Downsampling from 96k to 88.2k

We've been working on a project recording heavy distorted guitars and clean guitars into Sonar XI Producer @ 24/96.  The working drum files were sent to us by our engineer and come to find out, they were @ 24/88.2.

What I didn't immediately realize is that Sonar converts the files as they are imported, without asking.  So the default 24/96 project setting was on and all files that we were sent were thus converted.

We went on our merry way recording all tracks and now, I'm being told that we may have a serious problem converting what we've recorded now from 24/96 to 24/88.2 without likely artifacts and other audio issues.

Conceptually, I understand why this would happen, especially due to the sampling rates being non-multiples.  But the skeptic in me says otherwise?  I tested with a simple vocal and didn't hear anything bad when going from a 96k project file and importing it into an 88.2k project file...

Can anyone dispel the myths?

Thanks!

In terms of the math, resampling signals at a different sampling rate can be completely lossless. It is not required to have an integer relationship between the sampling rates.
However, accurate resampling is computationally expensive. Some simplifications in the math can make the process a lot more efficient. Thus, you may find different options in the software that allow you to prefer speed over accuracy.
For some reason, 96 kHz is very widely used. Given that many productions are intended to be put on a CD, 88.2 kHz is an excellent choice. I would think that the final down-conversion to 44.1 kHz could be implemented more efficiently from 88.2 kHz than from 96 kHz.
So, if you got the drum files first, the best approach would have been to also record your tracks in 88.2 kHz and make 88.2 kHz the sampling rate for the whole project.
Now, the up-sampling of the drum tracks to 96 kHz obviously worked okay, i.e. I don’t hear you complain. So, the best idea would be to leave the project in 96 kHz and be merry. The final export may require re-sampling to 44.1 kHz anyway. Thus, converting the project to 88.2 kHz know is pointless.
Just to answer your original question: Converting everything form 96 kHz to 88.2 kHz is possible, and it is not a problem. It will just take long, and it is an unnecessary step if 88.2 kHz is not the final format.

Reducing sampling rate and also bit length should always be the very last step.


If you want to google for it, some keywords are: sinc-interplation, re-sampling.
https://ccrma.stanford.edu/~jos/resample/
 
Don’t worry. Everything will work just fine.
 
Wilko
 

You will have no problem importing the file as long as the project and the file you are importing are both 24 bit (or the imported file bit depth is lower than the project). The khz rate doesn't matter at all. If the bit rate is 32 or higher, you should enable dithering (unless you have Sonar set to import @ 32bit). You always have to dither when reducing bit rate.

Are you sure he didn't mean the drum track was 88.2/32 and that's what he's worried about?

Edit: Had some facts backwards and fixed them.

Positive.  The drum tracks are at 24/88.2 and to paraphrase: "It would be very audible to upsample all the master drum tracks to 96/24 - you'd hear horrible artifacting, particularly on the cymbals, especially with the cross - SR conversion that has to happen (non-multiples etc)."

Bub


You will have no problem importing the file as long as the project and the file you are importing are both 24 bit (or the imported file bit depth is lower than the project). The khz rate doesn't matter at all. If the bit rate is 32 or higher, you should enable dithering (unless you have Sonar set to import @ 32bit). You always have to dither when reducing bit rate.

Are you sure he didn't mean the drum track was 88.2/32 and that's what he's worried about?

Edit: Had some facts backwards and fixed them.

Maybe the most effective counter-argument is to perform an A/B test for your engineer. Convert a good-sounding 88.2 file to 96 and play both of them for him. See if he can even tell the difference, much less hear "horrible artifacting".

This is indeed the next step... thank you.

bitflipper


Maybe the most effective counter-argument is to perform an A/B test for your engineer. Convert a good-sounding 88.2 file to 96 and play both of them for him. See if he can even tell the difference, much less hear "horrible artifacting".