APPENDIX 3
Curves 1 (208 Kb!)
Curves 2 (195 Kb!)
ANALYSIS OF THE ABNORMAL VALUES (PEAKS) NOTED FOR SOME PARAMETERS OF THE DFDR
A detailed study of the DFDR read-out shows that these peaks were not due to variation of the actual values during the flight but rather to de-synchronisations of the signal during read-out.
1. How a DFDR works
To explain this de-synchronisation phenomenon, it is necessary to review DFDR data acquisition and read-out. Sensors on board the aircraft enable to get various data. The data is transmitted to computers that deduce the value of a number of parameters which are characteristic of the flight.
Those parameters (in analogic form) are then coded, digitised and multiplexed by a job-oriented computer, the Flight Data Acquisition Unit (FDAU) and transmitted to the recorder itself as a sine wave form signal.
Multiplexing consists in presenting one after the other (in the form of a continuous signal) various parameters which are obtained simultaneously.
As far as the DFDR is concerned, all parameters are given in frame format, each parameter being repeated at the same place (or same "word") in each frame. The frame used by the FDAU of the DFDR is a four-second one, each frame being itself divided into four one-second sub-frames. Each sub-frame must contain 768 bits (0 or 1), i.e. sixty-four 12-bit words, and must begin by a specific word called "synch word".
At DFDR read-out, a series of computers decode the data in the reverse way they were recorded. So, after the signal is read, a computer cut it and convert it into bits and words. Then those words are demultiplexed so as to be finally translated by one more computer into actual values (the flight parameters).
During the demultiplexing operation, there is a verification of the binary signal: the demultiplexing computer searches for all the synch words (as defined above) and verifies that the number of bits between them is correct. If, for any reason, (damaged tape, bad reading, etc...) a synch word is not to be found at the right place, the computer will warn that there was de-synchronisation when that sub-cycle was concerned; in such a case, the conversion of the binary signal into actual values will be altered and the listings and graphics obtained subsequently will show abnormal values. There will be synchronisation again as soon as the computer detects two synch words that define two consecutive sub-frames, separated by the right number of bits.
2. Analysis of the de-synchronisations
Let (tf) be the last recorded second of the flight; a first de-synchronisation is noted at (tf-14 seconds). A more accurate analysis of the binary signal read at that second makes it clear that 2 bits are missing in the sub-frame (766 instead of 768 bits) and that these bits were lost at the beginning of the sub-frame (starting from the fifth word). All actual values calculated later on are therefore distorted (more than 90% of the data contained in that second are lost).
A second de-synchronisation occurs at times (tf-11s) and (tf-10s). These two seconds are de-synchronised because the computer has not found the synch word which is between the two sub-frames. Nevertheless, the analysis proves that the first 46 words (out of 64) at second (tf-11s) are consistent; so are the last 60 words at second (tf-10s). Therefore something wrong happened between the forty-seventh word of time (tf-11) and the fourth word of time (tf-10).
Finally, a last de-synchronisation occurs at time (tf-5s). During that sub-frame, the computer counted 770 bits (i.e. two additional bits). As early as the sixth word of that frame, there is inconsistent information (one additional bit at that moment). If we make the assumption that it is one bit too many, consistent data is to be found again, through manual calculation, until word '9, and then there are again abnormal values.
In this way we note that, several times along the fourteen seconds before the end of the recording, the binary signal calculated by the DFDR shows alterations that bring about a translation in actual values that is utterly wrong and in no case representative of the actions of the aircraft at these moments.
Another reading of the DFDR was made with a different tension of the tape on the playback heads. It was then noted that two out of the three de-synchronisations of the tape end had disappeared (those at times tf-14s and tf-5s). The parameters recovered for those two seconds were perfectly consistent with the rest of the flight.
De-synchronisations on one listing, partial recovery of consistent data on another ones are the proof, if needed, that these problems are due to the bad condition of the tape.
3 . Explanation of these de-synchronisations
The DFDR recovered after the accident was particularly damaged (impact evidence, rounded sides,etc). When it was opened, BEA specialists noted that the thermal insulation was damaged and, furthermore, that the outer loop of tape was severed, that it showed folds and that is had come out of the roller transport guides. We may suppose that that section of the tape was cut at the ground impact, because it was less protected, and that the tape slackened suddenly and hit mechanical parts (rollers, playback head, etc...) The signal on the tape may have then be damaged; this deterioration of the sinusoidal signal may have led to a defective binary transcription.
This DFDR works in such a way that the last recorded seconds are precisely located between one of the erasing heads and the corresponding far left roller. Between those two points, the tape Ad about 30 cm long. Since the tape recording speed is 0.43 in./s the last twenty-seven seconds, or so, are on this section.
To investigate that problem, BEA carried out an extensive analysis of the end of the original tape . The use of a detector material on that tape section made the following apparent :
The beginning of the detected blank space precisely corresponds to the end of the flight recording . 5.2 cm before, ( i .e . a 4. 7 s. recording), an important folding enables an explanation of the de-synchronisation found at time ( tf-5s ) on the first listing. Then, 11.1 cm before the end of the flight (i.e. a 10.2s recording),, the tape was cut, which is an indubitable explanation to the de-synchronisation between times (tf-10s) and (tf-11s). Lastly, 15 cm before the end of the flight (13.7s), the fact that the tape was slightly creased also explains the de-synchronisation at time (tf-14s) in the first listing.
APPENDIX 4
Model showing aircraft detachment
APPENDIX 5
Distribution map of the debris
APPENDIX 6
Distribution map of the major parts of main wreckage (Referenced C Appendix 4)
IDENTIFICATION OF THE MAJOR PARTS OF MAIN WRECKAGE
