Sampling Frequency

What sampling frequency do you use internally for recording?

Actually I wouldn’t use more than 48kHz: according to the Nyquist-Shannon sampling theorem that should be more than enough, more is voodoo.

But: If I increase the sampling frequency in Jack, the latency also decreases. What does this have to do with? Could this actually make recording at 192 kHz attractive?
How big are the audio files compared to 48kHz?

Rather than increase the sample-rate by a factor of 4, you might as well decrease the buffersize by a factor of 4.

You get the same effective latency and CPU time constraints with 256 sample/cycle at 48kHz sample-rate as with 1024 sample/cycle at 192kHz. (1024/192 == 256/48)

In both cases, every process cycle takes 5.3ms. Yet in the first case (48k) the CPU only needs to process 256 samples during that time, which results in lower DSP load compared to processing 1024 in the same time.

As discussed above, no.


Except … if you need ultra-low sub millisecond latency. Higher sample-rates can help while keeping buffer-size at 16 or 32 sample/cycle for more efficient processing (SSE/AVX SIMD). This is rather for scientific applications though.

Increasing the sample-rate may also lower the systemic latency of the soundcard a small amount. but the benefit of that is negligible compared to using a smaller buffersize.

Uncompressed raw audio: 4 times larger.

1 Like

ThanX a lot. Sehr hilfreich.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.