Dirac + DSD via Audirvana+?

[LBob]

I've been holding off on DSD for various reasons (e.g. cost, HD space). Now I'm thinking about getting some new titles in DSD. However, I wasn't sure if DSD and Dirac were compatible. I seem to remember reading on this forum that Dirac could not play DSD files. So I was under the impression that DSD wasn't possible with Dirac. For me that was a game breaker. I didn't want to give up the SQ advantages of Dirac. Then I read in one of Richard's thread that Roon would allow for DSD using Dirac.

 

At present, I'm using and enjoying Audirvana 2.2. I use a beta version of Dirac Live which allows for audio plug ins (I am very non-technical so please forgive me if this isn't the correct description) and my preferred Dirac filter is accessed via a plug in. I had not tried DSD for quite a while and I had never tried it on any of the A+ versions after 2.0, so I decided to try my one DSD album (Dvorack's 9th on Fidelio). I dropped the file into my music folder and A+ recognized it. Hooray!

 

The meta data was scrambled (no track numbers so they were out of order). I corrected that via the A+ editing feature. Then clicked on the first track. My dac first showed 176 then it changed to DSD 64 and the SQ was as good as I remembered. The streaming light was on in the Dirac window, but I wasn't sure that the signal was actually being processed so I opened the window with DSP gain and adjusted the slider. If Dirac was indeed in the signal path this should result in a volume change. It didn't so even though I got the music the filter apparently wasn't being used. I wondered if the A+ settings might be a factor. I had the setting on DoP which was what I had previously used. I changed to "automatic detection" and tried again. This time the DAC displayed 176 and when I adjusted the slider I got a volume change so I know that Dirac was being used. However, I am not sure if I was getting DSD or downsampled 176. I went back to the A+ preferences and the DSD 64 was not lit when auto detect was selected. When DoP was selected it lit up so I'm guessing that my second trial (where Dirac was engaged) the dac wasn't actually getting DSD.

 

So now I have two questions: Is it possible to play a DSD file with Dirac on A+? If not will Roon do it?

[Kal Rubinson]

DiracLive works only with PCM. This is independent of the music player.

Conversion of DSD to PCM is required and, fwiw, what I do by default.

[LBob]

DiracLive works only with PCM. This is independent of the music player.

Conversion of DSD to PCM is required and, fwiw, what I do by default.

 

That's what I thought. Do you have A+ set to automatically convert DSD to PCM?

[miguelito]

As pointed out, DSD effectively does not allow for any signal processing.

 

This is because the amplitude info is encoded in the time domain: In PCM you have a 16 or 24 bit word that represents a snapshot of the 'height' of the waveform at one particular point in time whereas in DSD that 'height' is built by a stream of 1's and 0's over the previous few samples telling the DAC 'higher' (1's) or 'lower' (0's). So to get a loud sound you need a bunch of 1's adding up to get the signal far up then a bunch of 0's to go far down, etc etc (*)

 

Imagine having to change amplitudes in realtime in the DSD domain (which is what a dsp algo like Dirac needs to do): very hard. There are ways but they are not trivial.

 

Long story short: you must transcode to PCM before using Dirac.

 

(*) Come to think of it, this means that DSD is such that resolution is lower for big signals and higher for small, and the opposite is true in PCM, isn't it???

[YashN]

DSD effectively does not allow for any signal processing.

 

Not if you're Miska...

[miguelito]

Not if you're Miska...

If Miska is doing some DSP in the DSD domain, it's probably in terms of time interpolation when upsampling. But AFAIK DSP like Dirac's requires a much more detailed handling of the data, and very likely can only be done in the PCM domain.

[Kal Rubinson]

That's what I thought. Do you have A+ set to automatically convert DSD to PCM?

 

Yes. I no longer use A+ but I generally have my player convert DSD to PCM for this reason.

[miguelito]

Yes. I no longer use A+ but I generally have my player convert DSD to PCM for this reason.

What do you use these days?

[Fitzcaraldo215]

So now I have two questions: Is it possible to play a DSD file with Dirac on A+? If not will Roon do it?

I think you have gotten most of your answer: Dirac absolutely and without question requires PCM.

You can maintain your library in DSD - ISO, DSF or DFF formats thereof - but playback software needs to convert to PCM on the fly during playback in order to use Dirac. Same goes for Roon. All you need is player software that converts on the fly.

I am very happy with a large DSD library in DSF format with playback via JRiver using its on the fly conversion to 88k PCM. To me, the advantages of Dirac room correction far outweigh any supposed sonic advantages of pure DSD playback. I have compared them several times.

HQPlayer has been mentioned, and I am sure it is a good product in many ways. It purports to be able to do room correction in DSD mode, eliminating the PCM conversion. However, finding no detailed description of these capabilities and owing to the huge technical, mathematical and representational problems inherent in doing EQ in DSD, I question whether it is a capable room EQ package. Even recording engineers have found no way to EQ or filter DSD recordings natively in DSD. They convert to DXD (which is ultra hi rez PCM), apply EQ, then convert back to DSD. If recording pros have not found a way to do it, I seriously doubt HQPlayer can.

[Kal Rubinson]

What do you use these days?

 

jRiver.

[ted_b]

HQPlayer has been mentioned, and I am sure it is a good product in many ways. It purports to be able to do room correction in DSD mode, eliminating the PCM conversion. However, finding no detailed description of these capabilities and owing to the huge technical, mathematical and representational problems inherent in doing EQ in DSD, I question whether it is a capable room EQ package. Even recording engineers have found no way to EQ or filter DSD recordings natively in DSD. They convert to DXD (which is ultra hi rez PCM), apply EQ, then convert back to DSD. If recording pros have not found a way to do it, I seriously doubt HQPlayer can.

 

First off, Jussi has created a few tools that the pros using Pyramix, etc only dream of (we at Native use some of his widgets cuz they are better than what Pyramix or Sonoma have currently, many of which were created 20 years ago). So yes, recording pros use whats out there, and are not bleeding edge by any stretch of the imagination.

 

Second, HQPlayer never touted itself as a room correction software suite. It allows convolution in DSD but is only as good as the impulse filter given it (which is simply how you do convolution in HQPlayer; you load 32 bit float wav impulse filters in any number of channels, then adjust the gain of the convolver to match the filter gain/loss). I do a simple 72hz filter (with a little 39hz bump) that my acoustical engineer Jeff Hedback derived from Room EQ at 24/96. I'm sure the results of Dirac are far better; but if you could boil your Dirac results down to a wav file impulse filter (I clearly know nothing of Dirac), then HQplayer could read it and convolve, in either DSD or PCM. I convolve all the way to DSD256 (takes a little cpu horsepower but nothing insane).

[YashN]

If Miska is doing some DSP in the DSD domain, it's probably in terms of time interpolation when upsampling. But AFAIK DSP like Dirac's requires a much more detailed handling of the data, and very likely can only be done in the PCM domain.

 

Why not ask Miska instead of speculating?

 

I wouldn't let any Dirac or any PCM process touch my DSD, ever.

[YashN]

First off, Jussi has created a few tools that the pros using Pyramix, etc only dream of (we at Native use some of his widgets cuz they are better than what Pyramix or Sonoma have currently, many of which were created 20 years ago). So yes, recording pros use whats out there, and are not bleeding edge by any stretch of the imagination.

 

Second, HQPlayer never touted itself as a room correction software suite. It allows convolution in DSD but is only as good as the impulse filter given it (which is simply how you do convolution in HQPlayer; you load 32 bit float wav impulse filters in any number of channels, then adjust the gain of the convolver to match the filter gain/loss). I do a simple 72hz filter (with a little 39hz bump) that my acoustical engineer Jeff Hedback derived from Room EQ at 24/96. I'm sure the results of Dirac are far better; but if you could boil your Dirac results down to a wav file impulse filter (I clearly know nothing of Dirac), then HQplayer could read it and convolve, in either DSD or PCM. I convolve all the way to DSD256 (takes a little cpu horsepower but nothing insane).

 

It certainly looks like he has a unique value proposition with his DSD-native processing and the recent inclusion of the routing and matrix processing in the betas is a great step forward.

[LBob]

Is there a preferred way of converting via Audirvana+? When I use auto detection I get a 176 output (which still sounds quite nice). Haven't tried the "convert to pcm" or the initial dcs method.