Out of sync even with timecode?

[dslrshooter]

Asking the pros:

 

Even with timecode boxes that don't drift, can audio and video be out of sync by almost a frame when editing?

 

If shooting at 24 fps that's 41 ms, way more than the ATSC recommendation of 15 ms for lagging video for example.

 

Here's my reasoning: if a sound recorder starts just as the frame number changed and the camera starts rolling  just before the frame upticks, the NLE will align the two clips (audio and video) right next to each other but they started almost a frame apart.  (For simplicity in this example the two devices started on the same frame).

 

What's the point for manufacturers to boast of boxes staying in sync by half a frame for a day long if you can be off by one frame right off the bat? Am I missing something?

 

Thanks!

[Bouke]

you are not missing something if you are a dslr shooter.
But, 'proper' cams have genlock to overcome this issue.
Then there is more math to be done. Transmitters introduce a delay. Even ProTools renders introduce a delay. (That could / should be counteracted....)
Distance of mic causes a delay. Front speakers to ear (in a large theatre) can introduce a HUGE delay.
Then, even the internal sound of DSLR's is often out of sync due to the cams being not made for pro video shooting.
But there is one bright point (shameless self promotion follows.)
If you record LTC on a sound track, and use my LTC convert software to alter the BWF timestamps to match the video TC, you are subframe accurate.
(As BWF timestamps are in samples.)
This of course is just as accurate as the internal sync of your DSLR.


 

[dslrshooter]

If I want subframe precision on the camera, I need to use audio LTC on the cam too, right? If not, the time granularity on the cam is still ±1 frame, even without camera drift.

 

Also, I looked up BWF TimeReference metada and can't figure how the number of samples since midnight is of any help if the camera doesn't know with a subframe precision "when is midnight".

 

If I understand the usefulness of your LTC convert, it joins both the audio file and the video file with a subframe accuracy using the AUX TC present on both files (given that the TC themselves are in sync). Is this correct?

[The Documentary Sound Guy]

I think you'd be better off asking about a real-life problem you are having rather than a theoretical one.  In practice, no, sometimes sync isn't perfect with complete subframe precision.  *Why* this happens comes down to any number of factors, including the specific devices you are using, how long it has been since jamming, temperature, etc. etc.  Also in practice, generally speaking we don't see "almost one frame" discrepancies very often if best practices are followed.  Also, sub-frame sync accuracy is hard to spot in casual viewing.

When I have seen errors on that scale, the most common cause has been internal processing delays within a camera aren't correctly compensated for in the camera, not problems with timecode.  (I.e. the camera itself may not maintain good video and audio sync.)  This happens mostly with low-end cameras where video isn't the primary use-case (i.e. cellphones, DSLRs and mirrorless cameras).

Here is a long, technical discussion of the issue I think you are pointing at.  It gets a bit contentious at times, and ultimately, I couldn't really get a firm answer to the question, but I don't think your assumption of how sync works is correct.  Specifically, the jamming process does usually synchronize the start of a video frame to the start of a timecode frame with much better than single-frame precision.  How much better probably comes down to the specific equipment.  I think it's possible there are cameras out there that work the way you describe, but in my experience, it's not a common issue.  Also note that this only applies when timecode is jammed ... the only way to guarantee that two devices will stay in sync (especially for very long rolls) is to have them share wordclocks by using genlock.

 

 

[Philip Perkins]

+-1 frame is all they get. 

[Bouke]

On 2/11/2024 at 6:40 PM, dslrshooter said:

If I understand the usefulness of your LTC convert, it joins both the audio file and the video file with a subframe accuracy using the AUX TC present on both files (given that the TC themselves are in sync). Is this correct?

Yes, if you use it to generate new files, it will be subframe accurate, BUT ONLY as accurate as the sound track is synced, and as stated, YMMV.
And, if you change the BWF timestamp to match the video TC, any half decent NLE will place it subframe accurate in synced clips / on a timeline. (BWF has no TC, but a sample count, thus can be positioned highly accurate.)

[Matthew Steel]

Since we are talking theoretical, an audio recorder that is receiving timecode (not just getting word clock that is derived from timecode) could actually know where the frame boundaries are.  In that case such a recorder could pad the recording (using a prerecord buffer if necessary) to always begin every recording on a frame boundary.

 

[dslrshooter]

On 2/11/2024 at 4:02 PM, The Documentary Sound Guy said:

  Specifically, the jamming process does usually synchronize the start of a video frame to the start of a timecode frame with much better than single-frame precision.

🤔 I searched this and I think it is wrong, TC does not usually synchronize the start of a video frame:

 

“ Timecode is not adjusting the video timing within the camera ” (DJ productions, YT)

 

“LTC (Time code) on the other hand is attaching time to the video. Where a Genlock signal is telling the video when to start, time code is assigned to the video of when it happened.” (videoengineering sub)

[The Documentary Sound Guy]

2 hours ago, dslrshooter said:

Specifically, the jamming process does usually synchronize the start of a video frame to the start of a timecode frame with much better than single-frame precision.

I stand corrected then.