Giri-nayaka das – Slovenia: Previous forensics analysis from august 26th 1997 did the following: “The analysis procedure included computer waveform analysis, spectrographic chart analysis, FFT spectrum frequency analysis and a critical aural review of the audio.”
The following analysis gives more in depth analysis. It includes before used methods, namely:
– computer waveform analysis,
– spectrographic chart analysis,
– FFT spectrum frequency analysis and a
– critical aural review of the audio,
and in addition includes other methods, namely:
– ambient analysis
– positional analysis
– spectral phase analysis
Let me also add that I, author of this analysis, am well acquainted with Srila Prabhupada’s way of speaking, through years of hearing his recordings. I’m also well acquainted with the way Srila Prabhupada’s recorded tapes sound, them being recorded on tape machine, which was following Srila Prabhupada from day to day. Official forensic from 1997 had no such benefit.
Recording used for analysis was obtained in this way:
= = = = =
PV Editor says:
April 14, 2012at2:15 pm
Not sure what quality this version is but please try
= = = = =
1. First impressions:
I captured audio from Youtube carefully, as it is playing, without recompression, in lossless audio format 44,1Khz, 24bit.
Audio is overloaded, up until last seconds. Could be it was overloaded while it was transferred to digital. At the end level is gradually reduced, which suggests that person making transfer realized he is overloading the signal. Therefore it looks as if noise reduces at the end of tape.
NOTE! This was first impression about the change at end. After careful analysis I came to different conclusion, as described below.
It would be better to have physical tape, and have the tape re-digitized. If anybody can provide the tape, I can re-digitize it on laboratory grade audio equipment, according to highest standards.
2. Initial observation:
Audio is overloaded, up until last seconds. Could be it was overloaded while it was transferred to digital. At the end level is gradually reduced, which suggests that person making transfer realized he is overloading the signal, and he gradually reduced level.
NOTE! This was incorrect first impression gradually reducing level. After careful analysis I came to different conclusion, as described below.
3. Initial conclusion:
After transfer I checked audio for noise discontinuities. Noise print is constant throughout recording, meaning that recording is not cut together from several tapes.
Only questionable thing is the last part, where noise changes. It seems due to recording level change, and change of noise color is result of noise reduction used on recording. Would need tape to compare.
NOTE! This was first impression about reducing level. After careful analysis I came to different conclusion, as described below.
4. Background noise flow and speech flow analysis:
Captured audio file is of length 2 minutes, 15 seconds.
For ease of analysis, I will break the transcript into 16 numbered sections, each containing one sentence.
TRANSCRIPT: (my changes to transcript after careful listening are in red)
|1||Satsvarupa dasa Goswami: Then our next question concerns, uh, initiations in the future, ah, particularly at that time when you are no longer with us. We want to know how a, a first and second initiation would be conducted.|
|2||Srila Prabhupada: Yes. I shall recommend some of you. After this is settled up I shall recommend some of you to act as officiating acarya.|
|3||TamalKrishnaGoswami: Is that called ritvik acarya?|
|4||Srila Prabhupada: Ritvik. Yes.|
|5||Satsvarupa dasa Goswami: Then what is the relationship of that person who gives the initiation and …|
|6||Srila Prabhupada: He’s guru. He’s guru.|
|7||Satsvarupa dasa Goswami: But he does it on your behalf.|
|8||Srila Prabhupada: Yes. That is formality. Because in my presence one should not become guru, so on my behalf. On my order, amara ajnaya guru hana, be actually guru. But by my order.|
|9||Satsvarupa dasa Goswami: So they may also be considered your disciples?|
|10||Srila Prabhupada: Yes, they are disciples, but consider, who …|
|11||TamalKrishnaGoswami: No. He is asking that these ritvik acaryas, they are officiating, giving diksa. Their … the people who they give diksa to, whose disciple are they?|
|12||Srila Prabhupada: They are his disciple.|
|13||TamalKrishnaGoswami: They are his disciples.|
|14||Srila Prabhupada: Who is initiating … he is grand-disciple …|
|15||Satsvarupa dasa Goswami: Then we have a question conc …|
|16||Srila Prabhupada: When I order you become guru, he becomes regular guru. That’s all. He becomes disciple of my disciple. That’s it.|
Up to section 15 background is consistent. At level bursts compression pumping of noise is perceivable, indicating that some form of dynamic processing of original material was used on recording. There is perceived change in background noise at section 15. Satsvarupa asks “Then we have a question conc…” . At this point noise level starts dropping, and changes in color.
There is no cut, no discontinuity of speech. Satsvarupa stops in the middle of word, at “conc…” and prolongs last “c”. This shows, that it is not edit, but that Srila Prabhupada jumped-in with “When I order you become guru, he becomes regular guru. That’s all. He becomes disciple of my disciple. That’s all.”, which is followed by affirmative “Yes”. Fading out of noise happens during sections 15 and 16. There is no sign of any cutting or montage.
5. Waveform analysis:
Waveform is constant up to section 15. After section 15 it changes gradually, smoothly, like fade out.
Careful observation of waveform shows no discontinuities in waveform. Left/right delay is constant throughout recording, indicating that recording is consisted from one occasion only. Only change is at section 16, after Satsvarupa’s “conc…”, when level of noise starts dropping. Phase relationship between channels remains constant to the end, meaning that sections 15 and 16 are not added from another recording, but are happening in same situation of environment and position of voices.
6. Ambient analysis:
Recording is a stereo file. It is not just mono on two channels, but left and right are actually slightly delayed. Same basic material is on both channels, but its acoustic imprint is slightly different, suggesting that recording was possibly made with stereo microphone, or two microphones put together at minimum distance. Delay is cca 6-7 samples @44100Hz, giving spatial distance of cca 5cm between channels, or time delay cca 0,14 milliseconds. Right channels are delayed by this amount, putting conversation slightly to the left in stereo field. In case of stereo microphone this means all three voices were positioned slightly to the left of microphone axis.
Srila Prabhupada is nearest to microphone, then Tamal Krishna, and Satsvarupa is further from microphone in the room.
Ambient is constant throughout recording. There are no abrupt changes in ambient, which shows that recording is done in the same room, with identical position of microphone, which shows only one time of recording (meaning it is not put together form several separate recordings).
There is strong background noise in recording, up to section 15, after which it fades out. It has peaks of 63Hz, 136Hz and 203Hz, indicating that it is not Mains electricity Hum, which is 50Hz/100Hz or 60Hz/120Hz. Careful listening of nature of noise reveals, that it has machinelike nature, as a motor. Analysis of part, where it gradually stops, shows that it main frequency peaks are reduced in both level and frequency, indicating at motor that was stopped or significantly reduced in speed. (see picture below, horizontal lines at 136Hz ad 203Hz disappearing, and line at 63Hz starting to go down in frequency)
Click is perceived, right in between sections 15 and 16, which could indicate turn off of a motor, and at the same time 203Hz line starts disappearing. Sound and nature of noise indicates that it could be a ventilation fan. Its noise is present throughout the recording, up to section 15. This presents unique noise footprint, which makes it impossible to montage recording from several different takes. This means that recording is made in one take, in same room, in same microphone position. The machine noise’s phase response is constant throughout recording, up to section 15, and gives proof of continuity f sections 1 to 15.
This conclusion refutes my previous initial observation, that record level was reduced due to overloading. Noise is coming from a machine, which was turned down, gradually reducing its speed, as seen from above spectrum graph of this section.
7. Spectral analysis:
Spectrum of background noise and speech is constant in character throughout recording. Only change is after section 15, after source of background noise is turned off, reducing its intensity gradually both in level and in frequency. (see picture below, spectrum of entire recording)
There is one notable gap at 1:20, in section 10, just slightly after Srila prabhupada saying “…, who…”. There is a click, followed by word “who” after 133 milliseconds. Gap starts 42 milliseconds after beginning of word “who”, and lasts for entire duration of word “who”. Because clic, start of word and gap don’t start at exactly same point, they cannot be due to cut in tape. Base on its length, shape and position it seems to be compression artifact, caused by tape recorder’s noise reduction system because of click (bump?) and word “who” appearing almost at the same time. (See picture below, zoom of section with gap)
In picture below, the gap is seen zoomed in on relevant spectra, and Srila Prabhupada’s “who” is seen as “wiggle” in the bottom centre of picture. Just before “who” there is a vertical line of lighter color, showing a “click”, possibly bump to microphone or some other “click”. It is visible that click, “who” and gap do not appear at the same time, so this is not a cut, or edit.
Besides this gap, there are no spectral discontinuities in recording, which means, it is not montage from several recordings, but one consistent recording. Absence of such discontinuities can also be taken as proof, that tape was not edited in analog, by cutting the tape or by overdubbing or by re-recording selected parts one after another in different order than it was recorded. It can be concluded, that recording is intact as a whole.
8. Positional analysis:
Relationship between phase of left and right channel in recording creates position of sound source in virtual sound plane. This is very sensitive to any changes in timing and phase, and any alteration in position of sound sources shows as discrepancy. It is a kind of footprint of recorded material, in regard of position of voices in recorded space. Any change in position of voices in space alters phase imprint.
In this recording, phase is consistent throughout recording. All three voices maintain positional characteristics throughout recording, meaning that recording is made with microphone in constant position, and speakers maintaining position.
This shows, that entire conversation happened in the same room, with voices being in same positions, without moving throughout the room. From this it can be concluded, that recording took place on the same day, at the same time.
9. Spectral Phase Analysis:
This is another analysis, which is very sensitive to any change in spectrum/phase relationship. It is another footprint, combining phase and frequency response, and takes into consideration phase of individual frequencies, and compares them across recording. If recording is montage from several recordings, it shows. On this recording spectral phase is consistent throughout recording, up to section 15, after which there is change due to background noise source being turned off.
This shows, that recording is made in the same room, with same microphone and speakers position. From this we can conclude, that it is not montage, but happened in one time and place.
After analyzing this recording carefully, conclusion can be made, that entire recording happened in one room, at one time. Entire recording material is of same sonic imprint, there are no fluctuations in the timbre of material. Conversation is flowing consistently, with no perceivable interruptions. Sound spectrum is consistent throughout recording, up to section 15. After section 15 certain source of noise is turned off, with gradual decline in level and frequency, indicating on mechanical slowing down, possibly from ventilator or similar motor operated device.
There are no perceivable cuts, which would indicate that material was intentionally cut. There is one spectral gap in section 10, as described in (7), but it is not due to cut or montage, because events around it happen in succession one after another.
It is impossible to claim that recording is all in all, all original parts of it included, and that no part of conversation was taken out, meaning that recording was shortened in some way. But there are no signs showing such cutting. Conversation is running smoothly, pauses between phrases are natural, rhythm of conversation seems natural. While Srila Prabhupada is speaking, there are often heard quiet responses of Satsvarupa and Tamal Krsna, ond opposite, such as “yes, mhm, hmm…”. Voices often overlap, making it hard to make unnoticeable cuts on analog equipment which was available cca 30 years ago.
In conclusion, recording appears to be one compact unit, without alteration. It happened as it is, quite possibly in one piece, certainly in one room and in one time.
Equipment used in this analysis:
– digital audio workstation ( DAW) Steinberg Cubase for capture
– EM-U audio card, to route youtube video digitally to DAW via ASIO
– Adobe Audition software for preview
– Diamond Cut Forensics ver 8.10, 2011 specialized software
– monitors Yamaha NS10M Studio for listening
– monitors Dynaudio Acoustics BM5
– headphones AKG K701
– headphones Beyerdynamics DT 770 Pro
– headphones Sennheiser HD 580
Analyzed by, Giri-nayaka das, April 14th, 2012