Upsampling in Roon 1.3

[Spindrift]

Would totally agree with Confused on impressions at 176.4, though a multiple of 2 rather than 4 (88.2) also sounds quite nice with a balance between articulation and relaxed presentation.  So far my experiments with dsd via dop with the mutec USB passing the dsd signal to the devs (I.e mat engine) have resulted in a slight loss of detail - too smooth for my tastes . . will have to try mutec USB conversion next.

Are you sure you are outputting DSD to the Devialet via the Mutec?  My understanding was that the Mutec can accept DSD/DoP but will only output PCM.  For example, if I send DSD64 to the Mutec and then check what the Devialet is receiving it indicates 176.4kHz.

I could well be wrong here.  Does anyone know for sure that the Mutec can output DSD?  Previously my understanding was that it can't.

As another update, I have just tried running the microRendu direct to the Devialet via USB, with Roon upsampling to DSD64.  In this set up the Devialet displays DSD64 indicating that the 'MAT' algorithm is indeed in use.  To my ears this did not sound very good at all, very disappointing.  Everything sounded somehow thinner and a touch harsh.  There was the also occasional 'cracks', very similar to that you would get with vinyl, so I am wondering if something was not working 100% as it should.  Strange, because this issue was not apparent running DSD64 to the Mutec.

My conclusion so far is that the best set-up is either to use Roon to upsample to 176.4kHz, or alternatively turn off Roon's DSP engine and send the Mutec good old native 16/44.1.  I would need more time to separate these two though.

EDIT: I found this on CA, a post by Mutec's Julian Davidin response to a similar DSD query:

"The MC-3+USB does not pass through DSD64 without converting to PCM as this would either require re-clocking the native DSD signal or disabling the re-clocking stage entirely. If we were to disable the re-clocking, then there wouldn't really be a benefit to having the MC-3+USB in the chain. What exactly are you trying to do?

Julian"

Thats what I thought as well . . and I think if you feed the Mutec USB via a USB signal it will always covert to PCM as Julian mentions.  However if I feed the Mutec an AES/EBU signal (which is the DSD/dop signal) it passes the PCM shell to the Devs without conversion as far as I can tell (the Devs report DSD64 on the display).  I have an older Mutec 1.2 USB converter from a couple years ago that I used to convert the USB signal from Roon to AES/EBU.  In any event - as you say - not an improvement . . I'm leaning towards turning off the DSP engine.

[marcor]

I tried  Roon Upsampling yesterday afternoon in different combinations.
First I did try PCM  up to 192/24 through ethernet. First album I tried was Eric Clapton - Unplugged. The different was quite perceivable, the sound was more airy and spaced, definitively an improvement. Just to make sure I wasn't imagining things I asked both my wife and my daughter to do a blind test and both of them perceived the difference.
Then I tried the same thing on other records both CD and Tidal and the results were a lot less noticeable and in certain cases not noticeable at all.
One thing I noticed is that you can have clipping even with PCM upsampling. I enabled headroom with value of 0 to check and the clipping red light started to go off. A bit of headroom did the trick. As an hindsight Probably Roon should have implemented the clipping check outside the headroom DSP engine.

Bottom line, depending from the source the mileage of the PCM upsampling could vary.

Then I did try the DSD upsampling using a MacPro connected to USB (admittedly the cable was a cheap USB cable as this is not my usual set up so I had to do with what I found at home). I found it very difficult to compare due to the difference in volume, I had to raise quite substantially the volume to get to the same level as PCM so a straight A ->B comparison was not possible. Listening at the same material on DSD left me a underwhelmed, I felt the sound too soft, lacking of punch, kind of muffled up. Not sure if this was due to the cable and as I don't have original DSD recording I cannot compare them with upsampled ones.

Bottom line, considered my set up I am not sure I want to go into the fuss of having a permanent USB connection to my Devialet to allow the play of DSD. I will wait patiently that DSD is available for streaming for Devialet, at the present I don't think I am missing a lot.

For completness my set-up is pretty basic
iMac 2012 (i5 quad core 3.1 GHz 24GB RAM) connected to a Devialet 250 PRO through giga ethernet using AIR release candidate 3.0. Speakers are Sonus Faber Amati Futura, speaker cables are Oyaide Tunam II
The computer connected to the USB was a MacBook Pro i5 dual core 2.7GHz 8GB RAM
With both the computer I didn't have any computing issue in running the upsampling.

Cheers
Marco

[Confused]

I tried  Roon Upsampling yesterday afternoon in different combinations.
First I did try PCM  up to 192/24 through ethernet. First album I tried was Eric Clapton - Unplugged. The different was quite perceivable, the sound was more airy and spaced, definitively an improvement. Just to make sure I wasn't imagining things I asked both my wife and my daughter to do a blind test and both of them perceived the difference.
Then I tried the same thing on other records both CD and Tidal and the results were a lot less noticeable and in certain cases not noticeable at all.
One thing I noticed is that you can have clipping even with PCM upsampling. I enabled headroom with value of 0 to check and the clipping red light started to go off. A bit of headroom did the trick. As an hindsight Probably Roon should have implemented the clipping check outside the headroom DSP engine.

Bottom line, depending from the source the mileage of the PCM upsampling could vary.

Then I did try the DSD upsampling using a MacPro connected to USB (admittedly the cable was a cheap USB cable as this is not my usual set up so I had to do with what I found at home). I found it very difficult to compare due to the difference in volume, I had to raise quite substantially the volume to get to the same level as PCM so a straight A ->B comparison was not possible. Listening at the same material on DSD left me a underwhelmed, I felt the sound too soft, lacking of punch, kind of muffled up. Not sure if this was due to the cable and as I don't have original DSD recording I cannot compare them with upsampled ones.

Bottom line, considered my set up I am not sure I want to go into the fuss of having a permanent USB connection to my Devialet to allow the play of DSD. I will wait patiently that DSD is available for streaming for Devialet, at the present I don't think I am missing a lot.

For completness my set-up is pretty basic
iMac 2012 (i5 quad core 3.1 GHz 24GB RAM) connected to a Devialet 250 PRO through giga ethernet using AIR release candidate 3.0. Speakers are Sonus Faber Amati Futura, speaker cables are Oyaide Tunam II
The computer connected to the USB was a MacBook Pro i5 dual core 2.7GHz 8GB RAM
With both the computer I didn't have any computing issue in running the upsampling.

Cheers
Marco

Marco, I am surprised to see you were getting clipping issues with PCM upsampling.  I did not notice this myself.  (although maybe I wasn't looking that hard!)  To clarify, this was Roon PCM upsampling only, no Parametric EQ or anything else?

Does anyone understand why PCM upsampling can cause clipping?  I can understand how PEQ can cause clipping, but with PCM upsampling alone this puzzles me.

[Dr Tone]

Marco, I am surprised to see you were getting clipping issues with PCM upsampling.  I did not notice this myself.  (although maybe I wasn't looking that hard!)  To clarify, this was Roon PCM upsampling only, no Parametric EQ or anything else?

Does anyone understand why PCM upsampling can cause clipping?  I can understand how PEQ can cause clipping, but with PCM upsampling alone this puzzles me.

Brian the brains behind everything pointed me to the wiki when I asked him.

http://kb.roonlabs.com/DSP_Engine:_Headroom_Management

[Confused]

Marco, I am surprised to see you were getting clipping issues with PCM upsampling.  I did not notice this myself.  (although maybe I wasn't looking that hard!)  To clarify, this was Roon PCM upsampling only, no Parametric EQ or anything else?

Does anyone understand why PCM upsampling can cause clipping?  I can understand how PEQ can cause clipping, but with PCM upsampling alone this puzzles me.

Brian the brains behind everything pointed me to the wiki when I asked him.

http://kb.roonlabs.com/DSP_Engine:_Headroom_Management

Ah, a case of inter-sample overs I suspect......  You live and learn, thanks for the link!

[Dr Tone]

Marco, I am surprised to see you were getting clipping issues with PCM upsampling.  I did not notice this myself.  (although maybe I wasn't looking that hard!)  To clarify, this was Roon PCM upsampling only, no Parametric EQ or anything else?

Does anyone understand why PCM upsampling can cause clipping?  I can understand how PEQ can cause clipping, but with PCM upsampling alone this puzzles me.

Brian the brains behind everything pointed me to the wiki when I asked him.

http://kb.roonlabs.com/DSP_Engine:_Headroom_Management

Ah, a case of inter-sample overs I suspect......  You live and learn, thanks for the link!

So far -2 db solved it for me.

[marcor]

Yes a minimum value solved my issue too. I started with -1 and decreased the value until there was no more clipping. I think Iended up to -2 as Dr Tone. I was using only upsampling and obviously headroom at 0 as mean of monitor.

[Dr Tone]

Having headroom management on to deal with clipping in pcm upsampling forces Roon to decode DSD instead of sending it direct to the Devialet.

Doh!

[Dr Tone]

Back playing with converting everything to DSD...

In this case I prefer the linear phase filter. Minimum phase softens everything up to much. No headroom management needed in this case and it matches the volume of DSD passed straight to the Devialet.

Most likely just run dspless in the end.

[Confused]

Myself and a couple of others on this thread have had issues with DSD upsampling in Roon giving a significant reduction in gain. The reasons and solution for this can be found in this thread in the Roon forum, from about post 158. (Posts from two Brians, Brian69 and Roon's Brian Luczkiewicz)

https://community.roonlabs.com/t/dsp-up-.../19293/192