What sample rate do you guys record and mix on?

I didn’t know there was a big market for Bats buying music - you live and learn

I wish I could remember the reason I was given, but at least one celeb engineer friend has told me that 88.2 is better than 96k. The reason was technical IIRC...maybe about being a multiple of 44.1? I'll shut up now.

I got a new computer a couple of years ago (M1 Pro MacBook Pro) and been working on tracks at 192 as an experiment. Just audio recording here, no software instruments...at about 18 tracks, it's starting to buckle. I'm looking forward to finishing it and going back to something more reasonable to see if there's any perceptible difference. I've read that there's literally no reason to record at 192...I've always wondered why it exists then? Anyway, looking forward to descending back down to 88.2 where I was before. Audio performance in general is a little dicey up here in the clouds of 192.

The first Pacific Microsonics converter ran at 88.2 and 44.1. It could convert 88.2 to 44.1 in real time and sounded better than anything else in the'90s if you didn't have all day to process each song.

Does running a higher sample rate change the latency on the Hearback?

Right. I forgot that limitation of lightpipe.

I realize you were asking a specific question, but I'll use the opportunity to comment on the general idea. People often cite decreased latency as a benefit of running at higher rates. The problem is that we don't have latency from sampling, we have it from buffering. Buffering is there for a reason. For computers, it's mainly because you need to limit how often the host system is doing context changes. At 48k, that would be every 21 microseconds, so if we buffer 256 samples, we're at 5.3 milliseconds. We need buffering both ways, so that gets us over 10 ms of latency, single buffered. Someone might say that at 192k, a 256 samples buffer will cover a quarter as much time, and our latency is back down to to about 2.5 ms. But if your system can handle 2.5 ms between context switches, it can handle a 64 sample buffer at 48k for the same latency.

And that's before noting that any processing (plugins, etc.) take four times the processor power at 192k, because the data rate is 4x.

The latency argument for higher sample rates is a red herring, but it's often cited as a legitimate reason (look at GS).

2. Higher rates...basically, the arguments fall into two categories, frequency and phase. Can we hear the added frequencies? For the people who think we can, there's an embarrassing lack of evidence—we'd expect it could be proven easily with double blind testing. So, until someone proves we (or some) can hear higher than 20k, I feel pretty good about this one not being an issue. Some also contend that natural instruments can have significant components up to ~40k, so we should use higher rates just to be sure we capture everything that might possibly be hearable...again, even though they can't demonstrate that they can hear it.

As for phase, it's mainly the idea that for the base rates (44.1/48k), the filters must be very steep. For analog filters, there is no choice but to have a large phase shift around the cutoff frequency. (A different issue with oversampling filters, where the first stage is digital and is often linear phase—for that, people worry about "pre-ringing". Feel free to ask if you want to discuss more.) The amount of phase shift is much greater than with analog gear and anything natural, so people contend that why digital audio doesn't sound as good as analog to them. Again, something that's not been proven with double blind testing, but it's an argument.

In case anyone missed it, I'm happy with the base rates, and it leaves me a lot more processing power, but 88.2/96k at least addresses the two arguments above. The frequency argument is probably not as strong, since it's so easy to prove if it were true. For the phase argument, it probably comes down to what sample rate would allow a filter that is hard to argue against. IIRC, Dan Lavry wrote article that put the optimal rate at around 66k, which makes a lot of sense—high enough to nuke arguments about the filter characteristics, without being excessive. We have 88.2k and 96k as standard rates—more wasteful, but they are the lowest that would address the filter characteristics arguments.

3. Why 192k+...Some people feel that the more samples, the closer to analog (due to lack of understanding, of course). One pro on another forum said he does everything in 192k, always. And the higher rates have been supported on recording interfaces since back when they all sort of sucked at any supported rate. 

I'm not shaming anyone for using higher rates, just explaining reasoning behind my choices!