TOPIC:

why resampling at first ... and then why 96 kHz
Resampling is not exactly a good thing. There are few theories, that say good resampling is so power hungry operation, that event modern PC CPU cannot do it in real time.

I personally couldn't hear a difference compared to another naim streamer, that don't resample ... bit still just curios

Hello Rusev,
to my knowledge, the internal sample rate has to be fixed to ensure that a predictable amount of DSP resources is used. A rate of 96 kHz was probably chosen as the highest at which the DSP chip was able to run the required program. Reasons to prefer higher rate could be various, e.g. the ability to accept more formats without downsampling, better DAC performance at higher sample rates, better quality of some DSP algorithms when implemented at higher sample rates, or something else.

These are just assumptions, it would be nice to know the official answer indeed.

Wanderer wrote:

Hello Rusev,
to my knowledge, the internal sample rate has to be fixed to ensure that a predictable amount of DSP resources is used.

If so, then why dont leave at least 44.1 unchanged. Its still the most common sample rate and the conversation in 96 its not exactly trivial to be done right.

It would, I suspect, be a "non-trivial" change to their software architecture. Also, anyone using advanced biquads or FIR filters would have to generate coefficients for every sample rate. OTOH resampling allows them to support non-standard sample rates as well, and if as you say you can't hear a difference....

44.1kHz is the worst sample rate for sound quality and there are technical reasons why it's hard to do good filtering at that rate. Without getting into sampled data theory it's difficult to explain why.

A simple example of the problem is that if you have a low pass filter which falls to zero at infinite theory with continuous signals, with sampled data that 'zero' is warped to the Nyquist frequency, which is at 22.05kHz for a sample rate of 44.1kHz. Having that zero at 22.05kHz instead of infinite frequency makes a huge difference to the rolloff of the low pass filter below 20kHz and well down into the audible frequency range. There are ways of faking the correct finite attenuation at the Nyquist frequency but they are approximations to what you really want. It is far easier to just raise the sample frequency to move the frequency warping away from the audible range.

I understand the internal 96 kHz, but is the output limited to 48 kHz? I've changed the output to 32 but 2-channel, but the DAC continues to read 48 kHz. I'm streaming Roon to it from my Bryston BDP-1, which outputs at 192.

Is 48 kHz the output limit?

To the answer why 96kHz or more simply, why a single sample rate, here are some comments:
- A single sample rate is always used for a DSP as all algorithms (FIR/IIR/delay... etc) are all dependent on that variables. If we had flexible sample rate, aka 44.1k in, 44.1k DSP, we'd need as many set of filters as we'd have potential sample rate. Forget Dirac live, forget REW, forget simplicity..
- SAmple rate conversion isn't as bad as seen/envisioned when it runs at Dynamic Range of 130dB as the one we use.. There are a lot worse issues to worries in an audio chain..

@ Steve, All processing (analog and digital) comes out at 96k. Our Audio Precision unit shows the correct rate as metadata so we're not sure what could be the reason for your DAC reading 48k. Maybe some odd metadata issue on your unit.. Maybe start a tech support ticket?

DevTeam

(withdrawn)

Something interesting i think ...
after the last update from volumio (jan'20) now the SHDstudio accept the resampling form roon. I still experimenting, but there is definitely a difference in the sound if it is resampled to 96khz from roon and then send to the SHDstudio. Sure, small difference but with the right equipment audible when switching back and froth. Can't say what is better or what is right or wrong, but obviously resampling is not the same.