This query is probably too specific to be relevant to very many people...

But I've been trying to get the best audio performance out of my computer. I am running Debian (Linux), using Jack, and an external soundcard (Focusrite Scarlett 18i20). By using this nifty program called "xruncounter" I was able to determine that my setup prefers a frames-per-period size as a multiple of the sample rate (48k, so 92 or 144), instead of the usual powers-of-two. However, PD absolutely hates that! Basic sine waves are very distorted (sounding like a sample rate mismatch). SuperCollider is much the same.

I don't know if there's a part of PD that I can adjust to compensate for this. I thought maybe making the Buffer Size also 48 or 92 might fix that issue... but, as we all know, PD only allows Buffer Sizes selected from the drop-down menu, all powers of 2.

Any thoughts? I'm just curious, in general, if anyone has anything to say. It's interesting that PD and SuperCollider don't work right with these settings, but other audio software seems to (though I haven't done extensive testing).

And, if you're using Jack and Linux and are curious about this, here's where I started:
https://linuxmusicians.com/viewtopic.php?f=27&t=19268
https://linuxmusicians.com/viewtopic.php?f=27&t=17033
and my thread:
https://linuxmusicians.com/viewtopic.php?f=27&t=19907

Quick sanity check-- are you using Jack to route the input or output of Pd to/from another piece of low-latency audio-processing software?

If the answer is "no" to that question, then you can remove Jack from your signal chain altogether.

That's a very good point... but yes, I sometimes do route PD to other software. I'm also trying to get my setup so that I can painlessly switch between programs, and/or interconnect them, without a lot of mess.

And - maybe it's just me, but I have been ALWAYS using Jack for a few years now... it's sort of the default setup for KXStudio and AVLinux, which is what I've been running. So it's the workflow I'm used to.

Someone at linuxmusicians pointed out that a 192 buffer size satisfies both 64 and 48... Math! Genius!