Go Back  PPRuNe Forums > Flight Deck Forums > Tech Log
Reload this Page >

AF 447 Thread No. 5

Tech Log The very best in practical technical discussion on the web

AF 447 Thread No. 5

Old 6th Aug 2011, 19:54
  #1701 (permalink)  
 
Join Date: Oct 2009
Location: UK
Posts: 1,271
Hi DozyWannabe,

And had they let go of the controls then the elevator demands would have ceased and the trim would have returned to neutral.
If only that were true.

Once the aircraft was climbing with fixed thrust, and with the stick pitch neutral, the FBW computers would maintain 1 g, and continue to trim the aircraft in pitch as the speed washed off.

In your glider, if you don't move the trim wheel, it will feel heavier in pitch the slower you fly - that feed back is missing with AB FBW.

When Normal Law is working it feels progressively heavy in pitch as speed reduces below alpha Protect, and impossibly heavy if you attempt to fly beyond alpha Max. ALT LAW has no such "protection" (Limits), nor any feel.

Last edited by rudderrudderrat; 6th Aug 2011 at 21:24. Reason: syntax
rudderrudderrat is offline  
Old 6th Aug 2011, 21:45
  #1702 (permalink)  
 
Join Date: Jul 2002
Location: UK
Posts: 3,182
Originally Posted by rudderrudderrat View Post
Once the aircraft was climbing with fixed thrust, and with the stick pitch neutral, the FBW computers would maintain 1 g, and continue to trim the aircraft in pitch as the speed washed off.
Could we get some confirmation from PJ2 or A33Zab?

(@RRR - I don't know about "belief" - I'm asking for some confirmation. Either way it doesn't alter the fact that without overcontrolling, the THS wouldn't have ended up in that position, or that it was possible to set the THS position either by maintaining nose-down inputs or moving the trim wheel. RWA seems to be under the misconception that autotrim commands a THS angle and holds it there regardless of input, when in fact it's just following elevator demand - a mistake I made when I first started looking into this)

Last edited by DozyWannabe; 6th Aug 2011 at 22:19.
DozyWannabe is offline  
Old 6th Aug 2011, 21:47
  #1703 (permalink)  
 
Join Date: Oct 2009
Location: UK
Posts: 1,271
Hi DW,

You can't believe it either can you?

Last edited by rudderrudderrat; 6th Aug 2011 at 22:10. Reason: syntax
rudderrudderrat is offline  
Old 6th Aug 2011, 22:23
  #1704 (permalink)  
 
Join Date: Jun 2009
Location: California
Age: 51
Posts: 203
It appears that rudderrat is correct.
If only Normal law had been preserved.



ALT law took out the following protections:
  • pitch attitude
  • low energy
  • alpha floor
  • high speed
  • turbulence damping
Low speed stability protection becomes active. Near stall this brings up a black and red barber pole (oblique stripes?) on the PFD speed. However a black and red ladder (horizontal stripes) is the overspeed!!!
This is surely a trap for mis-interpretation. Why on earth do they use the same colors for overspeed ladder as stall ladder in ALT law! Remember in Normal Law the stall ladder is amber.

Anyone else confirm this?

LOW SPEED STABILITY PROTECTION
An automatic nose down command is introduced to increase speed. No reference to AOA, only speed. Operates 5 to 10 kts above stall warning depending on weight & slat/flap configuration. The pilot can override.
"STALL" announces and crickets heard prior to stall speed being reached. PFD shows black and red barber’s pole for stall warning speed. V-apha-prot and V-alpha-max are replaced by Vsw (stall warning speed). No alpha-floor protection is available.
xcitation is offline  
Old 6th Aug 2011, 23:31
  #1705 (permalink)  
 
Join Date: Jun 2009
Location: NNW of Antipodes
Age: 77
Posts: 1,330
Ex BEA - Interim Report No.3 - page 76 ...

Until the end of the flight, the angle of attack values became successively valid and invalid. Each time that at least one value became valid, the stall warning triggered, and each time that the angles of attack were invalid, the warning stopped. Several nose-up inputs [nose-down inputs] caused a decrease in the pitch attitude and in the angle of attack whose values then became valid, so that a strong nose-down input led to the reactivation of the stall warning. It appears that the pilots then reacted by a nose-up input, whose consequences were an increase in the angle of attack, a decrease in measured speeds and, consequently, the cessation of the stall warning.
Bold text as written, and [bracketed] text as I believe was intended.

The French version of this statement on page 78 is more verbose, but has arguably led to a misrepresentation when subject to translation. However the French version is the work of reference, and I've changed the NU inputs to ND inputs in my English copy.
mm43 is offline  
Old 6th Aug 2011, 23:52
  #1706 (permalink)  
 
Join Date: May 2010
Location: USA
Age: 72
Posts: 18
Cockpit Confusion

Smiling Ed and Rudderrudderrat – you are patient and well spoken. I like your style.

I am a pilot, though just VFR, and wouldn’t presume to compare my piloting skills with yours or most of the contributors on this site. I do, however, know a thing or two about IT and implementing automated systems. I am not anti-Airbus, anti-Boeing, or anti-anything else. I am pro-arrival-at-my-destination. IT has taught me that thoroughly tested software can – and does – go wrong, and that just because something can be automated doesn’t mean it should be. Critical systems should, IMO, allow human beings to interrupt the automation when appropriate; i.e., “Give me the speed even if it is below your threshold!” or “Give me the AoA even if it is unreliable!” Perhaps that would have helped on AF447.
My hope is that the authorities will consider how decisions made years ago led to some of the confusion in the cockpit. There apparently was a mindset years ago – later reconsidered - that stall was no longer an issue. Did that misplaced confidence in automation pervade the builders and operators? What (if any) other design features are based on the assumption that automation will prevent bad situations? Several people have commented on design and training issues in which pilots lost to IT or bean counters (or perhaps other pilots). These include the stick issues including simulated feel, the lack of AoA indicator, visibility of pilot actions, etc. Dissecting how the old decisions were made, and whether they were good or bad, would be a good lesson in organizational dynamics. I also wonder:

1. What led to the decision to disable SW and displays below certain angles and speed thresholds? Several have said the AoA vanes work fine down to a light breeze. Were 30 deg and 60 knots (IIRC) chosen as limits because out-of-range values are unreliable or because the software would “protect” them from that situation?

2. Why no audible auto-trim? That seems quite an oversight to this novice. A lot of training emphasis was placed on avoiding manual trim. Did the designers think trim – like stall – was a non-issue? Maybe someone felt an audio feedback would just encourage pilots to be more aware of trim, and that was presumably a bad thing. The software, after all, would make that a non-issue.

3. What led to all the laws? The mixed-mode operations and varied levels of degradation seem confusing and counter-intuitive. I read it in the comfort of my chair with tequila at hand. How would it feel to be handed this situation at altitude, mid-Atlantic, 200 souls on board, improperly trained, for the first time?

IMO, too much false confidence in a number of areas led to the disaster. Designers were confident of “protections” that didn’t cover all the bases. AB failed to appreciate the significance of the pitot failures and the risk of simultaneous loss. AF was unable/unwilling/unaware re training issues. The gentlemen flying the aircraft were handed a basket of snakes. I’m not qualified to say anything about their actions, and hope never to be handed a situation like that.
Pardonne-moi pour le poste de longue.
MountainWest is offline  
Old 7th Aug 2011, 01:44
  #1707 (permalink)  
 
Join Date: Jun 2009
Location: Paris
Posts: 691
Baro-inertial Vertical Speed

Hi Blujet,
Welcome back, sir.
Originally Posted by Blujet
@takata: I have been trying to get info about VS indication on PFD(not data) and could not find any tests (AMM) were this system is checked. Maybe you have another source?
Did you check 34-14-00?
In fact, the Vertical Speed displayed is either "baro-inertial" (not only inertial) or "barometric". My first guess, from graph, was that no inertial V/S data was displayed passed -10,000 ft/mn. But it is not correct. The word range is +/- 32,768 ft for baro-inertial v/s, and only +/-20,000 ft for barometric V/S.

So, in fact, the reason should be simply because all ADRs were rejected by they respective IRs (this 0211 FLR ACARS discussed previously). It will become NCD and replaced by ADR source V/S on PFDs.

- Inertial vertical speed (365) and inertial altitude (361) computation:
The IR software contains a baro-inertial loop to compute the Inertial Vertical Speed and Inertial Altitude. This loop permits to take advantage of the different qualities of the inertial and air data systems. The IR brings its better behaviour in dynamic maneuvers while the ADR brings its stability in time (no drift of the outputs like in IR).

The ADR source to be used is selected according to the configuration of the discretes described previously. When no ADR source is available, Inertial Altitude and Inertial Vertical Speed are sent with their status matrix coded NCD (No Computed Data).

The principle of the baro-inertial loop is given here:

takata is offline  
Old 7th Aug 2011, 03:57
  #1708 (permalink)  
 
Join Date: Jan 2008
Location: Atlanta, GA, USA
Posts: 349
Not helpful -the only one showing a dismissive and arrogant attitude is you.
Arrogance is to assume that anyone who really flies worries about V_office-chair.
deSitter is offline  
Old 7th Aug 2011, 04:27
  #1709 (permalink)  
 
Join Date: Aug 2005
Location: fl
Posts: 2,561
ALT law took out the following protections:
pitch attitude
low energy
alpha floor
high speed
turbulence damping

I'm curious why the pilot pulled up with no overspeed warning. Since the CVR had no such warning why did he pull up? Inexperience or trapped pitot presure with his climb causing static pressure to indicate an overspeed with no warning.
bubbers44 is offline  
Old 7th Aug 2011, 04:50
  #1710 (permalink)  
 
Join Date: Jun 2009
Location: California
Age: 51
Posts: 203
Originally Posted by bubbers44
I'm curious why the pilot pulled up with no overspeed warning. Since the CVR had no such warning why did he pull up? Inexperience or trapped pitot presure with his climb causing static pressure to indicate an overspeed with no warning.
The initial pull up immediately after AP disconnect IMHO was because pitch was 0 deg, altitude descending. For some reason his 2/3 deflection pull up maintained a level altitude for a number of seconds (downdraft?).

Subsequent pull ups IMHO were because he saw the red and black barber stripes (low speed stability protections for stall only showing in ALT/DIR LAW) on his speed and mis-interpreted as the red and back horizontal striped ladder of overspeed. The actual speed showing as being low - but he understood speed had failed. In normal law the low speed bar is amber.

Happy if someone can debunk/support my speculations.
xcitation is offline  
Old 7th Aug 2011, 05:15
  #1711 (permalink)  
 
Join Date: Jul 2007
Location: Virginia, USA
Age: 82
Posts: 77
32,768

Just a number to many people. It is in takata's post a bit above, #1698. But it has a special significance in computer programming. It is half of a "64K" space, sometimes called a page space. In the simple program "Notepad" it was the maximum number of characters in any note you wished to jot down, for example (at an earlier time). Usually adequate for that, but if not, too bad.

Here it is the number of possible ft/min, up or down, that can be displayed, for vertical speed. I don't want to argue here whether it is an adequate number, but simply to point out the way in which the binary computer efficiency considerations can influence the range of "instrument" displays.

The numbers in this 64K space require 15 binary bits (ie 0 or 1) to represent them, plus a 16th to represent the sign. Computer words are commonly multiples of 8 bits in length these days.

Actually, in one version of implementation of this numbering, 32767 positive values are available, 32767 negative values, and two zero values. That's right, there are two zero values in the most simple arrangement-- plus zero, and minus zero. That wouldn't be so bad if level flight didn't require a zero value. I'm not being facetious here-- this one issue alone required a lot of attention in many computer applications over the years.

The point here is that there are many things in control by computer program that are subtly different from the way we are trained to think, and may occasionally lead to unanticipated results in service. This is going to be particularly true in computers pushed to the maximum in speed with a minimum of size, weight, and power consumption; using software written by people in organizations under similar pressures, to some extent.

The views of pilots as to how such software is working in practice therefore deserve some respect. Unfortunately, once the aircraft is sold, the rewriting of software doesn't contribute anything to the making of that sale.
Old Engineer is offline  
Old 7th Aug 2011, 05:58
  #1712 (permalink)  
 
Join Date: Jun 2009
Location: Paris
Posts: 691
Originally Posted by xcitation
The initial pull up immediately after AP disconnect IMHO was because pitch was 0 deg, altitude descending. For some reason his 2/3 deflection pull up maintained a level altitude for a number of seconds (downdraft?).
Illusion, deceleration, and weather.
- Illusion because altitude recorded is "indicated", then partially false at UAS point.
- Deceleration, because autothrust started to reduce N1 by 16% seven seconds before UAS event.
- Weather because there was an up/down/up.

Originally Posted by xcitation
Subsequent pull ups IMHO were because he saw the red and black barber stripes (low speed stability protections for stall only showing in ALT/DIR LAW) on his speed and mis-interpreted as the red and back horizontal striped ladder of overspeed. The actual speed showing as being low - but he understood speed had failed. In normal law the low speed bar is amber.
Happy if someone can debunk/support my speculations
Easy, there is no low/high speed stability in Alternate Law 2b, No red and black striped "barber/ladder".
I already mentioned that many times: SPEED LIMIT FLAG ON PFDs (both).
Neither did the speed "fail" at this point; they "failed" (NCD) when sensed under 30 kt during the stall sequence, much later.
takata is offline  
Old 7th Aug 2011, 06:55
  #1713 (permalink)  
 
Join Date: Jan 2005
Location: W of 30W
Posts: 1,939
Bea int rep 3 - p76 en - p78 fr

Originally Posted by mm43
The French version of this statement on page 78 is more verbose, but has arguably led to a misrepresentation when subject to translation. However the French version is the work of reference, and I've changed the NU inputs to ND inputs in my English copy.
You're absolutely correct.
CONF iture is offline  
Old 7th Aug 2011, 07:34
  #1714 (permalink)  
 
Join Date: Oct 2009
Location: UK
Posts: 1,271
Hi DozyWannabe,
Either way it doesn't alter the fact that without over controlling, the THS wouldn't have ended up in that position,
I disagree.
According to the description given in FCOM the aircraft will attempt to maintain pitch and 1g. Even stick free, whilst climbing and decelerating - it will trim more nose up. At the apogee, as it started to fall, provided the sensed g was greater than g/2, then it would trim even more nose up in an attempt to hold 1g.

Perhaps you wouldn't find any of that confusing because it behaved as it was designed, but with a lifetime of conventional aircraft handling qualities behind me - I would most definitely find it confusing.

Normally, when the automatics don't do as we want - we take manual control (as a simple starting base) and then add the automatics gradually.
This crew never had a simple manual control starting base.

Last edited by rudderrudderrat; 7th Aug 2011 at 07:55. Reason: spelling
rudderrudderrat is offline  
Old 7th Aug 2011, 08:13
  #1715 (permalink)  
 
Join Date: Jan 2008
Location: Bangkok, Paris
Posts: 53
I usually only read and don't post here, but...

Originally posted by Old Engineer
Actually, in one version of implementation of this numbering, 32767 positive values are available, 32767 negative values, and two zero values. That's right, there are two zero values in the most simple arrangement-- plus zero, and minus zero.
Been a systems engineer for almost 30 years now, and I've never heard of a signed 16-bit implementation different from the traditional -32768 to 32767 that has exactly one representation for zero (what you call "minus zero" actually being -32768). Any reference to this particular implementation you're mentioning?

Old engineer as well.

Last edited by alainthailande; 7th Aug 2011 at 08:14. Reason: Typos
alainthailande is offline  
Old 7th Aug 2011, 08:23
  #1716 (permalink)  
 
Join Date: Jan 2005
Location: W of 30W
Posts: 1,939
Originally Posted by takata
So, in fact, the reason should be simply because all ADRs were rejected by they respective IRs (this 0211 FLR ACARS discussed previously). It will become NCD and replaced by ADR source V/S on PFDs.
On a speed base, the IRs decide to reject the ADRs, but still prefer to be replaced by them on a V/S base ... it is getting complex again ...

The IR brings its better behaviour in dynamic maneuvers while the ADR brings its stability in time (no drift of the outputs like in IR).
Now, how to qualify the AF447 event ... ?

Anyway, thanks for your research takata, that's interesting documentation.
CONF iture is offline  
Old 7th Aug 2011, 10:24
  #1717 (permalink)  
 
Join Date: Jun 2009
Location: I am where I am and that's all where I am.
Posts: 660
3holelover - indeed it was a perfect storm. It was as if the events were tailored to send that specific pilot to the briny deep.

My imagination stretches farther than yours, I guess. I remember me when I got out of college. I surprised many of my new peers with my understanding of electronics. I still needed some pointers. Then I learned why I surprised them. I saw some other raw graduates who'd never fallen in love with their chosen profession and learned, avidly, everything they could. They knew the bare minimum to hold their grades. They were good. But, faced with something outside their specific schooling left them in shock until somebody older poked them the right way. (I think I alienated a lot of fellows when I'd tinker in the corner and finish the project they were working on while I was also tinkering with the project I had at hand. Thankfully nobody ever asked me to build something into neat little boxes. We had some very good lab technicians for that, thank God! I'm not quite helpless when faced with tools and scratch building. (sigh))

He was schooled. He knew what he was doing. But the precise sequence of events threw his schooling right out the window and, it seems, "I want UP" screamed through his system as logic left.
JD-EE is offline  
Old 7th Aug 2011, 10:40
  #1718 (permalink)  
 
Join Date: Jun 2009
Location: I am where I am and that's all where I am.
Posts: 660
Smilin Ed, out of curiosity what is the rate of change for the THR in manual control mode? How fast can it move 13 degrees? If 0.65 degrees per second is as good as it gets does that adjust your distaste for the auto-trim in any way, especially if that magic number and the actual time it takes to go from neutral to stop in both directions are taught very clearly in training?

If 0.65degrees/second is as fast as it gets and the pilots know this instinctively them holding the elevator is a one control option for controlling the elevator and spinning the trim wheel. You get the same effect.

I've been wondering if there is some place on that panel to put a cartoon of the plane that shows the control surface positions as reported by the feedback position sensors. It could be ignored until a pilot wonders if something is wrong or stuck. (Or perhaps if the PF was mucking it up.)

And, please, I am just asking. Knowing is better than guessing, not that I expect to ever be writing such software. Maybe somebody else who will be designing such software will read it and consider it fully.
JD-EE is offline  
Old 7th Aug 2011, 11:11
  #1719 (permalink)  
 
Join Date: Jun 2009
Location: I am where I am and that's all where I am.
Posts: 660
alainthailande, as an engineer practicing for even more years than that I have heard of the double zero implementation. Most of the old vacuum tube er empty state electronics computers used it. Sign magnitude is not particularly uncommon until the last 30 years or so. (Gawd - 1980 is THAT long ago! Maybe more like 40 years?)

(Honest, I still have some black hairs..... I'm not all the way white! Suddenly, though, I feel staggeringly old! Um, and when I was a kid the age I am now was considered to be quite old, too.)
JD-EE is offline  
Old 7th Aug 2011, 12:02
  #1720 (permalink)  
 
Join Date: Jun 2011
Location: Devonshire
Age: 92
Posts: 300
Marketing and performance

Old Engineer #1701 wrote:
" Once the aircraft is sold rewriting of software doesn't contribute anything to the making of that sale." True enough, if that is the only sale that you as a Salesman hope to make with that Purchaser...
The Marketing Department would wish from more sales from a satisfied and profitable Customer.
I have mentioned earlier that one manufacturer, B., sent me an overnight Telex on how to load their, our, aircraft. Another manufacturer, also called B. whose aircraft we operated, could not, or did not wish to do so... and did not...
Guess which manufacturer was successful in further orders...
Linktrained is offline  

Thread Tools
Search this Thread

Contact Us - Archive - Advertising - Cookie Policy - Privacy Statement - Terms of Service - Do Not Sell My Personal Information

Copyright © 2018 MH Sub I, LLC dba Internet Brands. All rights reserved. Use of this site indicates your consent to the Terms of Use.