Attenuated by phugoid damping?

DFDR barometric-height data (on Habsheim QFE)
 

Quote from me:
"The DFDR trace shows the a/c consistently lower, eventually below 30 ft QFE."

Quote from CONF_iture:
"I don't see any QFE altimeter reading in the DFDR traces, do you ?"

Sorry, Confit, that was sloppy of me. I should have acknowledged that the "trace" I was looking at is in the report that Airbus themselves did in 1995, not the BEA report.

However, Annexe VII of the BEA Rapport Finale, "Tome 1", represents one of the lists of second-by-second flight parameters decoded from the DFDR. Inconveniently, on my PDF copy the print is reminiscent of the blurred, bottom carbon-copy of a document created on a cheap portable typewriter. (How DID we ever manage in those days?) ;)

Column 2 is a second-by-second timeline.
Columns 3 - 5 are barometric readings, which I'll now try to explain in fundamental terms for our non-aviator readers.
Column 3 seems to be the flight level, sampled every 4 secs, and rounded down. (So "+018" means FL 018.)
Column 4 is either the QNE (i.e., the pressure altitude, based on the standard sea-level pressure of 1013.2 hPa) or the altitude above sea-level, based on the current QNH of 1012. (More likely the QNE.) I am ignoring that parameter.

Column 5 is almost certainly the actual reading on the captain's altimeter, the sub-scale of which would have been set to the QFE of 984 hPa.
(For the uninitiated, that would theoretically read between zero and -15 ft for an a/c parked on the airfield with a perfect altimeter. The lower figure of -15 ft would apply if the precise pressure at the airfield had been 984.5 hPA, which ATC would have broadcast as 984 hPa.) So it is the best reading of height above the airfield obtainable by barometric means. (See my previous post.)

At time 332.0 secs, the value (as far as I can see) is 0030.
At time 333.0 secs, the value (as far as I can see) is 0024 or 0026.
At time 334.0 secs, the value (as far as I can see) is 0024 or 0026.
At time 335.0 secs, the value (as far as I can see) is 0?48 or 0248.
(The latter may be the defined point of "impact".)
At time 336.0 secs, the value (as far as I can see) is 0000, and remains so until the end of the recording at 339.0 secs.

Understood (and I did read it the first time :}), however that's not what I was getting at. The FCTMs use very equivocal language when referring to the procedure of pulling back in High AoA Protection mode - e.g. "Alpha Max *may* be achieved" and "*Approximately* correspond[ing]".

Given Airbus's notorious pedantry when it comes to language, this choice of words seems to indicate that achieving the precise alpha value shown on the graph as "ɑMax Full aftstick - VɑMax" will not necessarily be the case if suitable conditions aren't met. Therefore CONF iture's argument that AF296's failure to achieve absolute Alpha Max is tantamount to a technical problem seems to be negated by the language used in the FCTM.

What are the 'suitable conditions' that must be met in order that the airplane can support its own weight at VɑMax/Vs1g?

Indeed, I think Conf_iture is comparing the "17.5° AoA = AlphaMax" from § 1.16.1.2 and the column labelled "INC" (short for Incidence = AoA) in the FDR annex (page 37 of the french original PDF report). We see on that chart that in the end of the flight (~12h45'33"), the AoA reaches and maintains 15°.
17.5 - 15 = 2.5. There we are.

Thanks for the value (16=max deflection), was looking for it. :)

Two remarks, regarding your point:

1/ I think the last column of values (in the FDR annex) are the decimals for the AoA. If I'm correct, then the last AoA value registerd is 15.4°, not 15°. The aircraft would have been then 2.1° short of AlphaMax, not 2.5.
[edit] This remark prooved to be null and void, as pointed by Conf below. Disregard, sorry.

2/ Between timestamps 1245.328 and 1245.335, the registered AoA varies between 2 extreme values of 13° and 15°. This variation led me to believe that, either:

• something prevented the aircraft to reach AlphaMax despite the crew demand (some are advocating this, but no you, unless I'm mistaken),
• or calculated AlphaMax at the time was lower than theorical max AlphaMax of 17.5° (DW... why?),
• or that the crew did not maintain full aft back stick. You said above that the crew did. Ahem... really?
  Do we have reliable sources on that topic? In fact, yes: the very same report. See: Why AF A320 didn’t reach AlphaMax @ Habsheim.

My conclusion: indeed, as Conf_iture said, the aircraft did not reach AlphaMax. Because the pilot didn't ask for it (unless, perhaps, at the very last second, i.e. too late).

Your explanation as to why is there:
That doesn't sound logical if the pilot had decided to rely on the AoA protection, because as you said yourself, to get AlphaMax you have to pull the stick full aft and maintain it there.

AZR,

Thanks, your post together with Chris Scott's was very helpful for reading the DFDR print-out. The graph below shows some parameters for the last 5 seconds of the accident flight.

My timeline differs 1 second from Chris, because the report puts t=0 at 12:45:39 and states that the airplane entered alpha protection mode at 12:45:34 (t-5) at 122 kIAS and alpha=13 degrees.

AlphaCmd in the graph represents the sidestick position PROFON_CAP, assuming a linear relation between AlphaProt=14.5 for SS=0 and AlphaMax=17.5 for SS=-16.

http://i.imgur.com/xDSJu8l.gif

P.S.
There is something that bothers me in the DFDR print, and that is that all parameter values are shown for the same timestamp in seconds, suggesting that all values are sampled simultaneously at the same instant. As I understand the recording format of current DFDR's, the parameters are sampled sequentially, i.e. each parameter sampling time is offset a fraction of a second from the time recorded in the first word of each 1-second (sub)frame of data, depending on the position of the parameter 'word' within the 'frame'. (See for example the DFDR read-out recently released by the NTSB for the Asiana crash). IOW, the timing of a parameter recorded near the end of a frame can differ nearly one second from that of a parameter recorded early in the frame.

Is that different for this recorder?

Hi HN39,

That graph is going to be a very useful tool, but I'm wondering if your time base might be one second adrift at the moment? I agree that the defined moment of "impact" is 12:45:39, but I'm still trying to find your quoted reference that Alpha Prot started at 12:45:34 (i.e., t -5), at an AoA of only +13.

Quote from the BEA (please excuse lack of accents):
"A t - 4 s, commutation sur la loi de pilotage en incidence, la valeur de 14.5 deg ayant ete atteinte, cette loi etant ensuite conservee."

Also, the DFDR print-out shows a corrupted reading ("D39"?) of longitudinal stick position ("commande profonde" or, in this case, alpha-cmd) at what I interpret as time 12:45;39. The two preceding values are U17 [sic] at 12:45:38 (t - 1) and U16 at 12:45:37 (t - 2)

Does the above make sense?


PS
Just noted your PS. I'm not sure, but I've been working on the assumption that the first value in a time frame is the exact second. So if there are 4 values in one second of a time frame, say 316.0 (s), the first would be at 316.0, the second at 316.25, the third at 316.5, and the fourth at 316.75. So, as you say, the values are sequential.
Likewise, with the hours and minutes, I've been assuming that the first line of "1245" refers to a time of 12:45:00. What I don't yet understand is why the associated value (column 2 on the same line) reads "296.0 (s)", but am hoping it doesn't matter!

By any chance, would you have a link for that Airbus report ?
It's not your PDF copy, it is how the BEA conveniently (?) published its report ...

At such level it is not attenuation but plain restriction.

There is no flight ending around 12h45'33" ... but apparently more confusion to come with the timings ...
I don't know what that column is, where the same sequence 1234 is reproduced, but it is not for the decimals.
Bottom line, the System had no intention to deliver those extra degrees.
It is very logical for a guy of his experience to not pull too early when thrust is not felt yet ... but that's another part of the topic ...

Hi Chris,

About the time base: On page 10, RH column, halfway down the page:
- les manettes de poussée des moteurs ...
A 12 h 45 mn 34 s la vitesse est de 122 kt et l'incidence 13 (deg). You are correct (I was wrong) that entry into alpha-prot mode occurred at t-4, one second later (paragraph 1.16.1.2 refers).

Re your comment on my P.S.: The DFDR recorded 200 parameters, including 141 discretes (yes/no) - para.1.11.4. We don't know if the sequence on the print is the same as the recording sequence, but the first word in the recorded frame normally contains the date/time of that frame. I read the second column headed TGEN (s) e.g. 296.0 as a frame counter starting from some arbitrary datum.

Makes sense too at FL350, but less at SL when the engines are supposedly approaching 90% N1 and Performance is requested NOW.

I understand Vs1g = 1.06 Vs but I'm not sure about VɑMax = Vs1g ?
Would you have the reference ?
What I get from the FCOM is VɑCLMax = Vs1g

I had told JT that I didn't want to participate in this thread, but I must admit that the debate has been to a much higher standard than I feared it might be.

At last I see what Confiture has been getting at, although I am not convinced that the difference matters much.

So far as 'phugoid damping' is concerned, the discussion has centred on the response in the last three or four seconds. The phugoid period at 118 kts is around 27 seconds - there is no way that phugoid motion is going to affect the issue.

The point made about timing deduced from the DFDR record is well made; since any one of the parameters can be up to a second adrift relative to another and the whole debate relates to a period of four or five seconds it is, IMHO, almost pointless to attempt to deduce much from that data.

In any case, the data available does not give enough detail to assess the motion, and in particular the effect of elevator lift on the motion. It has been said that the AI laws attempt to drive the aircraft so as to maximise performance. The usual way of improving pitch response would be to overdrive the elevator to improve the pitch acceleration and then to back it off to avoid overswing.
But when you apply a load of 'up' elevator the first thing that happens is that the aircraft starts to accelerate downwards because of the negative lift. It is only after the pitch/AoA has built up enough to overcome this negative lift that the aircraft starts to accelerate upwards. The net result is that there is a slight delay after application of elevator before the CG gets above the original height. The delay isn't much - from memory it would be around 0.5 to 1 second on something like an A320, which is unimportant in normal operations but if you are only one second away from disaster it is very important. And of course the back end will lag the CG in terms of height development!

At the very end of the BEA report there is a graph of an AI flight test to repeat the Habsheim event (done at a safe height of course). That graph, with pretty much the same stick input as the actual event, shows that once alpha-prot was triggered the AOA followed the stick position over the first five seconds at the end of which time the AOA was about 16 deg. It then took another 10 seconds with the stick held fully back until the AOA got to 17.5 deg. This is entirely consistent with the AI report mentioned by Chris Scott which says:

It also explains why the aircraft did not develop 17.5 deg AOA during the accident sequence - it was simply not given enough time to do so before contact with the trees.

That is one reason why I am not convinced that failure to achieve 17.5 deg AoA is important - the important thing IMO was the timing. The other reason is also related to timing; increasing AoA is not going to give an increase in flight path angle unless accompanied by enough thrust. Increasing AoA to 17.5 deg would indeed produce more lift but it would also increase the drag significantly and unless thrust were increased the aircraft would either decelerate further towards stall or sink. But if he had increased thrust earlier there would not have been a problem .......

Hi Owain Glyndwr,
I don't quite understand that statement.

If we consider the conversion of kinetic energy to potential energy from say 112 kts (achieved) to 107 kts (a guess of flying possibly 5 kts slower).
1/2 * m * Vsquared = m * g * h.
re-arranging, Delta h = {(189 ft per sec)squared - (180 ft per sec)squared} / 2 * 32.
By my maths, if the aircraft had swapped another 5 kts of airspeed for a gain in height they could have climbed 52 feet. Even then, they would still be below 100 radio.

Apologies - I quoted incorrectly from memory from the A330 FCOM which actually says in 3.04.10: "The FCOM uses VS for VS1g".

However, the explanation preceding that statement makes it quite clear IMHO that VɑMax is identical to VS1g:

Certification flight tests to determine the reference stall speed are conducted with all airplane systems functioning as designed, i.e. in normal law. VɑMax is the minimum speed that can be demonstrated by flight test in normal law for A320 and A330.

OK - I should have written sustained flight path angle.If they had zoomed to 85 feet they would then have the thrust they started with and the drag associated with the higher AOA - and as you get near to stalling AoA the drag goes up rapidly.

In a non precision approach MDA all authorities insist on +50ft. because when you execute a GA you are expected to loose that much and You are at Vapp at that point which is at least say 30 kts higher than alpha max. Then if you lost 100ft from alpha max is it surprising? this fly past was doomed even at 100ft because you will have no elevator left to effect flight path change. Some height loss had to be accepted.

No way? While I agree that the phugoid damping would not be sufficient to explain the apparently 'attenuated' response to the sidestick command, I'm not convinced that it could not have contributed to it.

As described in the Bilbao report " the EFCS behaves as a damper of the oscillations, commanding appropriate variations of angle of attack in a way that, when the aircraft is slowing down, makes it pitch downward and vice versa". In the last 5 seconds of the Habsheim flight, the airplane was decelerating at about 2 kt/sec. In the A340 airprox incident there was about 1 degree of alpha-prot bias between accelerating and decelerating parts of the trajectory for 2 kt/sec difference between acceleration and deceleration.

Originally posted by HN39
I'm not saying that phugoid damping won't affect long term response when under alpha-prot, only that in the first 4 or 5 seconds of a 27 sec period motion the effect will be very small and that the motion will be dominated by the short period mode.

Loss of height
 

Quote from vilas:
"In a non precision approach MDA all authorities insist on +50ft. because when you execute a GA you are expected to lose that much..."

You are not comparing apples with apples. Your instrument-approach analogy addresses a NPA of the modern variety, in which the descent fom the FAF (final approach fix), at the final approach speed, may be predicated on a steady approach angle in excess of 3 degrees to an MDA, at which point the G/A is initiated immediately (unless the descent can be continued visually). During the G/A, the a/c must not descend below the OCH (obstacle clearance height), so the chosen MDA may vary from type to type. The fly-past at Habsheim was not planned or briefed in that fashion.

Many years ago, the concept of MDA did not exist (as such). Even on precision approaaches like ILSs, the approach was flown to a CH ("critical height") or OCH, at which the a/c normally flew level until passing the threshold of the runway (usually assessed by stopwatch) unless the descent could be continued visually. Approaching the critical height, the pilot gently flared the a/c to fly level at it (but never below it - going below the CH/OCH on instruments was a fail-point on an Instrument-Rating test). In order to have a chance of "getting in" to the runway in marginal conditions of cloud ceiling and/or visibility, a large jet would have to fly the instrument descent from the FAF at a steeper angle than the average slope from FAF to threshold; in order to have manoeuvering space to level off, see the runway, and resume the descent. (In these days of reliable ground-speed information and - since 1983 on the A310 - the FPA "bird" - the concept of leveling off at the MDA/OCH/CH is no longer necessary on big jets.)

Quote:
"...and You are at Vapp at that point which is at least say 30 kts higher than alpha max. Then if you lost 100ft from alpha max is it surprising? this fly past was doomed even at 100ft because you will have no elevator left to effect flight path change. Some height loss had to be accepted."

No, that idea needs to be put to bed immediately. The PF would have planned to initiate a flare prior to reaching 100ft QFE or 100R. He was visual, and not constrained to descend steeply prior to levelling off. His descent below 100 ft was prolonged for nearly half a minute, and was therefore either deliberate, or due to sloppy flying.

If the PF was finding the prolonged application of back-stick physically difficult, he should have gone-around earlier. (Even if that was not the problem, he should have gone-around earlier...) The briefed fly-past at 100 ft, however that height was going to be assessed/defined, was a risky manoeuvre that led to an accident. As the whole game plan was voluntary, as well as plainly and irresponsibly risky, any excuses for its shoddy execution are incompatible with the essential concepts of command and responsibility. (End of sermon!)

Now, back to the minutiae of flight mechanics...

:ugh: What an idiot I'm, for not having seen this repetition of sequence. Of course you're right on that. Editing my previous post now. Thanks.:ok:

Hi Owain,

I find it hard to believe that the EFCS waits until it has diagnosed a phugoid motion of the airplane. As I understand the description in the Bilbao report, the EFCS in alpha-protect mode applies a bias to the commanded alpha as soon as it detects a deceleration or acceleration of the airplane, which by the way is good for stall protection as well as for phugoid damping.