My professionally-qualified conclusion based on the audio characteristics of the recording we have heard, is that it was most likely recorded from within the White House at the behest of Trump, via Mark Meadows and at least one technical worker, and that Meadows was likely behind the release of the recording. There are numerous possible explanations, about which more below.
I’m posting this story “for the record” in case it can contribute anything to journalists etc. who may be looking for further details of the events. This is in response to speculation about who made the recording and who released it.
Sources and labeling:
This story is based on the following sources of audio, that are posted on YouTube. After Youtube.com/ (after the forward-slash) copy & paste the following:
watch?v=o3hrN0cP58Y 4-minute 31-second exerpts posted by Washington Post. This will be referred to as “WaPo” below.
watch?v=AW_Bdf_jGaA 1-hour 2-minute 20-second entire recording posted by NBC News. This will be referred to as “NBC.”
watch?v=FBIG5Tv0fZk 1-hour 2-minute 31-second entire recording posted by TV station 11 Alive. This will be referred to as “11 Alive” below. Note, this appears to begin slightly earlier than the NBC full audio.
(I didn’t embed the links because a) embedded links have potential security issues and should be discouraged generally, and b) that would bloat up the page and waste storage capacity on your device; you can still go to the linked videos if you choose.)
Observations will be labeled O-(number) (capital letter O), Interpretations labeled I-(number), and Conclusions labeled C-(number).
O-1: Call setup. [Yes this is important.]
11 Alive’s audio begins 31 seconds prior to NBC’s, with conversation during setup of a conference call.
Male voice, not Mark Meadows: “The conference is now connected.”
00:00:11: Male voice later identified as Mark Meadows: “Mr. Raffensberger are you on the line?”
8-9 second pause, then different male voice at 00:00:20, “He was here a second ago.”
“This is Ryan Germany.”
Male voice: “Mr. Germany, Ms. Fuchs...”
Male voice, 17 seconds after Meadows asked, at 00:00:28: Raffensperger speaks up: “Brad Raffensperger.”
Meadows: “Mr. President, everybody is on the line, just so, this is Mark Meadows, the Chief of Staff. Just so we all are aware, on the line is Secretary of State and two other individuals, Jordan and Mr. Germany, with him you also have the attorneys who represent the President, Kurt and Alex and Cleta Mitchell, who is not the attorney of record but has been involved, myself, and the President...”
O-2: Audio volume and quality at start of call:
This also from 11 Alive audio:
First male voice: Medium volume level. Landline-grade audio quality.
Meadows: High volume level. Landline-grade audio.
Germany: Low volume level. Possible cellphone-grade audio.
Raffensperger: Medium volume level. Audio has emphasis in the bass frequency range, at or slightly below the bottom of the audio bandwith of telephony.
Trump: High volume level, landline-grade audio. There’s a slight room reverb on Trump’s voice. That reverb is not present on Meadows’ voice or any other voice in the “setup” phase of the call.
Note, audio bandwidth of conventional telephony is the range from 300 to 3,000 Hz. (cycles per second). This refers to frequency (pitch) and is not bit rate.
At 00:13:21 in 11 Alive audio, Mark Meadows chimes in at the same volume level as Trump but without the room reverb.
O-3: Signal artifact on NBC audio not present on 11 Alive:
At 00:12:28 on NBC audio, there are “ticks” or “clicks” or “scratches” in Trump’s audio as he speaks, and more of this shortly after when Meadows speaks. These sounds continue on the NBC audio but are not present on the 11 Alive audio.
This sound is clear at 00:12:37 where Trump says “far more certified than we need...”, on the words “we need.” This artifact is not present on 11 Alive’s audio at the same point in 11 Alive’s recording at 00:13:07.
This effect becomes more pronounced when Mark Meadows speaks on NBC’s recording at 00:12:50, “so Mr. President if I might...” particularly on the words “I might.” However this effect is not present on 11 Alive’s audio at the same relative point, at 00:13:22.
If you listen over headphones, the noise is located dead-center in the stereo field, “even more monophonic” than the monophonic audio of the voices themselves.
O-4: Raffensperger’s voice:
On the 11 Alive audio, at 00:14:03, Raffensperger speaks again.
There are four things of interest here.
One, there are not the signal distortion artifacts on his voice that occur on Trump’s and Meadows’ voices per O-3 above. This is definitive evidence that his audio signal level has a lower overall amplitude than theirs. As well, his audio amplitude (volume) is noticeably less than Trump’s and Meadows’.
Two, his audio fidelity is substantially lower than Meadows’ and Trump’s.
Three, it has a stronger “bass” component around the 200 Hz. frequency range: slightly below normal telephone audio bandwidth.
Four, there is an “aliasing” or high sideband sound on his voice, it’s at the top of the frequency range of his voice and sounds vaguely like a siblance: sss-sss-ss- etc.
However, the WaPo excerpt has Raffensperger’s voice substantially louder, at roughly equal volume to Trump: This is evident starting at 00:20 on the WaPo excerpt, where Trump speaks and then Raffensperger speaks shortly thereafter.
O-5: Cleta Mitchell’s voice.
Mitchell is first heard at 00:19:51 on the NBC audio.
The amplitude of her voice is about equal to that of Raffensperger. The audio quality of her voice is interesting: there’s a resonant peak in the range of about 600 Hz to 1,000 Hz, and the amplitude of her voice rolls off at the high end above about 2,000 Hz, in other words her “treble” is muted somewhat. There is not the “burbling underwater sound” of obvious “packetization” from mobile audio compression.
Her voice also has the high-end “aliasing” in the 2,500 Hz and above range, similar to what we hear with Raffensberger.
At 00:20:47 on the NBC audio, Trump cuts in while Mitchell is speaking. The amplitude of Mitchell’s voice decreases slightly but not substantially. Trump’s voice is substantially louder so it is more audible, but Mitchell’s voice does not decrease substantially.
O-6: Trump talking over Raffensperger:
At 00:25:25 on the NBC audio, we hear an exchange between Trump and Raffensperger, where Trump is talking over Raffensperger and causing Raffensperger’s audio to cut in and out. This begins again at 00:26:00.
O-7: Trump talking over Mitchell:
At 00:26:52 on the NBC audio, we hear an exchange between Trump and Mitchell, where Trump is talking over Mitchell. In this case, the audio behaves similarly to O-5: Mitchell’s voice decreases slightly but not significantly.
O-8: Meadows and Trump talking over German:
At 00:57:15, German speaks up, and by 00:57:30, Meadows is talking over German. German’s voice does not decrease significantly in volume. By 00:57:50, Trump talks over German: same result, German’s voice does not decrease significantly in volume.
O-9: End of call:
At 01:01:51 on NBC audio, the participants wrap up and begin to say goodbye and disconnect. At 01:02:00 we hear some hang-up clicks and then a quick “beep beep” to signal that the conference call has ended.
The 11 Alive audio cuts off after the Goodbyes, before the beeps.
I-1: Call setup.
The first voice on the line sounds like a conference operator, or could be an administrative worker in any individual’s office. This shows that they were using a conference bridge and an operator set up the call. Assuming Trump gave the order to set up the call (a reasonable assumption), the conference bridge could be part of the White House PBX (which itself is as large as a public telephone exchange) or an external public exchange.
What is remarkable is that Raffensperger was delayed by 17 seconds in announcing his presence on the call. He was on the call before the recording begins, then off the call, then back on again 17 seconds after Meadows asked for him. At this point I don’t have an interpretation of the delay, other than that it is interesting.
I-2: Audio volume & quality at start of call:
The relative volume levels suggest that the recording was made from a point that was closer to Trump’s end of the call, than Raffenperger’s end of the call. This is interesting and unexpected.
That would support the hypothesis that the recording was released by someone at Trump’s end. This differs from my earlier comment in another story, based on the WaPo recording, that it was more likely that the recording was released by someone at Raffenperger’s end. However at this point in the call it is too early to reach any conclusion.
Germany’s low audio volume suggests he is in a different room from Raffensperger.
Only Trump has room reverb on his voice. That suggests that Trump is using a speakerphone, that would typically pick up some quantity of room reverb.
Meadows does not have this reverb, which suggests that he is on a different phone in a different room.
I-3: Signal artifact on NBC audio not present on 11 Alive:
I interpret the signal artifact as digital distortion or clipping. It occurs when the amplitude (volume) of an audio signal exceeds the limits of the device that is processing the signal. In the present cases the excess amplitude is not extreme but it does provide interesting information:
The greater amount of distortion on Meadows’ voice than on Trump’s is clear evidence that Mark Meadows’ voice is louder than Donald Trump’s, and that Trump’s voice itself is loud.
I interpret that the distortion occurred during NBC’s processing of the audio for broadcast, since it is not present on 11 Alive’s version. There is no political or legal significance to this difference between the two broadcasters.
However, since the quantity of distortion is greater when Meadows speaks, it clearly shows Meadows’ signal is louder. That may have some additional bearing on the question of where the recording took place and who released it.
But caution is advised here since there are other possible explanations for the difference in volume between Meadows and Trump. For example a speakerphone typically transmits audio at a slightly lower volume than a telephone handset held normally to the ear and mouth: thus if Trump is on a speakerphone (as I’ve concluded from the room reverb) and Meadows is on a landline handset (as I’ve concluded from lack of reverb), then Meadows’ signal will be louder than Trump’s.
This by itself is only a comparison between Trump and Meadows, not a comparison with the signal levels of the other persons present, about which more below.
I-4: Raffensperger’s voice:
Re. the four items of interest in O-4:
One, the absence of the audio distortion artifacts is conclusive evidence that the amplitude (volume) of Raffensperger’s voice is lower than that of Trump’s and Meadows’ voices. The lower amplitude is audible: his voice is just not as loud as Trump’s or Meadows’.
Two, the lower audio fidelity strongly suggests that he is speaking on a cellphone. It’s actually pretty good for cellphone audio, the packet sampling rate is good (it isn’t “burbling” as if he’s underwater, a common cellphone audio artifact) but the frequency spectrum is muffled at the high end.
Three, the stronger bass element around 200 Hz suggests that he could be using a speakerphone over a VOIP circuit, such as if he was in his office. Recall that telephony is 300 — 3,000 Hz., so his bass frequencies are slightly below what I would expect to pass through the PSTN (public switched telephone network).
Four, the siblant high sideband sound (sss-sss-etc.) at the top of the frequency range of his voice: some of this sounds to me like it’s in the frequency range 2,500 Hz and up, so it could be passing over the PSTN.
All of this raises the question: was he at home on a mobile, or was he at the office using a speakerphone?
Further, the fact that the WaPo excerpt has Raffensperger’s audio at the same volume as Trump’s, indicates that WaPo may have processed the recording to get all of the volume levels closer to equal, for the purpose of auditory legibility. However it has the resulting effect of making Raffensperger “sound louder than Trump” due to the frequency distribution of Raffensperger’s voice.
Note, when Raffensperger is heard at this volume level, it appears more likely that Raffensperger is on a speakerphone at his office, which raises the interpreted probability that it was he who made the recording. That could have been done by him using a speakerphone with a “record to removable media” capability, a common feature now.
Raffensperger’s audio as heard on the WaPo excerpt was the basis for my conclusion in comments in another story, that it was most likely Raffensperger who released the recording. But now we see two news sources whose recordings have Raffensperger at lower volume than Trump (NBC and 11 Alive) and only one that has Raffensperger at equal or slightly louder than Trump (WaPo). So it’s more likely that NBC and 11 Alive did no further post-processing on the audio, whereas WaPo equalized the volume for auditory legibility.
All of the above raises the probability that the point in the circuit from which the call was recorded, was actually closer to Trump and Meadows than to Raffensperger. Especially if Raffensperger was on a mobile device at home.
I-5: Cleta Mitchell’s voice:
Due to the audio quality of Mitchell’s voice, it’s most likely that Mitchell was on the call from a personal cellphone.
Re. the fact that Mitchell’s voice does not decrease substantially in volume while Trump is talking: If the recording was being made with an external recorder and mic on the desk near Trump’s speakerphone, then when Trump started speaking, the gain-switching of the speakerphone would have very substantially reduced the volume of Mitchell’s audio.
Since the volume of Mitchell’s audio did not decrease substantially, we can conclusively rule out the possibility that the recording was being made by an external recorder on Trump’s desk. Therefore we can also rule out the possibility that Trump made the recording by himself, and by inference, rule out that Trump released it by himself. That does not rule out that someone else at the White House may have made and released the recording.
I-6: Trump talking over Raffensperger:
Trump’s louder voice causes Raffensperger’s voice to become substantially decreased in volume. This appears to indicate that Raffensperger is also using a speakerphone. If Raffensperger is using a personal mobile device, it is likely that he is using it in speakerphone mode.
This also demonstrates conclusively that the recording is not being made by a separate device on Raffensperger’s desk, because any such device would continue to record Raffensperger’s voice at his normal speech level, rather than a decreased level.
So now we have ruled out a separate recording device on Trump’s desk (item I-5 above) and on Raffenperger’s desk (the present item I-6).
I-7: Trump talking over Mitchell:
Further supports the conclusion that Trump was not recording with a separate device. Also is sufficient to demonstrate that Mitchell was not using any kind of speakerphone, or her voice would have decreased substantially as did Raffensperger’s.
I-8: Meadows and Trump talking over German:
German’s voice does not decrease significantly at these points, demonstrates that he is not using any kind of speakerphone either.
I-9: End of call.
The hangups happen too quickly to tell the order in which the participants hang up, and there is no “after conference” background conversation between one or two remaining participants.
However the two beeps at the end of the NBC audio are interesting. They are typical of signals heard after some or all parties in a conference have disconnected.
From that we can infer that the person who recorded the call was the last person to hang up on the conference, or that the recording was made by a trunk logging recorder or equivalent.
The likely array of terminal equipment is:
Conference operator: Landline phone with handset. Alternately, landline phone or console with noise-canceling headset that would also have resulted in his audio volume being lower, as observed.
Trump: Landline with speakerphone.
Meadows: Landline phone with handset.
Raffensperger: Probably a mobile in speakerphone mode. Less likely but possible, a desk speakerphone.
Germany: Probably a mobile, not in speakerphone mode.
Mitchell: Probably a mobile, not in speakerphone mode.
The likely audio processing in the source material:
WaPo excerpt: Dynamic range compression to get approximately uniform volume levels for intelligibility. However this led me to the mistaken impression (in comments in another story) that Raffensperger’s signal was louder than it is, and thus that the recording was most likely occurring at Raffensperger’s end of the call.
NBC: The distortion indicates that their copy was derived from someone else’s copy and was recorded with the input gain turned up just a tad too high. However, only the NBC audio has the very end of the call including the disconnect beeps. That could have been an editing choice. Helpfully, the distortion provides strong objective evidence that Trump’s and Meadows’ voices are actually louder than the others, and thus that the recording was made at their end.
11 Alive: No distortion so this was likely close to the original source of the recording. Also the only source with the very beginning of the call. However is missing the very end of the call (disconnect beeps), which was most likely an editing choice.
From the above it isn’t possible to say whether there was any interaction between NBC and 11 Alive in passing around copies of the audio. I haven’t dug into the various other recordings available online to try to ascertain “who went first” but the reporting on DK and elsewhere says that was WaPo.
How the call was recorded:
The audio on both NBC and 11 Alive has Trump’s and Meadows’ voices loudest of all by about 6 to 9 dB compared to the others on the call.
There is a technique for getting the audio volume roughly equal between the near and far parties on a phone call, that requires getting into the innards of the telephone set (details on request for journalists only): that is extremely unlikely in this case.
Without using that technique, the “near” party will always be substantially louder than the “far” party or parties, as we have heard in this call.
There are two likely ways that the call could have been recorded. One is to record to removable media on a local speakerphone. The other is to record to digital storage media on the line itself: this can be done via features in a PBX voicemail system or via a logging recorder installed on the trunk circuits (from a gov office PBX to the outside world).
(There is a third method that I would rate as highly unlikely, which is the use of a recorder coupler on the handset side of a telephone. Given the wide variety of proprietary designs for telephones at this point in history, successful recording with a passthrough connector from the handset jack would be unlikely to work. Unless of course, whoever was doing this had “hackerish” sensibilities, but again, unlikely.)
Using the voicemail system is less likely due to recording time limits in most such cases; using an external logging recorder is much more likely.
It’s reasonable to assume that all gov office buildings (the WH, and also in Georgia) are equipped with logging recorders for security reasons (e.g. capturing bomb threats).
The probability that the call was recorded from a location in Georgia is vanishingly small: whoever would have done that would also have had to “disguise” the fact by altering the volume levels of all participants consistently throughout the call, to make it appear to have been recorded from the White House. That would have been an unmanageable task for anyone who didn’t have extensive relevant experience such as in a recording studio.
All of these points are convergent. The call was most likely recorded from within the White House telephone infrastructure.
Who recorded the call and released the recording?
Within the White House there are three persons of interest. One is the conference operator, one is Trump, one is Meadows.
We can reasonably assume that Trump is not technologically sophisticated enough to have recorded it himself and released it himself.
The two beeps at the end of the call provide a clear clue. They are an “end of call” signal, that either all other parties have hang up, or that the conference operator has hung up. The latter point is strong evidence that the conference call operator was not the person making the recording. In any case his voice was at a lower volume level than those of Trump and Meadows.
One of two things occurred: either the person who made the recording remained on the line long enough to hear the two beeps, or, a logging recorder on the PBX remained on the line until after the very last person hung up, and captured those beeps.
To my mind everything points to Mark Meadows. Loud volume on the recording, handset rather than speakerphone, early on the call, and his official role. White House Chief of Staff is an “executive secretary” role: someone who is expected to “get it done” for the executive, tackling a wide range of tasks, including finding specialists where they are needed. Meadows would have been in the position to find the specialist needed to make the recording and deliver it to himself and Trump.
And, Meadows would have been in the position to deliver the recording to WaPo on Trump’s orders but “not for attribution.”
Alternately, he did the whole thing without Trump’s knowledge, to out-coup the coupmeister. I rate that as highly improbable because Trump would have flown into a rage over the recording and then sought out “the guilty party” and fired them immediately.
C-1: The call was recorded from inside the White House telecom infrastructure. Most likely by a logging recorder on the White House PBX. Alternately by removable storage media on Trump’s speakerphone. Less likely by a recorder coupler with a handset jack interface.
C-2: The order to record most likely came from Trump, and was carried out by Meadows, with the engagement of one or more technical personnel.
C-3: The release of the recording was most likely authorized by Trump and carried out by Meadows, “not for attribution.”