Sample rate conversion to 352.8 before DSP – why?

It seems that, whatever upsampling configuration I’ve chosen, when I use a DSP operation like EQ, a sample rate conversion step to 352.8kHz is inserted. This happens regardless of the sample rate of the source. Why this frequency? And why not a factor-of-2 ratio? I get conversion to 352.8kHz even if the source is 96kHz.

I’d like to understand the rationale behind that. Are the DSP filters (I’ve tried procedural and parametric EQ filters) unable to operate at different sample rates? Is this a configurable setting I haven’t found yet?

Shouldn’t be like that in my experience.

Would you care to give us details about your equipment setup, as well as screenshots of signal path(s) and device setup(s)?

Certainly. Mac mini running Roon Core, Topping D90 DAC attached via USB.

Signal path:

If you disable procedural EQ this problem will be solved

Yes, of course. But then I have the problem of not having procedural EQ :wink:

With 44.1 source material the same conversion happens:

And yes, without procedural EQ the signal path looks like this:

What I would like to understand, why 352.8kHz always? What’s special about that?

When u apply procedural EQ , it should be done in PCM not in DSD.instead of DSD if u upsample at 768 kHz or 705 . It will work in one go.

I’m starting out with PCM material (in these examples anyway), apply procedural EQ (still in PCM), then upsample the result to DSD265, which is what I want to send to the DAC.

For some reason, Roon insists on upconverting the source to 352.8kHz, regardless of source sample rate, before EQ. Why 352.8? Why not apply the EQ at the source sample rate? Or at 384kHz if the source is 96kHz? Or at 768kHz, which is the max the DAC can ingest?

In your examples, it can be seen that “volume levelling” and “headroom adjustment” are being done at source sample rate, but as soon as you insert the “speaker settings” step, the forced upconversion to 352.8kHz happens.

DSD is generated through DXD master which operates at 32 bit/ 348 KHZ/ 352.8Hence when we apply any EQ it goes upto 348/352.8first , then apply Procedural EQ and the DXD master up samples to DSD

That would make sense in a way – if it ever used 384kHz, which it doesn’t. Not even with 48 or 96kHz source material.

Ah, I think I can see why 352.8. It is an integer fraction on DSD256 (11.290Mhz), whereas 384 isn’t.