I think I am a bit stuck on the fact that it does not mention DSD anywhere. I initially thought that it would upsample to that. This is what it looks like now:
For example if you are playing DSD64 to a non-DSD DAC, with Roon 1.2 it was output to 24/176.
I now have the option to play this same DSD64 to 24/192 if I want (DSD Engine set to MAX PCM Rate) but is it best to stuck to the āpower of 2ā, i.e. 24/176?
Converting 24/176 to 24/192 is no good, or is it ok? What are the issues involved?
The SRC (PCM to DSD) in Roon seemed to sound better (more musical and analog sounding) than the hardware SRC in the Holo Spring DAC (OS DSD).
The Roon does a better job here and Iām sticking the Roon up-sampling algorithm(PCM to DSD) and set the Holo Spring DAC to NOS mode. Enjoy your music!
It is better to set āPower of 2ā this ensures multiple of 44.1k and 48k are SRC without truncation errors. For example 44.1k x2=88.2k but 44.1k x2 is not equal to 96k etcā¦ DSD64 (64x44.1k) normally convert to PCM 88.2k but 176.4k will probably be better. Note there is no truncation error here since DSD uses 44.1k base sample and multiply by 64 times. Hope this helps.
So Iām really having fun playing around with Roonās SRC (though Iām somewhat constrained in what I can do, since Iām running Roon Core on a mid-2011 i5-based Mac mini at the moment).
Almost everything with the upsampling and filtering is working as expected and sounding very good. Iām upsampling most material to DSD128, except for source material thatās 176.4kHz and up, which is more than the Mac mini can handle gracefully, so for those rates Iām doing power-of-2 PCM upsampling.
Just for informational purposes, Iām using a minimum-phase filter (most of the time) and a 5th-order modulator in Roon. Iām outputting via Ethernet to an exaSound PlayPoint and thence to an exa e32 DAC.
The one anomaly Iām seeing is when the source material is DSD64. In that case, it sure looks like Roon is converting DSD64 to PCM 352.8kHz before it converts to DSD128. Do I have something set wrong, or is that how itās sāposed to work? If the latter, thatās not cool, IMO. The intermediate conversion step exacts a pretty severe sonic penalty. To my ears, it sounds worse than just leaving the DSD64 unmolested.
@orgel, yes thatās what it does for now. If you think DSD64 sounds better, use the Custom configuration and tell it to pass through DSD unchanged. Weāll probably do something more bespoke for DSD->DSD in the future.
@lionel_Hightower1, Yeah, weāll probably do 768k that at the same time as DSD512 support. Thereās some discussion of that towards the beginning of the thread. Timeline uncertain, but itās not a big project.
Somewhat disappointed, but thanks for the response. Itās easy to do what youāre suggesting in āCustom.ā Consider this my vote for doing something more bespoke for DSD upsampling. I have a lot of DSD source material, so even though I think you guys have done a great job with SRC 1.0, Iāll stick with HQP for the nonce. BTW, super job on the configuration interface for SRC. Itās way more intuitive than anything else Iāve seen (with all due respect to Jussi).
After extensive evaluation with our server/streamer combo and DSD 256 DAC hereās our thoughts. Currently Daphile with the Clans7 modulator, Very high quality linear phase filter, and very steep checked, is the platinum standard. Second place is Hqplayer with DSD7256+fs modulator/Poly-sinc-short mp filter. Third is Roon.
Once Roon takes over the #1 position, we will gladly have all of our gear Roon certified. Until then, sound is #1 for us.
Iām sure it wonāt take all that much to accomplish as Daphile surpassed HQP in 1 week after we made the DSD-DSD upsampling request which is currently available in the latest beta, with a team of 2 working on it.
Small screen device running in Roon āphoneā mode donāt have access to control the DSP functions.
Larger tablets, PC and Mac that running the full interface do.
That said DSP operations once configured are available individually for each Roon audio zone.
No harm. If you go from 44 to 88 every other sample is identical to the original, and only the intermediate samples are interpolated. It seems closer to the original while adding āinformed interpolationsā vs when you go 44 to 96 youāre really interpolating every single sample. I have never really done a listening test, I tend to use power of two out of respect for the original data. My reasons might be completely wrong.
Iām having trouble upconverting to DSD256 DoP. Since this requires 768 pcm, is this to be expected? Tidal plays for 10 secs from start of track, breaks up, stops and goes to next track. Error message that there is a connectivity or network issue affecting tidal playback.