Why is my audio consistently a few frames off when I sync by timecode?

[bda92]

Hi everyone.

Recently, we acquired some Tentacle Sync timecode generators at my office job to be used in conjunction with our Sound Devices MixPre 6. We have a multicam shoot coming up where we will be using two Sony FX6s, a Sony FS7, and possibly even a Sony FX3. While testing our set up, I was able to sync the timecode between all of the cameras and the Sound Devices with no issue. When I synced the video and audio in post though, I noticed that the MixPre's audio was a couple of frames off from the camera's onboard audio.

 

To figure out what I did wrong, I then set up just one FX6 and the MixPre to do a simpler test. Still, when I synced the audio and video in post, I can visually see that the waveform from the MixPre is one or two frames ahead of the scratch track. I actually get a more accurate match if just eschew timecode and sync the clips by the audio waveform.

 

I have tried finding forum posts where others have had the same issue, but this has only lead me to further confusion. I've double-checked my frame rate and tried leaving the Tentacle Sync plugged into the FX6 rather than simply jamming it. Still, the sync is slightly off. My impression is that everyone I have seen do this either in person or on YouTube has no issues with sync. So what could be happening here? Is this latency maybe because the microphones are at different distances? Is there something I just fundamentally misunderstand about timecode? Am I expecting too perfect a sync? If so, why would I use timecode to sync audio in the first place? I'm just not sure what to think about this.

[acdave]

I had a similar issue with a MixPre3 v 2 recently. Also with Tentacles.

Please confirm that all devices, including cameras,  are running the latest firmware and report back. 

Good luck!

 

 

[Bouke]

No idea what the reason is, but something alike was discussed some time ago.
If the offset is constant, you can now use QTchange to do an offset to the BWF timestamp. (In frames, milliseconds or even samples.)

 

[The Documentary Sound Guy]

Is the scratch track actually in sync with the picture in camera (i.e. is the camera's internal A/V sync accurate)?  And are the camera's timecodes in sync with each other?  I'd be inclined to suspect the FX6/FX3.  Particularly when they were first released, I remember seeing reports that those cameras didn't always compensate properly for the internal processing delay, and would show 1-2 frame sync issues depending on exactly how the timecode was routed into the camera and how much internal processing the cameras were doing.  I recall one of the cameras would show a two frame discrepancy from the same timecode source depending on whether the timecode was plugged in to the dedicated timecode input or the audio input (I think it was an FX3, but I'm relying on memory here).

 

If you search, you'll probably dig up a couple threads on this forum about the issue.  I haven't seen anyone complain in a while, so my assumption is that the problem was eventually fixed in firmware, but I haven't had to deal with those cameras recently.  So, yes, update your firmware.

 

On a more practical note, if waveform sync is working and more accurate, just use that.  Timecode is useful, but only when it's working properly.  I'm not here to tell you it's perfect or that you need to use the one true timecode workflow.

[bda92]

I have done a few more tests since this last post and have a few updates to report. The Documentary Sound Guy's comment had me suspecting that some sort of processing delay was happening in the camera and my subsequent testing has made me even more suspicious.

 

I updated the firmware on the recorder and the Sony FX6 to the latest versions and tried another test, but the FX6 audio was still off from what I recorded with the Sound Devices MixPre-6. Meanwhile, I didn't seem to be having issues, or at least not as noticeable issues, with the FS7 and FX3.

 

I then got an idea and downloaded the Tentacle Timebar App so I could clearly display the timecode on my phone. I then sent timecode from the Tentacle Sync to the FX6 and filmed my phone's screen. In theory, each frame of the video should show timecode that perfectly matches the metadata of the file. However, when I pulled it into Resolve and compared, the metadata was consistently two frames behind what was displayed on camera. For another shot, I sent timecode out of the FX6 to the Tentacle Sync and in that case the metadata was three frames behind!

 

I am quite confused by all this. The good news is, it seems to be consistent and can be fixed (though I haven't checked with a longer clip to make sure it doesn't drift over time). Still, I'm shocked that the timecode presumably matches perfectly during recording, but then shifts two frames when the file is saved.

[The Documentary Sound Guy]

Another test you can try:  Film the timebar and the FX6 screen (with the timecode display active) side by side and see if the display timecodes match.  That could narrow down whether the problem is that the FX6 is jamming inaccurately, or if something is causing the metadata to lag after the timecode is jammed.

 

Trouble is, no matter how much you test, all you've really done is establish that the FX6 isn't reliable.  Beyond that, it's hard to fix it unless you can somehow push it up the chain at Sony to the point where they can release an engineering fix.

You might be able to compensate for that camera specifically if your tentacles can specify a timecode offset, but I don't think they allow that (and having to add a manual offset for specific cameras is extremely error prone, since once the offset is active in a timecode box, it's now wrong for every other camera.

Another test:  See if the frame offset changes at different frame rates.  Higher frame rates tend to have higher processing delays, so if it does change, that makes it more likely that the processing delay theory is correct.  Still doesn't help fix the problem though.

[Bouke]

Hmm.
Filming displays with TC 'used' to be a bad idea, as displays tend to have enormous lag as well, and the readout often walks a different path than the LTC out.

I have no clue what's in the FX6, but video undergoes a tremendous amount of processing in 'any' camera. (Only once the image is captured it will be processed.)
So I can imagine the muxing with sound is done afterwards when the lag of the video path is known (based on recording settings.)
And that part is omitted for the TC.
I like the idea of testing different frame rates, but also test stuff like skin smoothing / whatever fancy stuff is on the cam.
And of course, fix it in post!

[inspire]

You can make "clap test".
Audio and video sync: If all cameras and audio are tested, you can determine if all cameras are recording according to the same timeline or if one of them needs a different audio delay in post-production.