Audirvana Plus 3 for Windows (Official Thread)

[bodiebill]

I started using the mac version many years ago, and DSD64 => DSD 256 worked from the start.

 

@damien78

Maybe you can shed some light?

Main question is: can A+ for Windows upsample DSD64 to DSD256 and if yes, how?

[Jud]

By the way, I will gently disagree with whether upsampling DSD is essential, since conversion to PCM is necessary as an intermediate step. I would tend to purchase PCM versions (if the same mastering), since with DSD to DSD conversion the final conversion will be from the intermediate PCM anyway.

[Jud]

1 minute ago, bodiebill said:

 

I started using the mac version many years ago, and DSD64 => DSD 256 worked from the start.

 

 

I am virtually certain this was not the case years ago (DSD to DSD conversion on the Mac). I can always be wrong of course, but I don't think so....

[bodiebill]

You maybe right about that.

 

I have no preference for either DSD or PCM but I do know that my Lampizator Lite 7's DSD chain is way better than its PCM chain. And DSD64 upsampled (by either Audirvana mac or HQPlayer) to DSD256 sounds considerably better than the original DSD64.

 

On the other hand, PCM sounds absolutely gorgeous with my Audio Aéro DAC. It can compete with Lampizator DSD, but with a different sound signature (I call it 'creamy' but not in a bad way).

 

I was not aware that A+ (and HQPlayer?) upsample DSD => PCM => DSD.

[Jud]

51 minutes ago, bodiebill said:

I was not aware that A+ (and HQPlayer?) upsample DSD => PCM => DSD.

 

It's not possible to upsample 1-bit format (DSD) directly.

[firedog]

3 hours ago, Jud said:

 

It's not possible to upsample 1-bit format (DSD) directly.

You can get around that by converting it to multibit DSD.

[jamesg11]

As with what Miska does in HQP?

[Jud]

38 minutes ago, jamesg11 said:

As with what Miska does in HQP?

 

Convert to PCM.

[Jud]

1 hour ago, firedog said:

You can get around that by converting it to multibit DSD.

 

DSD, strictly defined, is 1-bit.  It's PDM or PWM, not PCM.  Anything with more than 1 bit has a PCM component.

[Miska]

Multi-level SDM is not the same as PCM. (And if you want, you can also upsample 1-bit DSD as is.)

 

Note that multi-level SDM can typically have for example 3 or 5 levels that doesn't match with any number of two's complement PCM bits. For example modulators in TI DAC chips have 5-level output, that is log2(5) = 2.3219 two's complement bits. Advantage of  multi-level SDM compared to PCM is that for all values except 0 and maximum, multi-level SDM has more than one representation for the same value. This is one property that multi-level SDM DAC chips extensively utilize. Same feature allows also good quality bit-perfect DSD conversion when used correctly.

 

6 hours ago, jamesg11 said:

As with what Miska does in HQP?

 

HQPlayer has two modulator variants, one that eats PCM inputs and another that eats SDM inputs (2 - 256 level). To be exact, there are two totally separate DSP engines, one for PCM outputs and another one for SDM outputs since these work in a different way. The SDM input one can remodulate directly from one SDM rate and number of levels to another SDM rate and number of levels. So for example rate is converted directly from 2.8 MHz to 11.3 MHz, without any intermediate rate. The modulator itself has gain property which allows adjusting volume as part of the process. You can also bake other kind of things into the modulator.

 

Also things like convolution are processed at the native rate, so for example if you do convolution for DSD256, it is processed at 11.3 MHz sampling rate.

 

5 hours ago, Jud said:

DSD, strictly defined, is 1-bit.  It's PDM or PWM, not PCM.  Anything with more than 1 bit has a PCM component.

 

So you define 2-level SDM very strictly and PCM very loosely.  If you are that loose, you can also say that DSD is just 1-bit PCM. Much simpler to call everything PCM! It would be better to be equally strict on both.

 

Equally strictly defined, PCM is two's complement (binary encoded) integer representation. So either DSD is also PCM, or HQPlayer doesn't convert to PCM.

 

 

[firedog]

5 hours ago, Jud said:

 

DSD, strictly defined, is 1-bit.  It's PDM or PWM, not PCM.  Anything with more than 1 bit has a PCM component.

And many in the field would say that's irrelevant. DSD is more of a marketing term of Sony's, than a relevant technical term. See Miska's post.

[bodiebill]

Interesting stuff! Especially the view that DSD can be seen as a 1-bit instance of PCM. Maybe that could make the DSD vs PCM discussion less black and white. After all, "what's in a name?". In the meantime, with much respect for the technical artistry to make this all work, I continue to trust my ears... 🙂

[Jud]

5 hours ago, Miska said:

Multi-level SDM is not the same as PCM. (And if you want, you can also upsample 1-bit DSD as is.)

 

Note that multi-level SDM can typically have for example 3 or 5 levels that doesn't match with any number of two's complement PCM bits. For example modulators in TI DAC chips have 5-level output, that is log2(5) = 2.3219 two's complement bits. Advantage of  multi-level SDM compared to PCM is that for all values except 0 and maximum, multi-level SDM has more than one representation for the same value. This is one property that multi-level SDM DAC chips extensively utilize. Same feature allows also good quality bit-perfect DSD conversion when used correctly.

 

 

HQPlayer has two modulator variants, one that eats PCM inputs and another that eats SDM inputs (2 - 256 level). To be exact, there are two totally separate DSP engines, one for PCM outputs and another one for SDM outputs since these work in a different way. The SDM input one can remodulate directly from one SDM rate and number of levels to another SDM rate and number of levels. So for example rate is converted directly from 2.8 MHz to 11.3 MHz, without any intermediate rate. The modulator itself has gain property which allows adjusting volume as part of the process. You can also bake other kind of things into the modulator.

 

Also things like convolution are processed at the native rate, so for example if you do convolution for DSD256, it is processed at 11.3 MHz sampling rate.

 

 

So you define 2-level SDM very strictly and PCM very loosely.  If you are that loose, you can also say that DSD is just 1-bit PCM. Much simpler to call everything PCM! It would be better to be equally strict on both.

 

Equally strictly defined, PCM is two's complement (binary encoded) integer representation. So either DSD is also PCM, or HQPlayer doesn't convert to PCM.

 

I thought I recalled something different from a conversation between you and mansr here on the forums, but I could be misremembering.

 

My notion of DSD is that it's 1-bit, with values obtained through pulse width/density (more 1s for higher values, more 0s for lower values), like adjustments to a steering wheel, whereas PCM "words" have assigned values like points on a graph or map, and bits within the "word" have different values depending on position, so 010 and 100 are different.

[Jud]

In any event, we still have the question whether A+ upsamples DSD input.

[Miska]

41 minutes ago, Jud said:

My notion of DSD is that it's 1-bit, with values obtained through pulse width/density (more 1s for higher values, more 0s for lower values), like adjustments to a steering wheel, whereas PCM "words" have assigned values like points on a graph or map, and bits within the "word" have different values depending on position, so 010 and 100 are different.

 

There can be smaller and larger pulses. Good example being DSC-1 DAC where after each bit the conversion output can have one of the 33 possible values. Yet it is totally bit-perfect DSD converter.

 

And you can have for example 1-bit word with values 0 and 1. And you can have 32-bit word with values 0 and 1. Both have same meaning. Or you can have letter 'a' and 'b' instead. So it won't make any difference. So in your generalization, then also DSD would be PCM.

 

41 minutes ago, Jud said:

bits within the "word" have different values depending on position, so 010 and 100 are different

 

That is not the case for multi-level SDM. You can also interchangeably transform between the two representations, to limited extent.

 

41 minutes ago, Jud said:

I thought I recalled something different from a conversation between you and mansr here on the forums, but I could be misremembering.

 

Can you give a pointer?

[Jud]

10 minutes ago, Miska said:

Can you give a pointer?

 

Wish I could! I would have to search, and if I am misremembering, I wouldn't find anything.   

 

The Wikipedia article gives a nice outline of the picture I have in my head (in the "DSD technique" section), as well as saying things are really more complex than that, especially these days.  https://en.wikipedia.org/wiki/Direct_Stream_Digital

 

[Miska]

6 minutes ago, Jud said:

 

Wish I could! I would have to search, and if I am misremembering, I wouldn't find anything.   

 

The Wikipedia article gives a nice outline of the picture I have in my head (in the "DSD technique" section), as well as saying things are really more complex than that, especially these days.  https://en.wikipedia.org/wiki/Direct_Stream_Digital

 

 

That is quite a bit of simplification. And I don't think it makes sense to attempt cover all the details in Wikipedia or forum post. There are books and research papers about SDM stuff, because it is so widely on A/D and D/A converters these days. Lot of companies don't tell much details how their chips work either.

 

[bodiebill]

53 minutes ago, Jud said:

In any event, we still have the question whether A+ upsamples DSD input.

 

+1 🙂

[tims]

I'm trying out the trial A+ for Windows but can't find a comprehensive manual like the 48 page A+ 3 for MAC manual on the Audirvana website, but only a A+ for Windows installation manual on the website.

For the A+ Windows, is there a comprehensive manual coming out or already out like the full MAC manual mentioned above?

 

Thanks

 

[Sam Ash]

Hi guys, I'm a hard-core Apple user, have always been but I do the accept the fact that Windows 10 has come a long way and there is hardly any clear contrast between the 2 platforms. In a way, I'm glad that there is a Window version of A+ 3 for windows because Windows machines have always been more cost effective when compared to Apple hardware. To that effect, is there a Mac Mini alternative that is cost effective, just as powerful (or more) and very reasonably priced ? A mini-PC machine that would be ideal to run A+3 without any problems ?

[Jud]

It depends. 🙂 

 

Were you thinking of something directly connected, or running A+ remotely and streaming?

 

If directly connected, I wonder whether a powerful miniPC that for example will upsample to DSD might, like a laptop, have the fan get noisy inside its small case.  If using it remotely for streaming, sure, no reason you can't find a powerful, relatively inexpensive machine.

[Sam Ash]

13 hours ago, Jud said:

It depends. 🙂 

 

Were you thinking of something directly connected, or running A+ remotely and streaming?

 

If directly connected, I wonder whether a powerful miniPC that for example will upsample to DSD might, like a laptop, have the fan get noisy inside its small case.  If using it remotely for streaming, sure, no reason you can't find a powerful, relatively inexpensive machine.

 

I suppose direct connection would give me reliability, are there not mini units designed to keep cool without fans? Having said that, I have a few questions:-

 

1. How would I connect a mini-PC to to an external dedicated 2-channel power amp ?

2. When remotely streaming via upnp/dlna, is there any loss due to comms compression ?

 

[Jud]

4 hours ago, Sam Ash said:

 

I suppose direct connection would give me reliability, are there not mini units designed to keep cool without fans? Having said that, I have a few questions:-

 

My guess is that the set of mini-PCs that are fanless does not have a great deal of overlap with the set of mini-PCs that will perform upsampling to DSD without strain, and even less overlap with the set of mini-PCs with that capability which are reasonably inexpensive.  Of course that doesn't prevent you from hunting for and possibly finding an exception which fits all three criteria.

 

4 hours ago, Sam Ash said:

1. How would I connect a mini-PC to to an external dedicated 2-channel power amp ?

 

mini-PC to DAC to amp.

 

4 hours ago, Sam Ash said:

2. When remotely streaming via upnp/dlna, is there any loss due to comms compression ?

 

 

No.  I subjectively felt the sound was better, and that has been the experience of others I know who've tried streaming.  Since there aren't any measurements I'm aware of saying streaming results in worse sound quality, you can believe the (absence of) measurements, or of course you are free to listen for yourself.

[Sam Ash]

On 1/28/2019 at 7:56 PM, Jud said:

 

My guess is that the set of mini-PCs that are fanless does not have a great deal of overlap with the set of mini-PCs that will perform upsampling to DSD without strain, and even less overlap with the set of mini-PCs with that capability which are reasonably inexpensive.  Of course that doesn't prevent you from hunting for and possibly finding an exception which fits all three criteria.

 

 

mini-PC to DAC to amp.

 

 

No.  I subjectively felt the sound was better, and that has been the experience of others I know who've tried streaming.  Since there aren't any measurements I'm aware of saying streaming results in worse sound quality, you can believe the (absence of) measurements, or of course you are free to listen for yourself.

 

Thank you very much for that information Jud, you've been a great help. I know this is an audiophile forum so no disrespect to fellow members but I've been playing some of my 320Kbps mp3 files and they sound fantastic too via A+. I know there is support for Qobuz and TIDAL streaming in A+ which is great but I wish the developer incorporated Spotify too as an option.

[Sam Ash]

I guess I could give TIDAL a go and might end up loving it.