Upsampling in Roon 1.3

[Confused]

Given that the Devialet Expert upsample all incoming bit rates to 24 bit 192 kHz PCM -- including incoming DSD64 -- is there any theoretical advantage to setting the ROON DSP to upsample to anything other than 24 bit 192 kHz for all sources? Is there any theoretical advantage to upsampling to DSD64 since this will be converted to a PCM stream by the Devialet DAC in any case? In essence, what are the theoretical "optimal" settings for the ROON 1.3 DSP engine feeding a Devialet Expert?

I think this is a very interesting question.  It would appear logical that because Devialet's ADH runs in PCM, then the optimum solution is to feed the Devialet PCM.  This does leave the question as to if it is best to perform this via Roon prior to the Devialet, or just send the Devialet 16/44.1 or whatever and let the Devialet do the work.  I think this is a case or try it and see (hear), I suspect that results will be system dependant and may vary between those feeding the Devialet via AIR, microRendu, Mutec or whatever else.

Again using logic, it would appear that there is no point upsampling to DSD.  Consider that if you take a PCM file, upsample it to DSD, send to the amp, and all that then happens is that the Devialet's 'MAT DSD Engine' will convert it back to PCM.  So you start with PCM, end with PCM, and surely all the processing in between can only be harmful?  This would be my firm view on the subject if it was not for the audition I had last year of the dCS Vivaldi upsampler and clock.  In this demo the Vivaldi was connected via AES/EBU to a Devialet D800.  I had a playlist of tracks on a USB thumb drive, mostly 16/44.1 material.  The guy from dCS stated that the system would sound best upsampling to DSD.  I argued that because the Devialet would simply convert this back to PCM this was pointless, the Devialet runs on PCM, so send it PCM.  We then agreed to try both, we ran through the same selection of tracks via upsampling to DSD, then again upsampling PCM, then again with some select tracks back on DSD.  I knew for sure it would sound better via PCM, the problem was that it didn't, it sounded best via DSD.  To be honest, I have no idea why, and I would like to spend a lot more time listening to understand this, logic still tells me upsampling PCM should be better.  So how does this relate to Roon?  My view is that we all need to go away and try this, I do not think that the end result is predicable.  Also, I do not think that my experience with the dCS can be extrapolated to extend to DSD via Roon.  The dCS Vivaldi and Clock retails for £24k in the UK (gulp), so I would hope this is a little closer to 'state of the art' than DSD upsampling via Roon on a £1k Windows PC, or whatever.  That's not to say that it is impossible that DSD via Roon could be superb, I just think that there are so many variable here that speculating is of limited use.  I suggest we all need to experiment, share results, and see if some kind of consensus emerges.

There are other 'head to head' tests to consider, such as how does Roon upsampling compare to HQPlayer?  All DSD files are DSD, but how they came to become DSD is not necessarily the same.

Then there is another 'head to head' to consider for Mutec MC3+USB users.  For example, I could upsample to DSD via Roon and send to the Devialet via microRendu for Devialet's MAT DSD engine to deal with.  Alternatively, I could send DSD DoP to via mR to the Mutec, and let the Mutec convert DSD to PCM.  This might be interesting, Mutec vs Devialet MAT?

So there is a LOT to try here to work out what works best with a Devilalet.  Here's hoping we can keep this thread reasonably on topic to sensibly share the results.