Audio Drifting (DSLR with external audio recorder)

G

Giganova

Well-known member
Hi guys --

My audio and video are having a serious drift which makes postproduction a major nightmare. Here's my setup:

Audio: Sanken Cos11B -> Sennheiser G3 wireless transmitter/receiver -> Tascam DR100, recording @ 24bit/48kHz
The output of the Tascam DR100 goes into the mic input on my Nikon D600 which records audio at 16bit/48kHz.

I'm using the (inferior) audio from the Nikon to sync the sound files of the external Tascam recorder. The sync point is a clapper board at the start of the recording.

The problem is that the audio is drifting 15 frames within 10 min (at 24fps), the audio of the Tascam being a bit too fast.

What causes the problem and how could I fix it?

Thanks!
 
I guess I found a quick & dirty solution: in post, I have to change the speed of the Tascam audio track by 99.9% to get the two audio tracks perfectly in sync.
 
This is a common challenge with non-sync, dual-system sound. In other words, your two devices are not sharing a master clock to keep drift from happening.

Each device, your recorder and your camera, has an internal clock that maintains the audio sample rate, in this case 48kHz. However, there are differences between separate devices with separate clocks and those differences can be quite considerable in low-cost, consumer devices. One may actually be running at a true sample rate of 48.002kHz and the other at 47.998kHz. Put them into your NLE, that plays them both back at 48.000kHz, and they will drift out of sync.

On big-budget, professional productions with dual-system sound, there will be a master clock to which both camera and sound recorder (both being capable of time code and external sync) will be jammed.

TASCAM handheld recorders are some of the worst offenders in my experience. The DR-680 isn't bad at all, and the FOSTEX FR-2LE is rock-solid.
 
Thanks for your replies! Makes sense they would drift in the absence of a master clock. Luckily, the drift seems to be pretty stable, so hopefully I'll get away changing the speed of the Tascam audio track by 99.9%. I think the culprit is the Nikon D600: I recently filmed a music video with multiple Nikon D600s and found the same 99.9% drift when trying to sync the sound with the master recording of the song from the recording studio.
 
I haven't tested my GH4, but my GH3 will stay within 3 frames per hour when sync'd with my Tascam DR-07 Mkii or DR-40 recorders set to 48 kHz 24-bit. ( re-syncing is easy with Sony Vegas Pro )
 
Alas, a drift of 15 frames in 10 minutes is not unusual for non-genlocked consumer gear.
TheDingo's experience of 3 frames in 60 minutes is MUCH better than average for consumer gear.
But, as TheDingo says, resyncing is easy enough in most any modern NLE.

Note that it is more common to "pull up" the video to match the audio. The audience will tolerate a duplicated frame (or dropped frame) every few minutes.
But changing the length of an audio recording by a slight amount can result in significant audible trauma unless you are lucky. NOT recommended.

Presumably, you are not editing 10-minute-long shots into the production. Any place they make a video edit is a golden opportunity to "pull up" the video to match the audio track.
I would NEVER EVER attempt to make fine adjustments to the playing speed (sample rate) of a digital audio recording.
 
Shoot things shot by shot and learn the discipline of putting a clapper board on each take and the drift should not be a problem unless you are doing hour long takes!

Sync check and match up each take before you start the edit as that is how film with sep sound has been done for over 60 years!
 
Richard Crowley said:
Note that it is more common to "pull up" the video to match the audio. The audience will tolerate a duplicated frame (or dropped frame) every few minutes.
But changing the length of an audio recording by a slight amount can result in significant audible trauma unless you are lucky. NOT recommended.

Presumably, you are not editing 10-minute-long shots into the production. Any place they make a video edit is a golden opportunity to "pull up" the video to match the audio track.
I would NEVER EVER attempt to make fine adjustments to the playing speed (sample rate) of a digital audio recording.
Click to expand...

Could you elaborate on the "audible trauma"? I'm guessing it isn't that people notice the change in pitch, but rather that the resampling somehow messes with the high frequency part of the spectrum. What does that sound like?
 
Richard Crowley said:
Note that it is more common to "pull up" the video to match the audio. The audience will tolerate a duplicated frame (or dropped frame) every few minutes.
But changing the length of an audio recording by a slight amount can result in significant audible trauma unless you are lucky. NOT recommended.
Click to expand...
With "pull down" the video, you mean change the speed of the video by 0.1% to get it in sync with sync the audio, correct? Would it be better to neither change the speed of the video/audio, just cut the video/audio in pieces and match the smaller segments, ignoring the drift? I'm filming interviews and the maximum duration my Nikon D600 allows me to record is 20 min, but I'll only use some sound bites from the interview, not the entire length.

Thanks!
 
Giganova said:
With "pull down" the video, you mean change the speed of the video by 0.1% to get it in sync with sync the audio, correct? Would it be better to neither change the speed of the video/audio, just cut the video/audio in pieces and match the smaller segments, ignoring the drift? I'm filming interviews and the maximum duration my Nikon D600 allows me to record is 20 min, but I'll only use some sound bites from the interview, not the entire length.

Thanks!
Click to expand...

Then it sounds like cutting the audio and shifting to match lip sync as required would be 'better', since then there is no stretching/resampling of the audio.
 
j1clark@ucsd.edu said:
Then it sounds like cutting the audio and shifting to match lip sync as required would be 'better', since then there is no stretching/resampling of the audio.
Click to expand...

I don't think all stretching/resampling is created equally. I've used the time-stretch feature of the cheapo GoldWave audio editor with excellent results, though I was adjusting the length of an hour or more of audio by only a few seconds. A 15 frame adjustment in a short clip would probably be quite noticeable.
 
In most NLEs or DAWs, stretching or squeezing a clip does not 'normally' change the pitch..(unless desired) The process does however produce artifacts. How severe, depends on the NLE's or plug-in's algorithm and the amount needed. In my experience usuing Vegas Pro with the Zplane élastique timestretch, a second of correction in a one minute, would not be audible to most folks.. at least on dialog. Your VMMV depending on other variables.
 
If you can get away with actually shrinking/stretching the audio track without complaint from the customer, then go for it.
No I am not talking about simple shrink/stretch. I am talking about the "smart" variety where the pitch is maintained.
However the result of that kind of manipulation can be expected to cause audible artifacts where you can actually here regular glitches. At lest with the tools that I was using.
If you have better tools that are artifact-free, then definitely go for it. Use whatever works for you.
I do mostly music productions and that is probably more sensitive to that kind of manipulation than dialog might be.

Please note that the term "pull-up" has nothing to do with the 3:2 "pull-down" kludge they devised to squeeze 24FPS cine into 30FPS video.
"Pull-up" refers to the act of moving the picture or sound relative to each other to re-synchronize them.
It was in use before video and NLE, and probably referred to actually tugging on the picture or mag audio strips of film.
I know it is a cliche, but newbies don't know how good they have it with computer-based electronic editing vs. the "bad old days"

Remember also that "drop-frame" has absolutely no affect on how the video or audio is recorded.
It is only a way of diddling the numbering scheme so that the frame numbers match the wall clock after color television changed the frame rate to 29.94 FPS.

I will repeat my previous statement. It seems very unlikely that you are using 10-minute long shots without interruption.
If you are editing the picture at all, then every edit point is a place where you can "pull-up" the picture to re-synchronize it with the audio track.
 
Wow really interesting that the .1% didn't jump out as the "answer". Yes drift is not unusual but what you have is a pull down issue. I have found consumer/prosumer gear to drift less than a few frames per hour generally. With older analog gear it was MUCH worse.

But the difference between 24fps and 23.976 is 0.1% same as it is between 30 and 29.97. In the ancient days (a few years back) this was a constant issue since film was shot on film and post sound was done to video. So the 24 FPS film was transferred to video using the 2:3 pull down (duplicating every third frame) to make it 30 fps and then played back at the video standard speed of 29.97 fps. So we were always working with slowed down picture and we had to speed up the sound when we went back and mixed to the projected film.

Film and video have frame rates, sound only has speed. So people got used to talking about sound speed as it related to fps which of course leads to some confusion.

FCP has a "bug" where if the capture rate doesn't match the project rate it will time stretch imported audio (and probably video) to match. Since you are possibly working at a true 24fps and you probably haven't looked at your capture settings in a long time since nobody "captures" any more this may be your issue. Assuming your on FCP. maybe other DLE's have a similar issue, though I haven't heard of any.

So in your case you would need to make sure both settings are the same and reimport the audio from the original files (the imported ones have already been changed so your SOL on them).

BTW the other big point of confusion is that camera manufacturers are fond of calling 23.976 "24p". So the other issue may be that you have set the NLE to 24 when your video is actually at 23.976.

Whatever the case I would bet money that the problem is a pull up/down that shouldn't be happening.
 
Richard Crowley said:
Please note that the term "pull-up" has nothing to do with the 3:2 "pull-down" kludge they devised to squeeze 24FPS cine into 30FPS video.
"Pull-up" refers to the act of moving the picture or sound relative to each other to re-synchronize them.
It was in use before video and NLE, and probably referred to actually tugging on the picture or mag audio strips of film.
I know it is a cliche, but newbies don't know how good they have it with computer-based electronic editing vs. the "bad old days"

Remember also that "drop-frame" has absolutely no affect on how the video or audio is recorded.
It is only a way of diddling the numbering scheme so that the frame numbers match the wall clock after color television changed the frame rate to 29.94 FPS....
Click to expand...

Actually there are two meanings to Pull- down/up in the film world. The one every body talks about is the 3:2 pull-down used for transferring film to NTSC video. You are right that that act has nothing to do with sound. However since the US went to color and changed playback speeds to 29.97 it does affect sound since once you slow down the picture you need to slow down the sound. That option in DAW's was pretty much universally labeled as a pull up/down. So it is not incorrect to talk of the sound as being pulled down/up, though it is a bit confusing. Better terminology was "film speed" and "US video speed", since that is what we are really talking about.

The other "pull down" is when you grab some of the start of one reel, copy it and "pull it down" to the tail of the reel before it. This was only done with the audio and it was so that you wouldn't have a gap with reel change overs during playback. It's around 3 sec. if I remember correctly. It always was a bit harrowing since it was a physical edit to the print master, so any mistake could be very costly.

Also I think you meant 29.97, not 29.94. But you are correct about drop frame. It could be a problem back when people were syncing analog tape using TC but not these days.
 
You must log in or register to reply here.
Back
Top