, therefore no useful F/D commands & no automatic movement towards & maintaining of G/A thrust. It appears that when the F/O relinquished control the Capt grabbed the controls with both hands but no-one was guarding the thrust levers.
Still, having said all this, I find it unbelievable that such an experienced crew could have allowed the situation to get this far. If this is true and there is nothing else nasty or dramatic involved, it really was a total loss of situational awareness on the part of 3 crew at a critical stage of flight.
Very strange & scary indeed.

[iskyfly]

In the initial report there is this paragraph:

"The voice recorder has shown that the crew were notified that the left radio altimeter was not
working correctly (via the warning signal “landing gear must go down”).
Provisional data indicates that this signal was not regarded to be a problem."

I imagine there will be some discussion about whether the warning signal "landing gear must go down" is indeed an adequate warning in the case of what is a fatal autopilot malfunction.

I believe the "landing gear must go down" is a translation error.

Anyway, I just would not have believed it to be possible for an automatic landing system to actually stall the aircraft like this and allow it to drop from the sky a whole km short of the runway!

I find it hard to believe that the pilots did not recognize the signs.

Fly the aircraft first.

[iskyfly]

FROM: THE BOEING COMPANY
TO: MOM [MESSAGE NUMBER:MOM-MOM-09-0063-01B] 04-Mar-2009 05:29:01 AM US PACIFIC TIME
Multi Operator Message

This message is sent to all 737-100,-200,-300,-400,-500,-600,-700,-800,-900,-BBJ customers and to respective Boeing Field Service bases, Regional Directors, the Air Transport Association, International Air Transport Association, and Airline Resident Representatives.

SERVICE REQUEST ID: 1-1228079803
ACCOUNT: Boeing Correspondence (MOM)
DUE DATE: 10-Mar-2009
PRODUCT TYPE: Airplane
PRODUCT LINE: 737
PRODUCT: 737-100,-200,-300,-400,-500,-600,-700,-800,-900,-BBJ
ATA: 3400-00

SUBJECT: 737-800 TC-JGE Accident at Schiphol Airport, Amsterdam - 25 February 2009

REFERENCES:
/A/ 1-1222489391 Dated 25 February 2009

Reference /A/ provides Boeing's previous fleet communication on the subject event. The US NTSB, FAA, Boeing, the Turkish DGCA, the operator, the UK AAIB, and the French BEA continue to actively support the Dutch Safety Board's (DSB) investigation of this accident.

The DSB has released a statement on the progress of the investigation and has approved the release of the following information.

While the complex investigation is just beginning, certain facts have emerged from work completed thus far:

- To date, no evidence has been found of bird strike, engine or airframe icing, wake turbulence or windshear.
- There was adequate fuel on board the airplane during the entire flight.
- Both engines responded normally to throttle inputs during the entire flight.
- The airplane responded normally to flight control inputs throughout the flight.


The Digital Flight Data Recorder (DFDR) data indicates that the crew was using autopilot B and the autothrottle for an ILS (Instrument Landing System) approach to runway 18R at Amsterdam Schiphol airport. During the approach, the right Low Range Radio Altimeter (LRRA) was providing accurate data and the left LRRA was providing an erroneous reading of -7 to -8 feet. When descending through approximately 2000 feet the autothrottle, which uses the left radio altimeter data, transitioned to landing flare mode and retarded the throttles to the idle stop. The throttles remained at the idle stop for approximately 100 seconds during which time the airspeed decreased to approximately 40 knots below the selected approach speed.

The two LRRA systems provide height above ground readings to several aircraft systems including the instrument displays, autothrottle, autopilots and configuration/ground proximity warning. If one LRRA provides erroneous altitude readings, typical flight deck effects, which require flight crew intervention whether or not accompanied by an LRRA fault flag, include:

- Large differences between displayed radio altitudes, including radio altitude readings of -8 feet in flight.
- Inability to engage both autopilots in dual channel APP (Approach) mode
- Unexpected removal of the Flight Director Command Bars during approach
- Unexpected Configuration Warnings during approach, go-around and initial climb after takeoff
- Premature FMA (Flight Mode Annunciation) indicating autothrottle RETARD mode during approach phase with the airplane above 27 feet AGL. There will also be corresponding throttle movement towards the idle stop. Additionally, the FMA will continue to indicate RETARD after the throttles have reached the idle stop

Boeing Recommended Action
- Boeing recommends operators inform flight crews of the above investigation details and the DSB interim report when it is released. In addition, crews should be reminded to carefully monitor primary flight instruments (airspeed, attitude etc.) and the FMA for autoflight modes. More information can be found in the Boeing 737 Flight Crew Training Manual and Flight Crew Operations Manual.

Operators who experience any of the flight deck effects described above should consult the troubleshooting instructions contained in the 737 Airplane Maintenance Manual. Further, 737-NG operators may wish to review 737NG-FTD-34-09001 which provides information specific for the 737-NG installation. Initial investigations suggest that a similar sequence of events and flight deck indications are theoretically possible on the 737-100/-200/-300/-400/-500. Consequently the above recommendations also apply to earlier 737 models.

[kitsap2]

If this is true, and the pilots, all three of them, flew this a/c into the ground, it leaves me utterly speechless. 

[joeyb747]

here is what I (think I) know: an airplane uses the air pressure to measure the altitude - above 10000feet the crew sets the altimeter to a default pressure (QNH) (1013hPa in europe I think) - below 10000feet the altimeter should be set to the local air pressure (information being received from ATC during approach). could be that the crew set their altimeter incorrectly - causing the aircraft to display a wrong altitude (ref. sea level).
however, if the airplane (at least airbus do that) gets below something like 2000feet - you also get the distance between ground and plane measured by a radar instrument. that one only displays wrong information if the system itself is broken...
if the aircraft gets below something like 100feet - there is either a hearble countdown (100fett / 50feet / 20feet / 10feet) or a "terrain" warning (depending on whether the airplance is "supposed" to be landing or not)
either way,
it's still strange three pilots in the cockpit did not notice the wrong altitude despite all the warnings, etc...

(if I just wrote a total incorrect nonsense, please correct my statements...)

I believe it's above 18,000 ft that they set to the standard altimeter setter(in the U.S. its 29.92), and below that it is set to local reading.  I don't think an incorrectly set altimeter could cause this. Variations in altimeter setting would only cause the sea level altimeter to be off by a few hundred feet. I've heard on some of the approach feeds something like "So-And So 1234, you a 500 below/above(what ever the case my be) your assigned altitude, current altimeter 30.12." (or what ever current reading is).

If that is incorrect, please correct me!

[kitsap2]

here is what I (think I) know: an airplane uses the air pressure to measure the altitude - above 10000feet the crew sets the altimeter to a default pressure (QNH) (1013hPa in europe I think) - below 10000feet the altimeter should be set to the local air pressure (information being received from ATC during approach). could be that the crew set their altimeter incorrectly - causing the aircraft to display a wrong altitude (ref. sea level).
however, if the airplane (at least airbus do that) gets below something like 2000feet - you also get the distance between ground and plane measured by a radar instrument. that one only displays wrong information if the system itself is broken...
if the aircraft gets below something like 100feet - there is either a hearble countdown (100fett / 50feet / 20feet / 10feet) or a "terrain" warning (depending on whether the airplance is "supposed" to be landing or not)
either way,
it's still strange three pilots in the cockpit did not notice the wrong altitude despite all the warnings, etc...

(if I just wrote a total incorrect nonsense, please correct my statements...)

I believe it's above 18,000 ft that they set to the standard altimeter setter(in the U.S. its 29.92), and below that it is set to local reading.  I don't think an incorrectly set altimeter could cause this. Variations in altimeter setting would only cause the sea level altimeter to be off by a few hundred feet. I've heard on some of the approach feeds something like "So-And So 1234, you a 500 below/above(what ever the case my be) your assigned altitude, current altimeter 30.12." (or what ever current reading is).

If that is incorrect, please correct me!

You are correct.  Aircraft altimeters are set to 29.92 at 18,000 ft and above.  Below 18,000 ft, they are set to local barometric readings.

An aircraft's mode C readout (altitude), as monitored at controller displays, must be within 300 ft of assigned altitude.  That is, less than 300 ft, for controllers to be able to use that information for separation purposes (terminal environment).  A controller must verify an a/c's mode c readout periodically.  If pilot states level at 9000, but his mode C, as depicted on the controllers display, is reading 8500, the controller will say something like, "N1234, verify at 9000, altimeter 3002"  If pilot states level at 9000, even after resetting to correct altimeter, and the a/c's mode C is still reading 8500, than the pilot should be told to, "stop altitude squawk, altitude differs by 500 ft."

[MathFox]

I am just pedantic here; but in the Netherlands the transition level is 3000ft. Altimeter setting would have added to the cockpit workload.

[snader]

here is what I (think I) know: an airplane uses the air pressure to measure the altitude - above 10000feet the crew sets the altimeter to a default pressure (QNH) (1013hPa in europe I think) - below 10000feet the altimeter should be set to the local air pressure (information being received from ATC during approach). could be that the crew set their altimeter incorrectly - causing the aircraft to display a wrong altitude (ref. sea level).
however, if the airplane (at least airbus do that) gets below something like 2000feet - you also get the distance between ground and plane measured by a radar instrument. that one only displays wrong information if the system itself is broken...
if the aircraft gets below something like 100feet - there is either a hearble countdown (100fett / 50feet / 20feet / 10feet) or a "terrain" warning (depending on whether the airplance is "supposed" to be landing or not)
either way,
it's still strange three pilots in the cockpit did not notice the wrong altitude despite all the warnings, etc...

(if I just wrote a total incorrect nonsense, please correct my statements...)

I believe it's above 18,000 ft that they set to the standard altimeter setter(in the U.S. its 29.92), and below that it is set to local reading.  I don't think an incorrectly set altimeter could cause this. Variations in altimeter setting would only cause the sea level altimeter to be off by a few hundred feet. I've heard on some of the approach feeds something like "So-And So 1234, you a 500 below/above(what ever the case my be) your assigned altitude, current altimeter 30.12." (or what ever current reading is).

If that is incorrect, please correct me!

You are correct.  Aircraft altimeters are set to 29.92 at 18,000 ft and above.  Below 18,000 ft, they are set to local barometric readings.

An aircraft's mode C readout (altitude), as monitored at controller displays, must be within 300 ft of assigned altitude.  That is, less than 300 ft, for controllers to be able to use that information for separation purposes (terminal environment).  A controller must verify an a/c's mode c readout periodically.  If pilot states level at 9000, but his mode C, as depicted on the controllers display, is reading 8500, the controller will say something like, "N1234, verify at 9000, altimeter 3002"  If pilot states level at 9000, even after resetting to correct altimeter, and the a/c's mode C is still reading 8500, than the pilot should be told to, "stop altitude squawk, altitude differs by 500 ft."

Hi all, the barometric altitude has no bearing on this subject.
The aircraft was on the final approach so APP was the active FD and autoplilot mode.
In this mode, the Flight Directors gives the steering commands for the autopilot to steer the aircraft to the glidepath.

Barometric altitude is not used for landing, nor does a QNH setting or QFE setting have any influence on ATC altitude.
The Flight directors in this situation only look for the G/S and LOC signals until the radio altitude drops down.

The altitude call outs come from radio altimeter #1.
Offcourse the flare command also comes from radio altimeter.

If the FCC sees the GS and LOC signals as 'on course' it will then wait for a decreasing radio altitude and at roughly 50 feet or so go into FLARE.

This is the way the system works and is cause for concern as this defect seems to happen quite a lot.

source: http://www.aircraftmech.com/pfd.html

[joeyb747]

With so much redundancy in modern aircraft today, only the #1 radio altimeter handles the entire approach? What if #1 was inop? Can #2 be set up to handle the same functions as #1?

According to the preliminary report iflysky linked to a few posts back(re-linked here),

http://www.onderzoeksraad.nl/docs/rapporten/Persverklaring_4_maart_GB.pdf

the last line in the report identifies the Left Radio Altimeter as the item that malfunctioned. I'm assuming, as with everyting else on a modern aircraft, the Left unit would also be identified as #1.

From report:
"With the exception of the malfunction of the left radio altimeter the investigators of the Dutch
Safety Board have not yet found any irregularities."

And again, given the crew members experience, it truly is sad that they didn't recognize the issue at hand, and hand-fly the airplane. If infact this is what caused the crash, then this entire event could have been avoided, leaving only a squawk in the logbook, not a destroyed aircraft, injuries, and death.

[another_one]

I believe the "landing gear must go down" is a translation error.

Indeed!! The original Dutch text is "landingsgestel moet uit":

landingsgestel = undercarriage

http://nl.wikipedia.org/wiki/Landingsgestel <----> http://en.wikipedia.org/wiki/Undercarriage

moet = must
 uit = out.

Does "Undercarriage must [go] out" make more sense?

[iskyfly]

Does "Undercarriage must [go] out" make more sense?

The GPWS warning is "TOO LOW GEAR".

[iskyfly]

With so much redundancy in modern aircraft today, only the #1 radio altimeter handles the entire approach? What if #1 was inop? Can #2 be set up to handle the same functions as #1?

They were not using the redundancy they had available to them. They were not doing a dual channel full autoland. Had they been doing a dual channel autoland the AP would have disconnected which would give the crew another indication that something was wrong.

[joeyb747]

With so much redundancy in modern aircraft today, only the #1 radio altimeter handles the entire approach? What if #1 was inop? Can #2 be set up to handle the same functions as #1?

They were not using the redundancy they had available to them. They were not doing a dual channel full autoland. Had they been doing a dual channel autoland the AP would have disconnected which would give the crew another indication that something was wrong.

Thanks iflysky. I was thinking in the back of my mind that either/both radio altimeters could handle the approach. I mean, modern aircraft have backups for the backups!  

[empiredude]

alright, just some more questions:

1. the airplane was on final approach when the accident happened. as far as I know, there is usually neither the A/P steering the aircraft nor the A/T engaged during final approach - unless, of course, there's low visibility and a automatic landing in progress (but this was not the case). so why was the airplane not flown manually?!
I thought A/P and A/T are usually (or even have to be) disengaged after the aircraft is established on the ILS track (except - as mentioned above - during low visibility procedures...)

2. do I get this right: it seems that the radio altimeter that measures the distance between aircraft and ground failed by indicating a wrong altitude above ground - causing the A/T to go into retard/flare mode (too early)?

it's really hard to believe no one (of three pilots) noticed this... I mean in order to stay on the glideslope, the attack angle of the aircraft must have been noticeably increased, not to speak of the speed massively decreasing - and all the warnings (like "too low" / "terrain" / "too slow" etc...)
I mean, is it already too late to recover an aircraft from impending stall when the warnings are triggered?

4. how well are the aircrafts being followed by ATC during final? does falling below the glideslope get noticed by the controller? (like with a warning on the radar screen or something?)

anyway, I know it's not really useful to say all those "unbelievable / how could they not... / how can it be that...." statements - but nevertheless, it is a little bit surprising indeed that such a malfunction during final approach did not get noticed - or only got noticed when it already was too late to recover...
I mean on one side there seems to be crew error -
but on the other side what are all the warnings good for if they warn you when it is already too late to intervene....?

[sykocus]

4. how well are the aircrafts being followed by ATC during final? does falling below the glideslope get noticed by the controller? (like with a warning on the radar screen or something?)

I don't know of any radar system that will detect when a plane is below the GS to a certain runway. The radar systems I've used all have a map of the terrain in the area programed in. Based on aircraft descent rate and trajectory the radar will give a visual and aural alarm. The system isn't perfect and often gives false alarms especially when landing.

[MathFox]

The weather would have been a factor too: cloudy and hazy

EHAM 250925Z 20010KT 4500 BR SCT007 BKN008 OVC010 04/03 Q1027 TEMPO 2500

When the radio altimeter malfunctioned the plane was above or in the cloud layer; with no easy visual clue for loss of speed.

[sykocus]

The weather would have been a factor too: cloudy and hazy

EHAM 250925Z 20010KT 4500 BR SCT007 BKN008 OVC010 04/03 Q1027 TEMPO 2500

When the radio altimeter malfunctioned the plane was above or in the cloud layer; with no easy visual clue for loss of speed.

Wouldn't those conditions tend to mean a crew is watching the instruments even more closely then if it were say 5SM SCT035

[empiredude]

just the facts please....

http://www.flightglobal.com/blogs/unusual-attitude/2009/03/turkish-airlines-crash-evidenc.html

* Capt LH seat, (fully qualified) FO in RH seat, observer FO on jumpseat

* FDR stores 25 hours, in this case 8 flights, same problem had occured twice previously before previous landings.

Am I getting this right: The same airplane that crashed had had the exact same problem (wrong radar altitude - A/T going into retard/flare mode too early) twice within 25h before ithe accident and

1. no one withdrew that plane for reparation?!

2. no one declared the A/T as INOP (at least for landing procedures)?!

3. no one even seemed to inform the crew about this?!

[iskyfly]

Am I getting this right: The same airplane that crashed had had the exact same problem (wrong radar altitude - A/T going into retard/flare mode too early) twice within 25h before ithe accident and

a malfunctioning radalt to quote this 737 captain is "something that happens all the time."

I encourage people to read his replies on this subject as they are very informative and give us some insight on what happens on the flightdeck when your approaching the runway in IMC, vectored by ATC onto a tight / short final, above GS, have checklists to go through, looking out the window for the runway....

http://www.airdisaster.info/forums/viewtopic.php?f=6&t=2098&st=0&sk=t&sd=a&start=175#p27794

[iskyfly]

alright, just some more questions:

1. the airplane was on final approach when the accident happened. as far as I know, there is usually neither the A/P steering the aircraft nor the A/T engaged during final approach - unless, of course, there's low visibility and a automatic landing in progress (but this was not the case). so why was the airplane not flown manually?!

Because they were in IMC conditions.

[empiredude]

thanks for your replies iflysky

you're right, there are often a couple of systems malfunctioning or inop on a plane (without being a real danger to flight operations). but still, isn't it a bit terrifying that they were using a system to control their airspeed during final that had been broken for at least a couple of flights earlier - and didn't even seem to pay special attention to this system.

but again, flight crew definitely has a high workload during final and there was definitely only very little time to react to the problem after it had been noticed - apparently too little...

[joeyb747]

My question is: If they knew there was a problem with radio altimeter number one, why wasn't the crew operating in dual mode?

With so much redundancy in modern aircraft today, only the #1 radio altimeter handles the entire approach? What if #1 was inop? Can #2 be set up to handle the same functions as #1?

They were not using the redundancy they had available to them. They were not doing a dual channel full autoland. Had they been doing a dual channel autoland the AP would have disconnected which would give the crew another indication that something was wrong.

At least then as iflysky pointed out, the autopilot would have sensed a discrepancy between the two rad alt and disconnected the autopilot.

Aircraft keep a logbook for these such reasons. When something is wrong with the bird, you note it and pass it on to the next crew, or MX at the home base. Does anyone know for sure if this item was noted in the logbook?? I've seen logbook entries for some pretty trivial squawks. I find it hard to believe that was not noted in the logbook.

"Radio Altimeter Number One Intermittent" seems pretty important information!

[toogd]

It is simply strange that the plane was allowed in the air with a "temperamental" altimeter hooked up to its hands free landing system. Surely that calls for some intensive soul-searching at Turkish Airlines...

That the crew end up depending on this port side altimeter seems to be because it they'd done it properly, with dual altimeter input, the autopilot would have disconnected (as it damned well should do!), but I can't find anything here about the other alternative.
-Was it possible to use the right hand side altimeter for the approach?
-Why did the pilot opt to use the documented "temperamental" unit to control this fateful glide?

Is there some logic to this that I am missing? What would a pilot be trained to do in these circumstances?

I can appreciate that it might be easy to miss detecting that the plane was loosing air speed until it was too late, but how do you end up relying on a piece of equipment that you know is faulty and not notice that it is failing you? I'm not getting it.