Digital Audio Drift
 

I noticed something disturbing about both my audio cards that I want to ask for peoples' opinion on.

I've been tranferring Star Wars from laserdisc to DVD recently using a two step process. First I capture the video and analog audio together through the Canon GL2. Everything's fine, in sync, etc.

Step 2 is I re-run the LD but record the PCM digital audio output into the sound blaster audigy2.

Well, what happens? They're out of sync significantly (several frames) by the end of the movie.

Either the sound blaster was dropping samples, or the LD player was. Either way it was bad.

But it gets worse. I used a different sound card, a "professional" quality one, Terreac I believe was the brand name. This time, it didn't drop samples. It had too many - the audio was oversampled.

Could it have been the LD player? If so, both cards should have had the exact same drift. But they were different. The pro card was closer to perfect, but still was off by a frame or two at the end.

What does this say about when we capture audio to HD recorders or minidiscs? How do we reliably transfer the audio to the computer without losing samples? Could I be doing something wrong?

Since it was an analog transfer, I'd have to look elsewhere. You stripped the header information out of the LD when you transferred it. Could be myriad things, but I'd suspect your clock on the soundcard first. Everything stayed at 48k/16bit throughout the process?

LD output is 44.1 KHz, I believe. At any rate that's what I captured it at. Could that make a difference?

Actually I found out LD output is exactly 44,056 Hz. I will have to see if that makes a difference (I doubt it will but who knows).

Yes, output at the LD is 44.1, but the Canon will convert it to 48k. Did you by chance, have the Canon set to 32K/12 bit?

The canon GL2 has no digital audio input. The analog output of the LD player into the camera has no drift. It is only the digital output of the LD into the sound blaster or the Terreac that has the drift. The problem must be with the sound cards since neither captures properly, but both are off by different amounts. The only thing I can try is setting the recording sample rate to 44,056 to see if that makes a difference. It is possible that the cards don't compensate properly for the slight difference between 44,056 and 44,100.

I guess I misunderstood. I thought you were taking an analog out from the LD into the Canon, using the canon to convert analog to dig, capturing in the computer.

Yes, the difference between 44.anything and 44.1 locked is going to make a huge difference. Just the .009 drift in an older XL1 causes audio to drift within 3 minutes.

The drift is not big enough to be attributable to the 44 hz difference. The drift amounts to less than 1 sample per second. So the sound card is clearly compensating for this difference by oversampling. It is just not doing so accurately. But if I set the sample record rate to exactly 44,056, which I believe I can do, that may solve the problem since the card won't have to oversample at all.

Alright, folks, serious issue here. The drift happens regardless of the audio source and regardless of the sample rate. It also happens regardless of the sound card or recording software.

Has anyone ever experienced this? You capture audio via optical or coaxial, and the resulting audio is either over or under sampled and out of sync (by a fraction of a sample per second) with the separately captured video. What could I possibly be doing wrong?

Well...you're certain it's not the camera, certain it's not the conversion device. So, I'm at a loss, since you feel it's not those devices, and haven't tried a different audio card.
I do this on a reasonably regular basis using my LD player and DVD into Vegas and Forge, using M-Audio and Echo audio products. Zero drift, ever. So, I'd be looking at the card inserting information or having a drifting clock, which happens to be quite common.
Are you certain the card isn't resampling incorrectly, and then your NLE or DAW attempting to correct the drifted audio via either resample or shift?

Maybe, I'm missing something, and if so, I'll back out and say how sorry i am for wasting peoples time. However, it seems to me that the incorporation of an analog capture chain (device) and the frame buffers associated with such a process are the issue.

No, you are right. Except Peter is saying that it's correct when previewed from the analog device, and correct when brought in from the Firewire device (camera) and only incorrect when he's transferring just audio separately. Why the two captures are happening this way, I don't know. There is little to zero loss in an analog to dig transfer when good cables, good converter card are used. I'd capture using a good converter such as the Canopus ADVC 300 and forget about it because there would be zero problems. (this is how I do this most of the time, unless the audio is 5.1, in which case I transfer audio using an Echo Layla)

I'm reasonably certain it's not the camera since it's happened using two different cameras. First time was when trying to match footage from a GL2. The audio from the tape was fine. But the audio from the external source (the original LD) through my soundcards drifted.

Now the source is from a DVX100A. Same thing. Audio from the tape is fine. Audio from my 6-track digital recorder? Drift.

", I'd be looking at the card inserting information or having a drifting clock, which happens to be quite common."

AH, sounds like the key. No resampling going on this time - source is 48 KHz, project is 48 KHz.

Jeff, sorry but I'm not sure what you mean. I'm comparing the audio from the DV tape with the audio from a separate digital source and imported into my computer. I line them up at the very beginning, both visually and aurally. By the end of the clip, they are quite noticeably out of sync. And if the clip is long enough (about an hour) the lip syncing is noticeably off.

Douglas,
I had the same thought as you but there's noticeable background noise when I go the analog route. I probably need better cables, or a better sound card / ADC.

About the clock drift problem, that seems the most likely culprit. Any good cards you would recommend that would not have such a problem?

I have two: the SB Audigy2, and the Terreac EWX 96/24. The Terreac is much much better than the SB, but still not perfect. The drift is substantially and consistently less though.

Incidentally, if there was a way to transfer the audio bit to bit using the SCSI interface of the Roland, I'd do that. But I don't think there's any PC software support for it (VS-880EX).

Peter, are you referring to the "Terratec" rather than Terreac?
These are consumer cards, labeled as "professional" because they use (used) Waldorf sampling hardware. But they are FAR from professional. Their Aereon (sp) is a joke. Can't comment on their other cards, but they are a video card company that makes other hardware as well.
Look to Echo, MOTU, or M-Audio for great solutions.