Some of the more recent posts have raised a few things that I thought had been settled by the two BEA reports. Looking back over the reports, nine points stuck out and I thought it was worth raising them in light of the most recent information we've got. Sorry for the monster post…
First, the problems started *before* the last known position: the first ACARS error (AP off) was received at 2:10:10 and time coded by the aircraft at 02:10. It and all of the next 14 messages are time coded 0210.
The about 6 sec between each reflects transmission time: as BEA wrote "[t]he messages were at least five or six seconds apart, which can be explained by the limited rate of communication by satellite." I assume that a period of unbroken contact of 6 secs is needed to transmit. This may be wrong -- eg it could be that a message can be sent if there is intermittent or poorer quality contact over a longer period. I also assume clocks roughly synchronized. I'm sure someone can correct if I'm wrong. If that's the case, what follows below is affected in specifics, but probably not generality.
Note that the message recording LKP (an AOC message) was received at 2:10:34 "between two maintenance messages. This can be explained by the fact that AOC messages take priority over maintenance messages.". I've assumed that it was generated while the previous message was in the process of being sent, then inserted into the queue and sent after it was finished. (I assume that if the AOC message had been generated during an even earlier transmission, it would have immediately followed that transmission, not the one it in fact followed -- stand to be corrected if wrong.)
The LKP message was received at 2:10:34 ("#0210/+2.98-30.59"). If the earlier postulate is right, it was generated at a time between 2:10:23 and 2:10:29, while the previous message was being sent, which was about 20 sec after AP off.
Second, BEA concludes likely significant orientation problems between 2:13:14 and 2:13:45, based on non-receipt of ACARS messages ("due, at least in part, to a temporary interruption in the communication link between the aircraft and the satellite"). That is a 31 second difference, but is a "gap" of 25 seconds, since the final 6 seconds or so was the time taken to transmit the message received at 2:13:45.
Now, JD-EE has recently written:
http://www.pprune.org/tech-log/39510...ml#post6354796
"A spinning condition would not necessarily stop ACARS. They were in a position for a very high angle view of the satellite. The ACARS antenna has a very big "main lobe". So a mere spin should not take the satellite out of the main lobe of the antenna. It would, perhaps, show up in satellite logs as degraded signal to noise ratios (technically speaking Eb/N0 - energy per bit divided by noise energy.)"
Postulate: between 2:13:14 and 2:13:39, although the main lobe was not pointing at the satellite sufficiently, the aircraft was likely *not* in a flat spin.
Third, there are also two earlier periods that look unusually long: 2:12:16-2:12:51 (35 sec, = 29 sec gap + 6 sec of message transfer) and 2:12:51-2:13:08: (17 sec = 11 second gap + 6 second transfer). After these periods, there are still messages coming in with time stamps of 0211 and 0212. BEA wrote: "There are two possible reasons for the longer gaps: either the aircraft did not have any messages to transmit, or it no longer had the means for doing so (loss of satellite communication performance, for example)." The fact that there were queued messages (with 2:11 and 2:12 stamps) suggests the former was not the case.
Although BEA did not positively suggest that these two gaps were also caused by orientation problems, let's mark them as possibles. An interesting pattern will emerge - see below.
Fourth, there were also some more long gaps directly after the 31 sec gap that were also not commented on by BEA. Following the message received at 2:13:45, there was the usual 6 sec difference for the next message (received at 2:13:51), but the following one was received at 2:14:14 -- 23 sec difference = a 17 second gap + 6 seconds for the message transmission. There is again a 6 second difference until the next (penultimate) message received at 2:14:20, and again for the final message received at 2:14:26.
Note that the penultimate message received was time coded 0213, but received at 02:14:20, again suggesting that there was some delay between when it was generated and when it could be sent.
Fifth, go back to the "speed or mach function" message received at 2:13:08. It is time coded 0211 but was received *after* the ADR disagree message (which was received 2:12:51 but time coded 2:12). So it was generated *before* the ADR message, but was held up.
This message *might* have a clue to altitude: BEA (report 2) says:
"The only cases of excursion outside the validity envelopes compatible with the CFR are: a CAS higher than 530 kt without the Mach value exceeding 1. This condition implies that the aircraft was at an altitude comprised between about 4,000 and 14,000 ft" or "a CAS such as the difference between the total and static pressures being lower than a given threshold. This case implies notably that the static pressure is higher than the total pressure. ... the “HARD” nature of the message indicates that the problem lasted longer than 2 seconds."
The second cause suggests bad pitots. The first cause requires consideration of speed and altitude - let's return to this below.
Sixth, the cabin advisory message received at 2:14:26 is time coded 0214. So it was likely no more than 20 secs old. It indicates "cabin altitude variation greater, as an absolute value, than 1,800 ft/min for five seconds."
This ground has been covered before. We assume that the cabin pressure was originally 8,000 feet. From report #2:
"The containers recovered closed showed that the passenger oxygen masks had not been released. There had been no cabin depressurisation. Note: Depressurisation means pressure inside the cabin corresponding to an altitude of more than 14,000 ft."
Hence, cabin was intact until at least 14,000 feet. There is no reason to think that the cabin didn't remain intact until impact, and BEA states: "The aircraft was probably intact on impact."
There has been an attempt to calculate numeric data from the message here:
The AF 447 Cabin Vertical Speed Advisory « Dark Matter I won't go into it here though; I trust mm43's post more
http://www.pprune.org/tech-log/39510...ml#post5870418. It does suggest altitude below 8000 feet.
Seventh, "[T]he last message was transmitted to the aircraft at 2 h 14 min 28 s and was effectively received". This was just after the "ADVISORY CABIN VERTICAL SPEED" message received at "2:14:26" with time code 02:14.
However, the "MAINTENANCE STATUS ADR2" message received at 2:14:14 should have been followed, one minute later, by the transmission of a class 2 fault message. This should have been received between 2 h 15 min 00 and 2 h 15 min 14. The aircraft therefore had, at 2 h 15 min 14 at the latest, one message to be transmitted. It was not received.
Report #2 notes that orientation was not the reason for this failure, based on the orientation at impact: "[t]he aircraft would therefore have been able, in the last seconds at least, to transmit an ACARS message." We know (p32) that the aircraft struck the surface of the water with a positive orientation, a low bank and a high rate of descent, and little sideslip on impact. The deformations of the fuselage frames at the root of the vertical stabiliser were not consistent with an aircraft nose-down orientation at the moment of impact. The aircraft was "in a straight line" (1.12.4).
The conclusion is end of flight between "2h 14min 26 and 2h 15 min 14". (I would have thought it should be after 2h 14min 28 sec, since that was when a message was successfully received by the aircraft.)
Eighth, contrary to recent speculation here (and the reason I went back to the reports), the vertical stabiliser was attached on impact:
"The damage due to separation from the fuselage was essentially located at the root of the vertical stabiliser". It "separated from the fuselage at the level of the three attachments". "Rib 2 was bent upwards with a right-left symmetry. The front of the fin showed signs of symmetrical compression damage … " "The vertical load pick-up arm in the rudder’s hinge axis (arm 36 g) broke at the level of the attachment lug on the rudder side. The size of this arm is calculated to withstand a maximum load of 120,000 N, corresponding to a relative acceleration of 36 g of the rudder in relation to the vertical stabilizer. … Shear cracks, along a top-down axis, can also be seen on the rudder hinge arm attachment fittings close to arm 36 g. These observations indicate that the vertical stabiliser was subjected to a load greater than 120,000 N in the rudder’s hinge axis."
36g and 120,000 N is impact damage, not free-fall damage.
At least some of the control surfaces were also attached:
"On the right-hand half elevator, four of the seven fittings attaching the elevator to the trimmable horizontal stabiliser were present. They had bottom-upwards deformations."
and
"Several parts of the flap extension mechanism fairing were found. There were marks on two of them (positioned at the level of flap track No. 3), made by the flap extension track on impact. Analysis of these marks (morphological and dimensional examinations) and comparison with an identical aircraft made it possible to determine that the flaps were in the “retracted” position at the time of impact with the water … "
Ninth, now let's look at the altitude data in the above:
- 2:10: No reason to think that altitude at 2:10 was other than 35,000 ft
- 2:11: the speed or mach message *might* have indicated altitude between 14,000 and 4,000 **OR** difference between the total and static pressures being lower than a given threshold. If the former is the cause, then the minimum average vertical velocity at this point would occur on a drop from FL350 to FL140 (assuming the aircraft is at the highest altitude in the band, ie shortest possible distance fallen) from 2:10:00 to 2:11:59 (which is the longest possible time for the message to have been generated and still have a 0211 stamp) = 10588 ft/min = 176 ft/sec. If it has dropped further, average velocity will have been much higher.
If it is the latter cause, then presumably no altitude hints from this message -- just an indication that pitots are still non functional.
- 2:15:14: latest time for impact.
If we disregard the 2:11 message for a moment, and focus on the first and last, then the smallest *average* rate of descent would be from 35,000 ft at 2:10:00 to 0 feet at 2:15:14, being 111 ft/sec. The 2:11 time is thus well above the ballpark for the calculation if it is speed, not pitot, related. Assume therefore that it is more likely to be pitot related, and gives no altitude or speed hints -- just that static was measured as higher than total pressure (still frozen up?).
Finally, where all the above comes together is putting the above into a timeline:
- 2:10:04-2:12:16 - cascade of faults generated but ACARS still able to be sent - 17 received.
- (in this period sometime between 2:12:00 - 2:12:45: ADR disagree message generated, it is received at 2:12:51)
- 2:12:16-2:12:45: (29 sec gap) possibly, orientation means no ACARS
- 2:12:45-2:13:14: (29 sec gap) orientation OK to send ACARS. 3 received. First started transmitting at 2:12:45 and was received at 2:12:51. Next two received at 2:13:08 and 2:13:14)
- 2:13:14-2:13:39: (25 sec gap) orientation problems significant enough to stop ACARS.
- 2:13:39-2:13:51: (12 sec gap) orientation OK to send ACARS. 2 received. First started transmitting at 2:13:39, received at 2:13:45. Second received 2:13:51.
- 2:13:51-2:14:08: (17 sec gap) possibly, orientation means no ACARS
- 2:14:08-2:14:26: (18 sec gap) orientation OK to send ACARS. 3 received. First started transmitting at 2:14:08, received at 2:14:14. Next two at 2:14:20 and 2:14:26.
- (in this period, possibly 2:14:20, the cabin advisory message is generated)
- 2:15:14: latest time for impact for followup ACARS fault message not to be received.
If these additional ACARS window outages are right, does this not suggest two distinct phases:
1. initially, upset -- eg pitot related -- then high vertical sink, but orientation within normal bounds that still allows ACARS. About 2 mins 12 sec in duration. (Could be a flat spin or a deep stall?)
2. then a 2min 10 sec period in which orientation varies sufficiently that ACARS is interrupted: can't send (29 sec), then can (29 sec), then can't (31 sec), then can (16 sec), then can't (17 sec), then can (12 sec).
Note that the periods are roughly equivalent, but shortening. Then some unknown time, max 54 sec, to impact, but effectively flat and able to send ACARS.
At some point, either before (within about 16 sec), at the time of, or during the first attitudinal upset sufficient to cause loss of ACARS, there is the ADR disagree, followed by PRIM 1 and SEC 1 fault - either command or a failure. (Crew trying to reset? Or everything goes down?) There seems to be a correlation between it and attitudinal upset -- is there any likely causation?
The new key piece of information is that we now know that the aircraft didn't travel very far horizontally -- perhaps 10 nm(?) (I recall that figure earlier this week). This suggests mostly vertical movement -- would there have been enough control to loop back?
More interestingly, if the above is right, what orientation would cause a cycling of available/unavailable ACARS window, if not a spin? I'm trained as a scientist, not a commercial pilot. I assume the most likely cause is pitch? More or less nose down, trying to recover from spin? Or a spiralling/barrel rolling descent mostly nose down, where the aircraft's back points towards, then away, then towards, then away from satellite? Since the aircraft strikes flat, have they finally pulled out of a dive/spiral, but have run out of sky?
(Loss of satellite contact caused by roll, while dropping vertically, doesn't sound as plausible, does it?)
(The above has borne in mind the BEA note: "message-timing by the CMC is accurate to within one minute" but "the order in which these messages are transmitted does not necessarily correspond to the associated sequence of events" and the time coded in the fault messages are with hour and minute only (p48).)