BOLT BERANEK AND NEWMAN INC.
18 February 1983
G. Robert Blakey
Professor of Law
Notre Dame Law School
Notre Dame, Indiana 46556
This letter conveys our suggestions for further studies of the acoustical evidence that we examined for the House Select Committee on Assassinations.
We have read the report that Norman Ramsey's committee wrote for the National Academy of Sciences (NAS). That report can be divided into two parts. one part analyzes some data that pertain to the Barber hypothesis, while the order part contains opinions about our report. Only within the former part do we find important and original information.
Barber discovered a very weak spoken phrase on the DPD Dictabelt recording that is heard at about the time of the sound impulses we concluded were probably caused by the fourth shot. The NAS Committee has shown to our satisfaction that this phrase has the same origin as the same phrase hears also on the Audograph recording. The Audograph recording was originally made from the Channel 2 radio. The common phrase is heard on Channel 2 about a minute after the assassination would appear, from context, to have taken place. Therefore, it would seem, and the NAS Committee concludes, that the sounds that we connected which gunfire were made about a minute after the assassination shots were fired.
Upon reading the NAS report, we did a brief analysis of the Audograph dub that was made by the NAS Committee and loaned to us by them. We found some enigmatic features of this recording that occur at about the time that individuals react to the assassination. Therefore, we have doubt about the time synchronization of events on that recording, and so we doubt that the Barber hypothesis is proven. The NAS Committee did not examine the several items of evidence that corroborated our original findings, so that we still agree with the House Select Committee on Assassinations conclusion that our findings were corroborated.
It appears to us that further analysis is needed to decide whether the Barber hypothesis is correct. If it can be proven that our indications of shots do not coincide with the assassination, then that would invalidate our findings. But if this cannot be proven, then it still seems necessary to further the acoustical analysis to increase confidence in the conclusions to be derived from the acoustical evidence. In the following paragraphs we present our ideas on how these further studies should be done.
A. STUDY ORIGINALITY AND INTEGRITY OF RECORDED ACOUSTICAL EVIDENCE
1. Phase demodulate the power hum on both recordings.
The phase of the 60 Hz power mains to which both recorders (Dictabelt and Audograph) were connected would contain modulations (perturbations) caused by various electrical equipment on the grid. We recommend both channels to be analyzed for phase modulation using a velocity tracking phase lock loop (PLL) which uses an analog phase detector and a servo response optimized to estimate 60Hz phase deviations within about a 100Hz modulation bandwidth. The PLL should be preceded by a very sharp bandpass filter, centered at the power frequency (nominally 60Hz).
2. Compare phase modulation on the two channels as a function of time.
The Channel 2 recording is discontinuous in time, owing to the voice-actuated recording switch. The task is to match the phase modulations on the two channels, so as to determine where each part of the Channel 2 recording fits on the Channel 1 recording - if anywhere.
3. Study the gain discontinuity and power-hum harmonic-strength discontinuities seen on the Channel 2 recording at about 12:30 (nominal time of the assassination).
Look for phase discontinuities in hum here. Look for other indications of a splice or overlay.
4. Look for indications of double recording on both channels, by examining the number of hum signals that have been recorded.
This can be done by very narrow filtering or by looking for beats in the hum waveforms.
The results of these four studies should determine whether both recordings are single recordings or whether they have been superposed. The correct time alignments between Channel 1 and the various segments of the Channel 2 recording should be established also. If the common phase was recorded originally and simultaneously on both channels, and if both recordings are faithful copies of the originals, then the Barber hypothesis is true. And if true, the recorded sounds that we decided were probably gunfire would not be gunfire. But, if not provable as true, then the truth must be found as the most likely choice between the Barber hypothesis, our interpretation of the acoustical evidence, or any other interpretation that may enjoy corroboration of any kind.
The studies listed under "A" can probably show whether the Barber hypothesis is proven or not. If not proven we recommend studies "B".
B. ENHANCED ECHO PATTERN MATCHING
Owing to time and funding limitations, our original study devoted only about three days to the process of matching the acoustical reconstruction echo patterns with the sound patterns we had found on the DPD recording. This study "B" would be undertaken to add some additional features to the matching process and to perform the matches in a systematic way on a computer. We suggest the use of crosscorrelation, where both waveforms (echo patterns and DPD recording) would be pre-processed as follows: Each 10 ms time window would be characterized by the amplitude and phase of the largest peak occurring within the window. Amplitude would be quantized into two or three values, separated by about 10 dB. The best quantization would need to be studied in light of the expected uncertainty in the directive pattern of the motorcycle-mounted microphone.
The matching process would amount to the continuous crosscorrelation of each of the echo patterns (about 500) with 60 sec of the DPD recording (6000 crosscorrelation values per echo pattern). The results of this effort would provide the data necessary to recompute the probabilities that four shots are uniquely revealed by the DPD recording. In other words, if the Barber hypothesis cannot be proved, the results of this study would increase the confidence with which the acoustical evidence can be said to reveal gunshot times and origins. If the results show that gunshot sounds are probably not on the DPD tape, then our original conclusions would be rejected. If the results show that gunshots probably are on the tape, then we suggest the following studies "C" to further increase confidence in the result, and in particular, the conclusions about the third shot.
C. ADDITIONAL WORK OF THE WEISS AND ASCHKENASY (WA) TYPE
Owing to time and funding limitations, Weiss and Aschkenasy applied their analytical extension of our pattern matching procedure to the third shot only. We suggest that their procedure be applied to the other three shots as well. The WA procedure will need to be extended to include the third (vertical) dimension for the TSBD-originated shots, because the source-reflector-receiver is not in a horizontal plane. There is little doubt about the origin and approximate timing of these other shots and so if their procedure works for them too, then their procedure will be validated independently of the controversial third shot.
We found that WA had placed the onset of the third shot echo pattern at about 300 ms earlier than had our pattern matching procedure. The NAS Committee decided that this inconsistency was an error, and therefore supported their view that either or both of the tests were wrong. We pointed out to the NAS Committee that both the relevant signals on the DPD tape and the echo pattern contain two similar bursts of echoes that are separated by about 300 ms, and that we had simply matched on the second. The second burst is an echo of the first burst after its sound is reflected from the face of the DCRB, which is nearly perpendicular to a line connecting the Knoll with the microphone and about 150 ft from the microphone. We suggest that the WA procedure be repeated with explicit attention paid to this reflecting feature.
If the results of "C" show that the other three shots do not allow an accurate position of the motorcycle microphone to be found, or if the positions that are found are inconsistent with our knowledge of the motorcycle trajectory, then the truth of the third shot from the Knoll would be in question. If the results do produce an acceptable motorcycle trajectory, then the truth of the third shot from the Knoll would be greatly enhanced. In this latter event, we suggest studies "D" that are designed to improve the credibility of the acoustical evidence by answering questions that have arisen about the compatibility of our interpretation of the acoustical evidence with non-acoustical evidence.
D. COMPATIBILITY OF ACOUSTICAL WITH NON-ACOUSTICAL EVIDENCE.
1. Comparison of shot instants with "jiggle" analyses.
There are at least three separate analyses of the jiggle of the Zapruder camera, presumably in reaction to the gun shots. These three analyses have yielded different results, so our first suggestion is to compare them. A quantitative analysis of camera/operator motion induced by both muzzle blast and shock wave is needed first to enable those jiggle features to be identified that are most likely to have been caused by gunfire directly - as opposed to features caused by psychomotor action.
Once a best estimate is made from the jiggle data about the timing of each event that could be a shot, then these times would be compared to the times found by studies "C". These comparisons must include tape speed corrections as well as time-of-flight corrections.
2. Estimate motorcycle speed
We believe that a harmonic analysis of the amplitude modulation on the motorcycle noise can determine the motorcycle speed as a function of time on the DPD recording (with ambiguities due to gear position). We suggest that this analysis be done, to show the speed trajectory during times around 12:30. This trajectory would establish whether the motorcycle was moving along in the motorcade or not. We note that if the NAS Committee were correct, then the now-apparent slowing of the motorcycle did not occur until about one minute after the assassination. If this analysis confirms the now-apparent slowing of the motorcycle, then the NAS idea that the motorcycle radio passed a Channel 2 receiver to hear the "hold everything" would be voided.
3. Measure doppler shift on siren sounds
A few minutes after the assassination, the sounds of sirens are heard on the Channel 1 recording. There appears to be a doppler shift as the sounds first intensify and then fade away. We suggest that the instantaneous frequency as a function of time be estimated and plotted. From this can be estimated the closest point of approach and relative speeds of the motorcycle and the siren vehicle. These estimates will enable one to chose among the various explanations of the origins of the siren sounds.
4. Photographic examination of McLain's trajectory
It is important to compare our best photographic estimates of McLain's trajectory with our best acoustical estimates - those from "C". We understand that further work had been done here, but not having seen it we don't know if an adequate analysis has been done.
You have also asked us to estimate the cost of doing the studies that we have suggested. We offer the following figures as a guide. It is assumed that those who do the work have the necessary equipment and skills in hand.
A.1 1 mm (scientist) and 1 mm (technician)
A.2 1/2 mm (scientist)
A.3 1 mm (scientist) and 1/2 mm (technician)
A.4 1/2 mm (scientist) and 1/2 mm (technician)
B 5 mm (programmer) and 4 mm (scientist)
C 2 mm (scientist) and 6 mm (technician)
D.1 2 mm (scientist) and 2 mm (technician)
D.2 1 mm (scientist) and 3 mm (technician)
D.3 1 mm (scientist) and 2 mm (technician)
D.4 1/2 mm (scientist) and 2 mm (technician)
As a rough estimate, the cost of a programmer or a technician would be about $50/hour, and the cost of a scientist would be about $70/hour.
The NAS Committee report contained a list of 13 studies that they suggested could shed additional light on the subject. We are in general agreement with their suggestions. All of their suggestions are incorporated in some way in our suggestions, except for their Nos. 2, 6, and 10.
We hope this list of topics for further study will be useful to you. Although we do not wish to perform any of these studies ourselves, we will discuss our ideas with whichever investigators you decide upon.
Very truly yours,
James E. Barger