Level increase when exporting and resampling?

[jono grant]

Hi there,

I'm exporting a final master from Cakewalk. I have the final output of the limiter peaking at - 0.2, when I export at the native sample rate of 48 kHz, the peak is the same.

However; when I export the file and choose to change the sample rate to 44.1, the exported file goes over that maximum peak and hits zero.

Why is that happening and is there a better way to re sample the file, retaining the accurate peak maximum?

Thanks

Jono

 

[jono grant]

Do I need to re sample the file first and then bring it into a new cakewalk session at 44.1 kHz and then redo the mastering process again?

J

[jono grant]

Anyone know about this?

Thanks!

[Gswitz]

Interesting. I've never seen that.

Any chance other audio is getting in? 

If you import and do a null test does it match? It should.

Edited August 17, 2019 by Gswitz

[jono grant]

10 minutes ago, Gswitz said:

Interesting. I've never seen that.

Any chance other audio is getting in? 

If you import and do a null test does it match? It should.

Thanks! Not sure what you mean about the null test... no other audio going in, it's the same process for the 48 kHz export except for the re-sample.

After exporting, I bring the file into sound forge (it has a good wave form preview) and the re-sampled track has peaks over the set limiter setting of -0.2. The one that isn't re-sampled sits perfectly into the -0.2 range.

I think when it does all the mastering processing AND gets re sampled at the same time, it goes over. I'm going to try to re-sample the track before the mastering processing and I'm guessing it will export properly. It's odd to me though.

I'll report back.

Thanks

Jono

[jono grant]

Confirmed... if you re-sample the track and do the mastering/limiting work in a new file with the new sample rate, the same exact mastering process does not peak.

So, re-sampling while exporting a track with processing will not sit inside the threshold. Too bad, it's a time saver!

If anyone knows any more about it, please let me know!

I will try the same process using a track without any plugins and see if the re-sample causes any change to the peaks.

Cheers

Jono

[Gswitz]

You do a null test by re-importing the file you just exported into a new track exactly lined up with the clip you exported.

Do this by using ctrl+g to move to the beginning of the selected clip, set focus on the new track then import the file you exported.

Now you have two copied of the same clip aligned. If you flip the polarity (phase) on the track you just imported and play the two together you should get silence. If you bounce the two clips together and normalize the new clip you should get flat silence.

[Quick Math]

Sample rate conversion can increase the peak.

Technically SRC is done with a filter internally which is a low pass filter and It's a liner phase filter as far as I remember.

And because of that, the peak slightly shifts as in EQ cut. EQ cut actually may increase the peak, even though It's a cut.

I assume resampling is also the same thing so that's probably why the peak changes.

Edited August 17, 2019 by mikannohako

[jono grant]

18 hours ago, Gswitz said:

You do a null test by re-importing the file you just exported into a new track exactly lined up with the clip you exported.

Do this by using ctrl+g to move to the beginning of the selected clip, set focus on the new track then import the file you exported.

Now you have two copied of the same clip aligned. If you flip the polarity (phase) on the track you just imported and play the two together you should get silence. If you bounce the two clips together and normalize the new clip you should get flat silence.

Right, okay, thanks! A couple of things may not work here though. Lining up the file might be difficult as the original is going through a bunch of plugins and may have a bit of latency, even with plugin delay comp happening. Also, to bring the 44.1 k file back into the 48 k session,  it wants to be re-sampled to the session rate of 48k so I'm not sure the experiment will tell me much. I can tell visually that the exported/re-sampled file is different.

What I found is that you need to re-sample the audio first and work in a session that has the same sample rate before adding your mastering processes. Perhaps it has something to do with the mastering plugin processing getting re-sampled and truncated or something, causing the file to peak a little beyond the ceiling that was set in the limiter.

Anyhow, thanks

J

Edited August 17, 2019 by jono grant
add more info

[Gswitz]

I wonder if our difference in process might explain some of the diffs in result.

I usually fast bounce to a new track then export that clip with no processing. This makes the export go super fast and I can export to different bit-depths and sample rates where that will be the only difference.

I don't know if this helps but it might help you isolate the difference to only the exports.

 

[jono grant]

On 8/17/2019 at 7:44 PM, Gswitz said:

I wonder if our difference in process might explain some of the diffs in result.

I usually fast bounce to a new track then export that clip with no processing. This makes the export go super fast and I can export to different bit-depths and sample rates where that will be the only difference.

I don't know if this helps but it might help you isolate the difference to only the exports.

 

That's a good way to do it! I thought it might make sense to do all the master processing while the audio was at it's highest sample rate. I guess I can't really do that unless I leave the final limiter/maximizer to be the final process after SR conversion.