I want to use a 192k sample rate for input from dac~, do some processing that requires the high sample rate, and then have the rest of the patch operate at 48k, because I don't think my computer can handle the whole patch running at 192k and I don't need 192k for most of what I'm doing.

Is it as simple as the project being set to 192k, the main patch running at that, and then all the lower sr stuff being embedded in a patch with a block~ object downsampling to .25? I won't lie; up/down sampling kind of bends my brain, and I'm not sure if there will be a bunch of inintended audible consequences, or if the downsampling/block strategy will save me as much cpu as I'm hoping.
Any hints or comments?
Thanks

Before downsampling, lowpass the signal to remove frequencies above the new Nyquist. If your target SR is 48kHz, any frequencies in the signal above 24kHz will fold over into the range 0-24kHz, which you don't want. Your ADC (hardware) filters the analog signal before sampling for this reason.

I'm not an expert in this type of filter design.

Since you will lose everything above 24kHz anyway, what if you ran your system at 48kHz and up-sampled just for that bit where the higher SR is necessary?

(FWIW I'm skeptical of high sample rates for most applications. They make sense for sound effect libraries, where pitching down is a very common operation. Most other cases, it's gravy -- so I'm curious what is the use case here, for bringing the gravy into the system but then downsampling it away.)

hjh

Use a larger block size?

@ddw_music
I'm doing some experiments with cross correlation; hoping I can locate a point source in a stereo field. The higher the sample rate, the better the resolution I can get. So I need the original data at high frequency, but after I do the cross correlation calculation, I'll process it in more normal audio ways that don't require anything above 48k. That's why I think hardware input needs to be at 192k, rather than upsampling a 48k input.

Thanks for your pointer about filtering!

@dreamer would that be for reducing cpu further after I've downsampled?

@mbbaker said:

@ddw_music
I'm doing some experiments with cross correlation; hoping I can locate a point source in a stereo field. The higher the sample rate, the better the resolution I can get.

That's fair -- thanks for explaining.

hjh