Zoom F6 Audio Drift Despite Matching Timecode

[23haro]

I'm experiencing a very puzzling issue regarding externally recorded audio files exponentially drifting when sync'd with my video files. All audio and video files are at a matching timebase of 23.98. 

I'm using Sony FX9, FX6, and A1 XAVC-L files along with 32 bit Float audio files recorded externally on a Zoom F6.

At the time of this posting I am using the latest and hopefully greatest version of Premiere (23.1). 

Any insight on how to remedy this would be much appreciated. Thank you.

[Remote Index]

Hello 23haro,

It's difficult to troubleshoot such issues, and you haven't provided much info.

One additional piece of info that might help you figure it out is if you can come up with a figure for the amount of drift. How many frames (short or long) per minute? 10mins? per hour? Sometimes the drift amount starts to point in useful directions for problem-solving.

R.

[23haro]

The drift occurs as soon as the video begins. It’s subtlety off from the first frame. Then it is very out of sync within the first minute. The sync issue is far worse off after ten minutes and even worse at twenty minutes.

It plays as if the audio is being interpreted at 29.97.

[Remote Index]

Can you quantify it a bit more precisely?

Try working in a timeline with both image and sound clips.

Find a hard sync point (something obvious in both image and sound) at the start of a clip; line them up so they are directly in sync. Put a clip marker in image and sound. Then find a hard sync point at the end of a clip. Put a clip marker in image and sound.

1. Using in and out markers, measure the duration from marker to marker on image, and the same on sound. What are the two values?

2. Using a speed effect, can you change the speed of one clip to match up the markers (or get close). What is the speed value that acheives this?

If you do the above for more than one clip, do the relationships look the same?

The above will not solve the issue, but it will give you some information to work with.

Also - what sampling rate did you record at? What is Premiere Pro reporting as the sampling rate for the audio clips?

R.

[23haro]

Thank you for your suggestion, R. I attempted your solution but unfortunately the issue was not solved by adjusting the speed in the timeline. Any speed adjustment would divert the drift to the opposite side of the video clip (i.e. I would adjust the speed accordingly to sync the back half of the timeline, only to make the front half of the timeline suffer from drift). I've never experienced this issue in Premiere and presume there is either a bug within my Zoom F6 or Premiere.

The sampling rate is 48hz. Premiere is reporting the rate as such.

I found this video which solved my issue but is a workaround.

[23haro]

One more thing - I have been able to perform a solve in Premire using the 'rate stretch' tool. So all is well. But again, the root of this problem remains a complete mystery to me.

[Remote Index]

I am glad you found a solution.

Out of interest - what sort of (revised) speeds did you end up with after using the "rate stretch" tool?

I found the manual for the Zoom F6 here.

The Zoom F6 has some advanced functions regarding sample rates - some rates are "stamped" or identifed in metadata as slightly different than the actual recording rate, and I had suspected that this is what was happening with your material - but the variance should only be about .1% (or 1 frame in a 1000). (If you do a search for "setting the sampling rate" on that doc you can read it, though it is probably not very clear if you're not aware of the function to begn with). And of course it's difficult to say if this is the issue you were facing without more info.

R.

[23haro]

I'll take a peek into the Zoom manual. Thanks for suggesting that, R.

I've attached a few screengrabs for a few of my different interviews showing the varying differences of my corrected clips vs what was externally recorded. The numbers are inconsistent.

[23haro]

From the Zoom manual:

44.1 kHz, 48 kHz, 88.2 kHz, 96 kHz, 192 kHz: These are standard sampling rates.
47.952 kHz: Select this when recording video at 23.976 frames per second in order to edit later at 24 frames per second.
48.048 kHz: Select this when recording video at 24 frames per second in order to edit later at NTSC 29.97 or 23.98 HD.
47.952 kHz(F), 48.048 kHz(F): These function the same as the two above, but the sampling rate metadata will be recorded as 48 kHz for . This enables playback and editing with devices and software that do not support 47.952 kHz and 48.048 kHz WAV files. Playback, however, will occur at the ±0.1% speed at which the file was recorded.

Strangely, I recorded everything at 23.98 so I am still perplexed. I suppose I need to conduct some tests with my gear in Premiere.

[Remote Index]

Hi 23haro,

Yes that is the section of the manual I was referring to.

The framerate or timebase (in your case 23.976) is distinct from the sample rate of an audio recording. Think of them as 2 different "rates" or quantities per second. In other words, you can record sound at 23.976fps which is 48KHz, but you can also record sound at 23.976fps which is 47.952kHz. One is indicating samples per second, the other is indicating frames per second. Setting one correctly does not mean the other is set correctly.

What the manual is telling you (in the section you've quoted) is that it is possible to record at one sample rate, but "flag" or "stamp" the file to be played back at a different rate. The framerate setting is just a way that the file will be labelled with timecode (after the speed change); in other words, the frame rate and the sample rate setting on this recorder does not guarantee the file will be played back at the same speed it is recorded.

If this was set incorrectly, the speed difference is (as I said) 0.1%. This incorrect setting can be corrected fairly simply with software like Wave Agent from Sound Devices (but you have to know what you're doing).

Again, I am not sure that this is what is happening with your files, but it is one likely possibility.

The screen grabs you've provided don't really help. They show an offset, but again, the information that is needed is the relationship between the original and corrected clip (so for example, the original length of a clip and the length after you've used 'rate stretch'; or alternately just the value / percentage of the 'rate stretch').

R.

[23haro]

Hey there,

Revisiting this thread. It turns out that my Zoom F6 recorded the 32 bit float files in 48.048 KHZ not 48KHZ. I suspect this is the source problem. In Premiere I have eyeball adjusted (via rate stretch tool) every associated video clip to be .1% slower. This only affects clips of a certain length, however, as it seems that it needs a clip of a longer duration than say 2 minutes to produce noticeable time drift.

I found the following info on how to correctly setup a production workflow around 48.048 to not give anyone in post-production a heart attack: https://www.trewaudio.com/articles/48048-khz/

[Remote Index]

Hi 23haro,

Good of you to return with an update and additional info. Hopefully it will help others faced with similar problems.

If you didn't check it out, have a look at that Wave Agent software.

R.

[23haro]

Thanks for the reply. Unforutnately Wave Agent states they don't support 32 bit float files, which is what I'm using.