Sync & Drift for Digital Audio Recorders

[Ben Moore]

What is the best way to deal with drift using digital audio recorders? I have a 20min clip from a Olympus DS-40 and it will not stay in sync with the audio from the camera (VX-2100). I converted the file to 48k 16bit. and played with the speed of the clip but so far no luck. Best thing I can think of is to chop it up every couple minutes and resync. What is everybody else doing on this issue. (Using Premiere Pro CS3)


Thanks
Ben

[John McClain]

You should read this if you haven't already:
http://www.dvinfo.net/conf/all-thing...ers-drift.html

That recorder isn't designed for what you are using it for and the internal clocks just won't line up. Short of chopping it up I don't know of another way to align the audio; I worked on a feature where the director used a consumer recorder like this but he made sure not to record more than 2 minutes at a time so he could more easily line up the audio. John.

[Rick Reineke]

You could try time-stretching the new external audio track on the timeline to match the camera audio trk.
PTs, Vegas, Sonar, ect, and other DAWs support this function.

[Michael Liebergot]

Rule number one of audio recording, is to make sure that you are recoding your audio at the same speed as what you are recording to tape at.

Converting the file after the fact won't solve the problem as the damage has already been done. You are able to try to figure our what the timing difference is and speed up your audio to match appropriately.

This is the problem with trying to use some thing like the DS-40 for video purposes. As the recorder can only record at a max of 44k WMA file format.

If I was to use the DS-40 it would only be for vocal capture like a ceremony vows and such. Then I would manually cut up the track and sync it in post in small portions.

In the future I would look to something else for your critical audio, either wireless or a more advanced recorder capable of at least 16/48k recordability.

We sue many different audio capture tchniques in the field, ranging from Sennheiser wireless setups, to small recorders such as Zoom H2 for instrumental, and Marantz PMD620 and Edirol R09 for vocal and occasional board capture. Of course we also utilize multi track recorders such as the Edirol R-44 and wireless mic and hard wired mic inputs.

Just for the record there is a real good deal on the Marantz PMD620 at B&H for $299. You simply have to select email better price and they will send it to you. But this would be one of my top recommended small recorders that you could use on a person with a lav mic or use the internal mics.

[Ben Moore]

Thanks Guys

Yeah that is a great deal on the Marantz PMD620!, I have the Edirol R09HR. I just have not tried snyc it on anthing longer than couple minutes yet, but my guess is that it snyc's much better.
I guess it may be best to dump the DS-40 and pick up the Marantz. That way I will always be recording 48K WAV.

I currently use wireless for ceremony's but with the whole white space thing would like to get away from it.

I tried the time stretch thing using Soundbooth, but no luck. I think Michael is right, it needs to start out 48k and not be converted to have a chance over time.

Anyone need a cheap DS-40?? :-P

Ben

[Roger Shore]

There's no reason why you can't convert your remotely recorded audio track to 48KHz PCM, to match the camera audio, in post. This should do the job -and it's free!:

Audio Sound File Converter Software- Convert to wav files and mp3.

And remember that however good your audio recorder is, it has to match the 'master' audio track recorded from the camera. That's the one that is in sync with the video. Even if it's the camera 'clock' that is slightly adrift, that has to remain the master, and you will have to change the length of the remote audio recording -even if it was spot on to start with!!

Here's a procedure I wrote a while back to help with the problem.

My Video Problems :: View topic - Synchronise external and camera audio tracks.

Uses free programs, but you can of course just adapt the concept to suit whatever you use.

Saves having to 'sync' in small sections at a time.

[Ben Moore]

Thanks Roger, I will give that a try.

Ben

[Steve House]

Originally Posted by Roger Shore

There's no reason why you can't convert your remotely recorded audio track to 48KHz PCM, to match the camera audio, in post. This should do the job -and it's free!:

....

Just FYI - the various file compression schemes - wma, mp3, etc - achieve their size reduction in part by manipulating the signal's clock. Once compressed, convertng back may fail to completely recover the original clock sufficiently to maintain proper sync over any signifigant shot length. It's worth a try, of course, but the odds are not good. Once information is lost, it's gone for good and variations in playback accuracy that would have no audible effect at all in casual listening can render the recording completely useless for sync sound.

[Roger Shore]

Sorry, I didn't make clear that I was specifically responding to the OP's question.

The Olympus DS40, as far as I can tell, uses the same WMA compression as the WS200S I have used myself for experimentation, and those files import as 44.1KHz 16 bit files, which can be converted to 48KHz LPCM and adjusted really quite successfully, in my experience.

I'm sure you're right concerning files that have perhaps been 'squashed' by various 'on the fly' compression during recording, and the more extreme the variation from the required final format, then the more troublesome the conversion is likely to be.

The point I was trying to get across is that, even if your remote audio recorder has an absolutely accurate sample clock, it would still be the track that required modification, if it varied from the 'master' camera audio track.

[Steve House]

Originally Posted by Roger Shore

Sorry, I didn't make clear that I was specifically responding to the OP's question.

The Olympus DS40, as far as I can tell, uses the same WMA compression as the WS200S I have used myself for experimentation, and those files import as 44.1KHz 16 bit files, which can be converted to 48KHz LPCM and adjusted really quite successfully, in my experience.

I'm sure you're right concerning files that have perhaps been 'squashed' by various 'on the fly' compression during recording, and the more extreme the variation from the required final format, then the more troublesome the conversion is likely to be.

The point I was trying to get across is that, even if your remote audio recorder has an absolutely accurate sample clock, it would still be the track that required modification, if it varied from the 'master' camera audio track.

Yep, and the more extreme the compression, ie, lower bit rate, in the recorder the more problematic subsequent conversion to proper audio-for-video file formats becomes.

Ben, you might also consider hooking up the recorder's analog output to your sound card's line in, play the file and re-record it in the computer in the proper format - 16 bit, 48 kHz, WAV file. Sometimes this is more successful than trying to convert the file in the computer in software because you're playing back on the same device that made the original recording and that helps cancel out timing errors. What was recorded too slow, for example, plays back too fast by the same percentage, effectively cancelling out the speed error. That's the reason some people report fewer problems with cheap mini-disk recorders than they do with cheap file based recorders.

And you're definitely correct, Roger, that the recorder and camera must match. In order to avoid the problem in the first place both the camera and the recorder must share the same timebase, that is, be driven by a common clock. Higher level equipment has provisions such as genlock input to the camera and wordclock on the audio recorder, with devices such as master clocks and Lockit boxes to provide clock to both, or audio recorders that accept video blackburst or tri-level sync signals from the camera as the master for their internal clocks.

[Ken Marsley]

This is a VERY misunderstood subject. And it is also a very miscommunicated subject. Some suggestions in this thread might perpetuate this.

The word "drift" IS entirely appropriate, here. The original poster is using a clock in an video device and a clock in an audio device to capture things happening in the real world. These two clocks (quartz crystals) are not connected or talking to each other in any way, so they not operating at the same rate.

The proper solution to this is:
A master clock feeds GENLOCK to all cameras shooting.
Same master clock send WORD CLOCK to all audio devices shooting.
Result: The camera and audio recorder both follow a stable external source.

Another solution is that you take the video output of your camera, resolve it to WORD CLOCK using a Clocking device with video input sync.
Result: The audio recorder follows the camera source, whether stable or not.

Another solution is to buy an audio device with direct video sync (essentially, an onboard clocking converter).
Result: The audio recorder follows the camera source, whether stable or not.



Redigitizing the audio will not do anything at all regarding the drift between the audio and video clocks. I don't know why this was suggested. You can't go back in time and capture the original source material using video clock.

You have to stretch or crunch the video or audio in post. Add or remove frames/samples.


The reason two different I-Rivers or any other 2 device from the same mfgr will have little drift is because they contain, ostensibly, the same clock equipment internals. So each device may drift from a professional clocking device, but they are drifting at near the same rate, so they stay relatively close to each other.


Audio should be recorded at at 48kHz WAV. There's no negotiation about that if you want to talk about doing things in sync.


And just to clarify: none of this has anything to do with timecode. This is digital clock sync.


This post has the right principles:
My Video Problems :: View topic - Synchronise external and camera audio tracks.

[John McClain]

Originally Posted by Ken Marsley

This is a VERY misunderstood subject. And it is also a very miscommunicated subject. Some suggestions in this thread might perpetuate this.

The word "drift" IS entirely appropriate, here. The original poster is using a clock in an video device and a clock in an audio device to capture things happening in the real world. These two clocks (quartz crystals) are not connected or talking to each other in any way, so they not operating at the same rate.

I have to respectfully disagree on the use of the term 'drift'. Since the audio recorder the op was using does not have the capability to lock to another device I feel we should not use the term 'drift' because we normally use that term to describe two devices that are properly connected but still 'drift' for some other reason (I feel semantics are very important in this regard). Rather than drift, what the op is experiencing is what happens when one uses a device for other than its intended purpose. Other than that, I completely agree with your post Ken. I think what you have written in regards to proper 'clockable' devices and attempting to re-sync the unclocked audio should be made a sticky. john.

[Steve House]

QUOTE=Ken Marsley;980312]This is a VERY misunderstood subject. And it is also a very miscommunicated subject. Some suggestions in this thread might perpetuate this.[/quote]

Very good post Ken. The reason redigitizing was offered is the OP is stuck with a WMA file recorded at who knows what real sample rate on a low-level consumer device never intended for serious sound recording. Sometimes (and sometimes not) redigitizing by playing the file back on the original recorder and re-recording the resulting analog signal at the proper sample rate can get you closer than trying to import the WMA file into the computer, converting it to WAV and changing the sample rate to 48kHz in software. Neither approach is truly satisfactory compared to doing it right the first time as you point out.

[Ken Marsley]

Well, my apologies, I have to eat my words: I couldn't find anything that was blatantly misleading in this thread, aside from the normal differences in workflow, best practices, and terminology. I should have said that someone new to sync issues might misread some of the suggestions, here and on many other threads. And since we're talking about super low budget stuff, workflow and best practices go out the window.

John McClain - Fair enough. Discussing the semantics of the word "drift" certainly elucidates the subject further, which is of value to everyone. Very good point that we can't refer to a device as having a "drift-malfunction" if it was never designed to sync in the first place.

Steve H - I suppose analog playback, and re-digitizing is a possibility. It's certainly can't hurt to try. Anyone trying this would want to try to replicate the exact conditions under which the recording was done: ambient temperature, proximity to heat-generating lights and equipment, as heat is a major factor in digital clock variations. Advil might be necessary, too. But I'm imagining that this will be done with mini-stereo jacks, running the chance of inducing more noise into the recording.


Anyway, for people just starting out, using the Tascam's HD-P2 audio recorder ($800 and up) is probably the simplest way to solve A/V syncing issues during long-form shoots. Take the video output and plug it into the Tascam. Simple.

Doing long-form recording without directly syncing your audio and video devices will always make for a tricky post situation. Square Peg, Round Hole. Get out your hammer.

(Btw, partly I was responding to many other posts outside of this thread..)
Oh, and let me say one more thing for those who are lurking and trying to wrap their heads around this subject:

A)
Audio Recorder set to record at 44.1kHz.
Audio Recorder is syncing to the blackburst/tri-level sync from the video recorder


B)
Audio Recorder is set to record at 48kHz
Audio Recorder is not synced to the video recorder


Option (A) is better than (B) regarding sync. Both options will suck if the person hasn't taken the time to study a little about mic placement. Remember, 48kHz doesn't guarantee anything IF IT'S NOT RECEIVING SYNC FROM THE VIDEO SOURCE! (like putting 97 octane in a 20 year old car) Converting 44.1kHz to 48kHz after the fact will do nothing for your sync issues.

I'm in full agreement that recording directly to a compressed format that has no professional precedent will leave the recordist at the whim of the consumer technology. No help or suggestions there. MP3 is a delivery format, not a recording format. Memory is just too cheap these days to go down that road. Period.

And again, for the newbies:
This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode.

To those that are new at this, if you are confused, that means you're on the right path!

[John McClain]

Originally Posted by ken marsley

this has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode. This has nothing to do with timecode.

To those that are new at this, if you are confused, that means you're on the right path!

lol!lol!lol!...