TWA Flight 800
The Four Missing Seconds Analysis from the
Flight Data Recorder
26 September, 2000
Glen H. Schulze
Littleton, Colorado
Section One
Introduction
Condensed Version of the 4 Missing Seconds Analysis, TWA FL 800 FDR Tape 26 Sep., 2000
When the TWA FL 800 43 data word subframes were blocked into 64 word record blocks a repetitive and orderly set of word group patterns were generated within each record block. This blocking, necessary to support the burst-recording and Checkstroke FDR functions, produced 43 different 64 word block groups each with a unique set of words. Of these 43 unique word block groups only one block starts with subframe word 1, only 1 block starts with subframe word no. 2, etc. Also, only one block ends with subframe word no.1, only one block ends with subframe word no. 2, etc.
The NTSB published their Tabular Data in a format, which preserved the record block groups, separated usually by horizontal lines in their Data Tables. Although the NTSB did not publish the entire 43 word subframe contents recovered from the 64 word record blocks of FL 800 they did publish a significant number of flight parameter words. This enabled analysts to easily identify the unique word block group associated with each record block entry in the NTSB Tables.
By arbitrarily defining the unique record block, which starts with subframe word no. 1 as Superframe Block no. 1, we find the last NTSB entry block in their FL 800 Tabular Data Charts No.1 and No. 2 is Superframe data block no. 31. Therefore, the implication is the end of the FDR data stream occurred at the end of data block 31 in the 17th and last Superframe of FL 800 data.
But NTSB Addendum No. 2, Fig. 3, shows a different endpoint of the FDR tape. This Fig., an actual pictorial of the waveforms found at the very end of the FL 800 tape, is annotated with the following "Bits 1-9 of Word 4". This last word 4 was followed by an 8 bit postamble thereby verifying that word 4 was the last word in the last FL 800 record block. But, only one unique record block ends in subframe word 4, and it's block 35, not block no. 31!
Accordingly, waveforms for data record blocks 32, 33, 34 and 35 exist pictorially at the end of the FL 800 FDR tape but their flight parameter values have been withheld from the public dockets by the FBI/NTSB.
The flight parameter values in these last 3,000 bits of the FDR data stream, particularly the vertical and longitudinal accelerations, may well hold the answer as to what happened in these last four seconds of FL 800 to cause it's disintegration at 13,700 feet.
Note: The terms "data record block", "data block", "record block" and "block" are
used interchangeably in this analysis.
Section Two
Background for the 4 Missing Seconds Analysis for the TWA Flight 800 FDR Tape
The TWA FL 800 Flight Data Recorder, FDR, tape-recorded a very unique frame of flight parameter data. One DATA SUBFRAME contains 43 twelve-bit words, allocated over 18 flight performance categories and 10 documentation or "housekeeping" parameters. See NTSB Exhibit 1.
The TWA FL 800 FDR "burst-mode" recorded the 43 word data subframes after they were serially assembled into DATA RECORD BLOCKS, each record block comprising 64 twelve-bit words. The FDR recorded these data blocks at the average rate of one block per second, but at a burst rate writing speed over 11,000 bits per second at 5-ips tape speed.
A Checkstroke QC feature of the FDR provided for stopping and rewinding the tape after the recording of each one-second data block. During rewind, the last data block just written was read back and it's contents verified. Subsequent data blocks were then recorded after insertion of small inter-record blank gaps on the tape.
Before recording, all 12 bit-word formatting was organized into a single serial bit stream, which was recorded onto the active single tape track of an 8-track 1/4-inch wide tape. The serial bit stream record blocks contained 12 X 64 or 768 data bits per second, at a one block per second rate. See NTSB Exhibit 2.
In addition to the 768 data bits each record block contained a directionally symmetrical set of pre- and postambles, each containing 8 bits, i.e. 0000 0001/64 data words/1000 0000. These fixed 8 bit words were electronically attached/spliced to the beginning and end of each data block, before recording, for ease of establishing playback data block boundaries.
Playback processing and publishing of the FDR data by the NTSB was accomplished in unique and orderly one-second data block groupings. In each published one-second data block, 64/43 data subframes were included, i.e. about 1 1/2 data subframes for each data block grouping. For ease of reading, the NTSB published only selected flight parameters, with no documentation parameters, in the one-second data block groupings. See NTSB Exhibits 3A and 3B.
The FDR tape contains no time track per se. However, GMT months, days, hours, minutes and seconds are embedded in the FDR data frame as documentation words 4 through 8. Although the FL 800 GMT words were never decoded and published by the NTSB they were found to be cyclically and properly incrementing by citizen inspection of Freedom Of Information Act FDR waveform sets. The NTSB reluctantly and belatedly released this FOIA data. (Ref. 1)
The NTSB time tagged each of their published one-second data block flight parameter groups with a time-of-flight in hours, minutes and seconds derived from NTSB microphone keying event comparisons with the CVR transcript. The CVR tape contained no time track per se and CVR transcript time was established by NTSB speech event comparisons with the Boston Air Traffic Control Tower Tape. The ATC tower tape does contain an "E code" time track containing Julian days, hours, minutes and seconds.
FL 800 FDR PCM and Tape Format Synopsis
Channel Code: Bi-phase mark
Coding: BCD, binary coded decimal
Bit order: LSB first
Character order: LSC first
Sync word: 0111 111 1110
Pre/post ambles: 000 0001, 1000 0000
Words/Subframe: 43
Words/Record block: 64
Superframe: 43 data blocks
Superframe Start: Subframe word 1, in word 1 position of the record block
Average record data bit rate: 43 X 64 bps, one record block per second
Burst record bit rate: 11,400 bps
Tape write speed: 5 ips
Tape width: 1/4 in.
Tape tracks: 8
Record block length: 0.344 in.
Inter record gap length: approx. 0.07 in.
Record time: 3 1/8 Hrs per track, 25 Hrs per tape.
Tape track pattern: Sewing machine style, FWD, REV, FWD, etc.
Erasure gap: approx. 3 in.
NTSB Exhibit 1, Parameter List, p. 9, FDRFR Rev. 1, Feb. 15, 2000
NTSB Exhibit 2. MagnaSee Footprints, Fig. 2, FDR Add. 2, Feb. 15, 2000
NTSB Exhibits 3A and 3B. Tabular Data No. 1, p. 41, 42, FDRFR Rev. 1. Feb. 15, 2000
Ref. 1, FOIA 99-297
Section Three
TWA Flight 800 Data Word Playback Location Patterns
The playback 43 word DATA SUBFRAMES, when serially assembled and blocked into 64 word RECORD BLOCKS, produce unique word pattern groups on playback.
These unique patterns are lost if the playback processing is "seamless" with time, i.e. the pre- and post-ambles are discarded and the one-second record blocks are rejoined/spliced perfectly together with the inter-record playback time gaps removed. The playback processing preserves the data word inventory and word group patterns only when the data words are presented in one second, stand alone groups of the recovered record blocks.
The NTSB Flight Parameter Tabular Data No. 1, Revised 15 Feb., 2000, does present the processed flight parameter data words preserved in the one-second groups or record blocks in which they were burst recorded. This enables the playback word group patterns to be used as a potential embedded elapsed time measurement tool.
This potential is confirmed by inspection of NTSB Exhibit 3A, which shows a unique pattern in the lower page right side for data words 34, 35, 36 and 37, the 4 Engine Pressure Ratios. After appearing in preceding one-second data blocks in a clear and repeatable inventory pattern of 1, 2, 1, 2 , 1, 2-- the pattern is uniquely altered to 1, 2, 1, 2, 1, 1, 2, 1, 2 --. Moreover the unique "1, 1 " inventory pattern can be seen to move sequentially and orderly in time from word 37, to 36, to 35, to 34, in perfect descending order of word number sequence. See highlighted words in NTSB Exhibit 3A.
Inspection of the Tabular Data No. 1 contents precisely 43 data blocks earlier shows an exact duplication of the unique 1, 1 appearance pattern for words 34, 35, 36 and 37. This discovery indicates that the data word playback location patterns have a 43 record block and 43 second periodicity. See NTSB Exhibit 4A & 4B, ( highlighted words).
These discoveries led to the construction of a FL 800 SUPERFRAME Matrix showing the detailed appearance patterns of the 43 data subframe words in the 64 word data record blocks. See GS Exhibit - 1.
This SUPERFRAME Matrix was "synchronized" or reset to a convenient start pattern by initially allowing data subframe word 1 to start in record block word 1 position. This synchronization allows the individual one-second data record blocks to be sequentially numbered from to 1 to 43. At sequential block position 44, data subframe word 1 will reappear in word 1 of the one-second data record block, starting Superframe No. 2, etc.
This SUPERFRAME Matrix was then used as a check or inspection tool to see if the NTSB Tabular Data No. 1 of flight parameter word grouping patterns on playback conformed to the constructed Matrix. The word pattern groups matched perfectly.
NTSB Exhibits 4A and 4B, Tabular Data No. 1, p. 39 and 40, FDRFR, Feb. 15, 2000
GS Exhibit 1, TWA FL 800 Last Superframe Word Patterns, June 23, 2000
Section Four
The Various Versions of the Final One-Second Data Block from TWA Flight 800.
The initial version of the processed FDR tape Flight Parameter Tabular Data No.1 ended with the last one-second data record block grouping containing an expectedly aberrant set of flight parameter engineering values. This expectation was based on the lead FBI investigator announcing early on that both black box tapes ended with "not normal" signal contents. See NTSB Exhibit 5, (highlighted words).
Shortly after publication of Exhibit 5 in Dec., 1997, parameter Tabular Data No. 1 was revised and the last one-second data record block, still legible, was lined through the center with a note "End of FL 800 Data". See NTSB Exhibit 6, (highlighted words).
The NTSB later explained this redactive editing as resulting from their inadvertently processing the FDR tape as if the FL 800 data had not stopped at the leading edge of the three-inch erase gap. This failure to stop the processing at the end of the last valid data record block before the erasure gap resulted in data from FL 803, recorded 25 tape-hours earlier, being blended with the last data from FL 800. The NTSB mistake in publishing the unsynchronized FL 803 data, as if it were the last "not normal" data from FL 800, accounted for producing the anomalous values in what was originally published as the last one-second data record block from FL 800, Ref. CVR Time 20:31:12.
The NTSB later made the elimination of the last 800/803 blended data record block from the Parameter Tabular Data No. 1 table complete by total redaction of this anomalous data block. See NTSB Exhibit 7, (highlighted area).
The last complete one-second data record block in Exhibit 7 then showed a set of flight parameter engineering values totally in agreement with the immediate predecessor blocks of flight data, i.e. no anomalous or aberrant values. Also, this block contains a unique and expected word group pattern consistent with it's immediate predecessor data record blocks, i.e. no evidence was found of time or data record block sequence discontinuities. This fact was established using the data record block playback word group patterns as an embedded elapsed time measurement tool.
Accordingly, using the most current NTSB Tabular Data No. 1, the last valid and legible one-second data record block word pattern from FL 800 clearly matched the 31st sequential data block word pattern in Superframe 17! See GS Exhibit -2, (highlighted).
The last word in data block 31 is always data word 6., i.e. GMT hours!
NTSB Exhibit 5, p. 42, Tabular Data No. 1,Created Jan. 09, 1997; Report Feb. 28, 1997
NTSB Exhibit 6, p. 42, Tabular Data No. 1,Created Jan. 09, 1997; Add. 2, Feb.15, 2000
NTSB Exhibit 7, p. 42, Tabular Data No. 1, Rev Feb. 15, 2000; Rpt.Rev.1, Feb.15, 2000
GS Exhibit -2, TWA Fl 800 Last Superframe Word Patterns-Extended, Jun. 23, 2000
Section Five
Firmly Establishing the Last TWA FL 800 Data Record Block as Block 31 of Superframe 17
(Based on Flight Parameter Tabular Data No.1, NTSB Exhibits 3A & 3B.)
Rigorous proof is required to support the stated goal of this section of the analysis.
This proof is supplied by using the unique and orderly patterns of the number of times a particular flight parameter word is expected to appear, data block-by-data block, as predicted by the Superframe Matrix, GS Exhibit -1. This method can be called a "unique word inventory pattern" based analysis. We also use a supplemental "record block continuity" method for time saving.
The fact that flight parameter data words do not always appear an equal number of times, from data block-to-data block, provides a recognizable change in pattern useful for data block aligning/synchronizing the NTSB Tabular Data with the GS -1 Matrix. The following analysis steps will illustrate the use of the "unique word inventory pattern" analysis tool, as well as providing the necessary rigor for the "Block 31 of Superframe 17" statement of above.