Audio drift when recorded on different equipment

Question

Level 1

45 points

Audio drift when recorded on different equipment

For me, this is a very inexplicable problem.
Composing music on Logic Pro. The bas player (who lives in another city) receives the Logic project file (with midi lines of the music), records his bas lines on Logic Pro and sends the recorded bas files (uncompressed). When the bas recording is imported into the master Logic project, the imported file drifts. Gradually it starts getting audibly out of synch, more and more as time goes on. Nope... on his Mac it sounds great, tight, correct, but when imported into the master-project where the other audio files are recorded, it drifts. Why? How?

Not only in Logic.
Concert video. Video taken with 5 HD cameras, camera's audio at 48KHz. Audio was also separately recorded in a Pro Tools system, also 48 KHz.
The video was later captured into FCP using the standard HD 1080 50i setting. Syncing up the videos was no big hassle. All the cameras audio-files read as they should. But since these are the camera-mics, the sound quality is far from usable. So, we add the "real" audio, the audio we recorded on ProTools during the concert. When adding the ProTools audio, it drifts out of sync... slowly but surely.
I found in the FCP manual about the fact that if audio and video is not synced using a master sync / time generator during the recording, the audio can drift in editing.
In my mind, the master program (in this case FCP) should read exactly 48000 / second regardless if the audio is captured on the cam or on any other 48KHz device. But it does not do this... it reads maybe 48090 - or whatnot - samples per second. The result is that the audio drifts ahead of the video and other audio tracks.

But how is this possible? I want to understand. I can understand the need for an external sync during recording, to more easily sync up the start point for each track when editing, but one second is one second long, exactly, no more, no less... and 48000 samples is 48000 samples, wether recorded on one device or another. 48KHz is 48KHz and 1 second is 1 second, 1 hour is 1 hour, summing up the exact amount of samples whatever the device. 1 hour is 172 800 000 samples read at 48KHz. Nothing more. Nothing less. And a digital device should be able to read this at that accuracy. Or?
In my mind, a bit is a bit, a byte is a byte, a sample is a sample... the master-program should read each sample in its place, 48000 / second, in a quantified "time grid" where each "grid snap" is 1/48000th of a second. So, in my mind, this drifting just cant happen, but it does. Its like it has a separate clock or a smaller "grid" for each track, instead of one clock, one grid for the whole project. I could understand if the whole project drifts (but that would not be noticeable to the ear since we are talking about a drift of 2 seconds per hour). But I cant understand how one track drifts independent of the other tracks. And it is not intermittent. It drifts with total consistency, in a linear mode. It does not depend on if I have on "Play" from start to end, or if I randomly move the "play-head" to a part of the project... the drift is linear and exact, so at the end of the project the amount of drift is the same, regardless if played for start to end or jumped to end.

Can anyone explain why, and if there is anything one can to in the program (FCP or Logic) when reading the files to ensure actual sample-time accuracy when reading imported files?

iMac C2D 2GHz and 2 Gb RAM, Mac OS X (10.4.10), a couple of G5 2x2GHz/2.5Gb, iBook G4 966, 10.3.9, a couple of old PMs and iMacs

Posted on Mar 11, 2008 5:19 PM

Reply

Answer 1

Mar 11, 2008 6:39 PM in response to Per Nordin

hi Per,
you might, and only "might", have more luck if both machines were identical in spec.
for short pop songs this inaccuracy might be tolerated.
but for AV? i think not.
you could cut the file up and sync it section by section.
lotza work, and you really shouldn't have to.
a click trk is always a good thing.
the same and worse used to happen with analog tape, in fact it was very much worse.
other than the two machines running at slightly differing "versions" of the sample rate, i can't help you.
DR9.

Reply

Answer 2

iSchwartz

Level 5

4,321 points

Mar 11, 2008 9:14 PM in response to Per Nordin

Regaring your problem with the audio from your video shoot being out of sync, I'm going to venture a guess that it has to do with some kind of pull-down that happens upon importing your footage into FCP.

Regarding your problem with Logic, you said that his track brought into your session drifts out of time (which I'll assume for now means that it progressively plays behind the beat). It would be inevitable, then, that the track should also be out of tune with your track (assuming the bass part was in tune to begin with). Is it out of tune?

If you really want to get to the bottom of this, have your friend record a test tone at 880 Hz from Logic's test oscillator. The recording he makes should be exactly 60 seconds long. Exactly. That's 2,880,000 samples, not a sample more or less. This can be achieved by setting the tempo to 60 BPM, setting the sample rate to 48K and recording 60 bars-worth of the test tone. Then he should trim the file in the sample editor and confirm the length of the audio file from the data shown in the audio bin, which should be 2,880,000 samples long.

Once you get the file, load it into Logic, set to 48K. Open the bin and look at the size of the file. It should be 2,880,000 samples long.

Install a test tone oscillator into a channel and set it to 880 Hz. Play back Logic's test tone along with his audio track of his test tone. Over the course of those 60 seconds you should hear a static sound of two test tones playing at once. If you hear beating or wavering in the sound, there's a problem with one of your system's clock. If this happens, put a tuner on the track you imported and check the pitch. What does it read?

Reply

Answer 3

Per Nordin Author

Level 1

45 points

Mar 12, 2008 4:43 AM in response to David Robinson9

David,

"for short pop songs this inaccuracy might be tolerated.
but for AV? i think not."

Right. We are talking about symphonic rock, not short pop songs. On this concert the longest song is 26 minutes and the whole concert is close to 3 hours.

"you could cut the file up and sync it section by section.
lotza work, and you really shouldn't have to."

Thats what we are doing... cutting and adjusting the video a few frames every two minutes or so to bring it back to sync with the audio.

"a click trk is always a good thing."

Does not have any help on this type of situation. It is simply that the machines have an intolerably bad precision in their internal clock.

"the same and worse used to happen with analog tape, in fact it was very much worse."

Yes, I know. And I can understand that. But I cant understand how a modern computer has a worse clock accuracy than a normal wrist-watch does. We are talking about a drift of some 2 seconds / hour!

Anyhow, thanks for the response.

Reply

Answer 4

Per Nordin Author

Level 1

45 points

Mar 12, 2008 4:56 AM in response to iSchwartz

iSchwartz,

Thank you for your input.

"Regaring your problem with the audio from your video shoot being out of sync, I'm going to venture a guess that it has to do with some kind of pull-down that happens upon importing your footage into FCP."

I guess you are thinking NTSC. Being in europe, we use PAL.

I have come to the conclusions that it simply is that a modern computer does not have very good accuracy in its internal clock, actually less accuracy than a cheap 20 year old digital watch.
So instead of actually sampling exactly 48000 samples each exact second, some machines sample 48002 samples each second, or 48010 or 47999.... So when the files from different sources are placed in the same project, and thus all played using one clock, these differences are noticed.

Reply

Answer 5

iSchwartz

Level 5

4,321 points

Mar 12, 2008 5:11 AM in response to Per Nordin

Hi PN,

You're welcome!

I would like to suggest that you not yet feel married to the conclusion you've come to. Maybe an "informal engagement" for now... 😉

I don't think that what you're experiencing is due to a difference in clocking tolerances between computers. If this were the case, it would literally be impossible for any two (random) people to exchange files created on any two (random) computers. In all of my years working with digital audio and exchanging files with clients -- or even playing back files I've made on different computers in my studio -- I've never run into this problem. If there were such a problem? I'd hear it, whether it was a miniscule difference in pitch or timing.

Even back in the days when we were syncing two multitrack decks via Lynx these kinds of problems were truly rare.

I think you should try out that test I described with the 880Hz tone. That will tell you absolute volumes about what's going on.

Message was edited by: iSchwartz

Reply

Answer 6

Mar 12, 2008 5:24 AM in response to Per Nordin

hi again,
when i was a cough PC user, back in the stoned age, i had a little app that read out the exact sample rate.
the variation was as much as 100samples/second.
so your theory about dodgy clocks is really fact.
i also think that the clock is not steady and fluctuates over time.
some projects i do, hang perfectly from the get go, whereas with others, i'm continually micro-shifting events/clips.......
DR9.

Reply

Answer 7

Per Nordin Author

Level 1

45 points

Mar 12, 2008 6:06 AM in response to iSchwartz

iSchartz,

I had heard about this problem before, from a film-editor and from some musician friends. I had not encountered it myself before, because I normally worked all alone, using internal sound sources and such.

The drift is so small, that it would not be noticeable in pure audio in a short pop song. But in a 30 minute symph-rock piece it becomes evident. And in a pure audio situation, it is not as evident as in a AV situation, since the ear-to-ear is more tolerant than the eye-to-ear. The eye-to-ear notices the unsynch sooner than the ear-to-ear does.
As for hearing a pitch-shift, the difference in sample-rate (and therefore playback speed = pitch) is so minute that it is not very evident when dealing with acoustic instruments since there are so many more things happening when a person sets his flabby finger on the grip (as well as the physics of a string) which affects the pitch sensation. But sure, if using two sinus-tones (or even better yet two tones with steady harmonics) one can better pinpoint the exact amount of difference in the sample-rate of the two machines. But at this point, I feel it is of purely academic interest. I understand now the cause of the problem and since I am not a programmer nor computer designer, I have no use of such analysis. What I do need is a genlock.

I posted the same question in the Final Cut Pro forum and got that explanation, confirming my theory. It simply is a fact that digital clocks in computers and/or dedicated hard-disk recording devices are not as accurate as one may desire. One may work at 47999.97 samples per second, another at maybe 48000.3...
The cameras where locked to a genlock, but not the pro-tools system, so all the cameras are forced to work at one sample rate, but the pro-tools was left to its own clock. So the cameras do not drift amongst each-other, but the un-genlocked pro-tools system beats by its own drum, so to say.
So, when - as in this case - having a video take of a 3 hour concert and having the master audio recording on a separate recording device, the only solution is to every few minutes adjusting the video a frame to bring the audio and video back into synch. Sad, but true.

Reply

Answer 8

iSchwartz

Level 5

4,321 points

Mar 12, 2008 6:20 AM in response to Per Nordin

Per Nordin wrote:
The drift is so small, that it would not be noticeable in pure audio in a short pop song. But in a 30 minute symph-rock piece it becomes evident. And in a pure audio situation, it is not as evident as in a AV situation, since the ear-to-ear is more tolerant than the eye-to-ear. The eye-to-ear notices the unsynch sooner than the ear-to-ear does.

Tolerances of this kind are dependent on what your profession is as well as your skill level in that profession. If something's as little as 1 cent out of tune I'll hear it. Other people can't. And this world is literally filled with picture editors who can't tell when picture and dialog are a frame or two out of sync. I think you probably know what I mean.

As for hearing a pitch-shift, the difference in sample-rate (and therefore playback speed = pitch) is so minute that it is not very evident when dealing with acoustic instruments since there are so many more things happening when a person sets his flabby finger on the grip (as well as the physics of a string) which affects the pitch sensation. But sure, if using two sinus-tones (or even better yet two tones with steady harmonics) one can better pinpoint the exact amount of difference in the sample-rate of the two machines. But at this point, I feel it is of purely academic interest. I understand now the cause of the problem and since I am not a programmer nor computer designer, I have no use of such analysis. What I do need is a genlock.

What you're doing is relying on a theory -- which is more or less hearsay -- to explain the problems you're having. If you really want to get to the bottom of it, do the test. (And here I'm referring to your bass track, not the multi-hour concert recording). Trust me, I didn't take the time to suggest that test to you for the purpose of steering you through an academic exercise. Rather, it's basic troubleshooting that could help you pinpoint the problem.

Anyway, whoever overlooked the need for genlock during a 3-hour (or whatever it was) concert recording should be shot. Or at least taunted mercilessly for at least a year.

😉

Reply

Answer 9

juhani h.

Level 2

264 points

Mar 12, 2008 6:37 AM in response to iSchwartz

Interesting discussion. I once bumped in to a similar problem: a friend of mine added some guitars at home to drums I had recorded in the studio. Later, I was surprised to hear that the guitars drifted out of time over the course of the song.

I couldn´t understand why until I realized he had re-recorded the drums into some type of digital machine in order to add the guitars. No wonder things started to drift.

The solution I tried was to calculate exactly when the last chord was supposed to be struck, which was easy cause the drums also played simultanously, and then time stretch the guitars in the Sample Editor.

Unfortunately, it didn´t work. I needed to shorten the guitars by something like 1,5 sec over three minutes. It never worked, the Sample Editor choked. Maybe because this was ten years ago or so. Maybe it would work today? Haven´t had a reason to do this since...

The solution was to use the original drums, by the way. The rough mix of the drums ended up on the record. So I learned two things in the process: 1) digital sync is good and 2) make good rough mixes!

/juhani

Reply

Answer 10

Per Nordin Author

Level 1

45 points

Mar 12, 2008 11:53 AM in response to iSchwartz

David: Thank you very much for the confirmation. What was that gadget? It would be nifty to have.

iSchwartz: I understand your point. Thank you.

As for pitch, being a trained and examined piano-tuner I too have very acute hearing. Actually too acute. To me, everything sounds out of tune. But I have come to realize that those with such acute hearing are very rare, and there are so many other factors that play in, when playing e.g. a bass. Show me ONE bas player who can play a fret-less bas with a 1 cent tolerance in his intonation. Then we have the fact of voluntary vibrato (which per definition is a pitch variation). Or show me one fret-bas (or guitar) that has such precision in its frets. Or a guitar that actually can be tuned?!?. Or a string that has a stable harmonics series (or even a linear set of harmonics... strings have a logarithmic set of harmonics). Or a string that holds the same tone over time from the attack to the full decay, even if it is a fixed string like an open guitar string or a piano string...
And when a bas, 2 guitars, various percussion, a full heavy drum kit, a load of fat keyboards with sampled strings, mellotron, chorused synth-pads, a fender Rhodes piano (which has metal tone-rods, meaning they have a very erratic and un-harmonious set of harmonics) not to mention the notoriously untuned moog, four part harmony vocals, etc, is all added together, the reality is that the small pitch-shift of the bas is actually less noticeable than the timing at the accumulated drift.
So sure, using an actual straight tone generator, I can hear the pitch-shift, but in an acoustic reality of all the physics of playing on a string with fingers, etc,etc, the very minute pitch-shift is negligible in the grander scheme of things. Hearing the result, I am fully convinced you would agree.

"+Anyway, whoever overlooked the need for genlock during a 3-hour (or whatever it was) concert recording should be shot. Or at least taunted mercilessly for at least a year.+"
Yes. True. And they where not directly gratis. And the camera work is not top-notch. The concert was filmed in Holland, and in hindsight I wonder if the Dutch camera crew had shared a joint before they went to work...

Juhani: Thank you very much for your input and for sharing.

Reply

Answer 11

iSchwartz

Level 5

4,321 points

Mar 12, 2008 12:05 PM in response to Per Nordin

The concert was filmed in Holland, and in hindsight I wonder if the Dutch camera crew had shared a joint before they went to work...

Don't they have laws against that kind of thing in Holland? Especially in Amsterdam? And specifically, in that tea shop that's on... oh, what's the name of that street again? 😉

Reply

Answer 12

Mar 12, 2008 2:05 PM in response to iSchwartz

HIGH STREET, of course!!!!!!!!
DR9.

Reply

Answer 13

Mar 12, 2008 2:13 PM in response to Per Nordin

"David: Thank you very much for the confirmation. What was that gadget? It would be nifty to have."

yes it would, but i cannot remember it's name.
the guy who wrote it had other software out as well and it was just part of a package. it was in the late 90's.....i was young, foolish, and using a PC (pentiumII).
now i'm older, wiser, and use a mac. (so what's changed?).
DR9.

Reply