Jump to content
IGNORED

Any DSP that supports SDM & DXD


Recommended Posts

Hi,

 

Any of you know of a DSP software that can be used with a player such as A+ that supports both PCM @384kHz, aka DXD, and SDM?

 

Seems like the most advanced one is DIRAC that supports now up to 192kHz...

I think you can use JRiver to convert them to PCM, but I haven't tried it personally, so no promises.

DSP happens using PCM, so it will need to be converted at some point in the chain.

Link to comment
I think you can use JRiver to convert them to PCM, but I haven't tried it personally, so no promises.

DSP happens using PCM, so it will need to be converted at some point in the chain.

 

Yes, that's my understanding also... as well as there are other players that can do the same including the free Foobar 2000 with its SACD plugin:

 

SACDpluginFoobar.jpg

 

Ciao, Flavio

Warning: My posts may be biased even if in good faith, I work for Dirac Research :-)

Link to comment

Hi,

 

tkx for your reply.... already was aware of that.

 

What I'm actually looking for is a DSP room correction software that could work, natively without any transcoding on the fly from SDM to PCM (or worse even by downsampling the sample rate to 192, that seems to be the highest sample rate available for DSM room correction software these days).

Link to comment
Hi,

 

tkx for your reply.... already was aware of that.

 

What I'm actually looking for is a DSP room correction software that could work, natively without any transcoding on the fly from SDM to PCM (or worse even by downsampling the sample rate to 192, that seems to be the highest sample rate available for DSM room correction software these days).

 

That would be bad. For technical reasons, native SDM is quite lossy for doing any DSP aside from delay.

 

The best option is to convert it to PCM, and do as much processing as you want in hi-res PCM, without much loss.

 

You can forgo doing DSP, and send the SDM straight to the DAC without conversion. That will achieve SDM purity, but I definitely consider that to be less desirable than goodies like room correction, bass management, EQ, active crossovers, etc. Improvements from stuff like that far outweigh the tiny loss in the SDM->hi-res PCM conversion, IMO.

Link to comment
What I'm actually looking for is a DSP room correction software that could work, natively without any transcoding on the fly from SDM to PCM (or worse even by downsampling the sample rate to 192, that seems to be the highest sample rate available for DSM room correction software these days).

 

When you say "SDM" I think you are referring to DSD files like .dff and .dsf? If so, it is mathematically impossible to apply DSP to a one bit file. It must be converted to some kind of multi-bit file. I think the closest thing might be the way Jussi does it in HQplayer. He claims to be able to apply DSP natively to DSD files. I think he uses some form of SDM to apply the DSP.

 

Seems like the most advanced one is DIRAC that supports now up to 192kHz...

 

I have a lot of experience with both DIRAC and Acourate. I have to disagree with this statement. Acourate is more sophisticated, IMO. I believe the filters are better and much more transparent.

THINK OUTSIDE THE BOX

Link to comment
When you say "SDM" I think you are referring to DSD files like .dff and .dsf? If so, it is mathematically impossible to apply DSP to a one bit file. It must be converted to some kind of multi-bit file. I think the closest thing might be the way Jussi does it in HQplayer. He claims to be able to apply DSP natively to DSD files. I think he uses some form of SDM to apply the DSP.

I interpret SDM to be any Sigma Delta Modulation variant, like DSD.

 

I have read about exotic ways to do DSP on SDM natively, but they are really cumbersome and have lots of artifacts. It's a cure that's worse than the disease.

 

The purist dream is to take the SDM signal straight from an ADC and feed it straight to a DAC, and avoid conversion to and from PCM, and associated loss.

1) purist dream of SDM everything:

DAC -> SDM -> deliver content as SDM -> DAC

 

Reality is far from that dream, since you need DSP to create content.

 

The real options are:

 

2) conventional PCM creation and delivery:

DAC -> SDM -> PCM -> DSP -> deliver content as PCM -> DSP -> DAC

 

2) real SDM creation and delivery:

DAC -> SDM -> PCM -> DSP -> SDM -> deliver content as SDM -> PCM -> DSP ->DAC

 

3) real SDM creation and delivery, without DSP on playback (ouch):

DAC -> SDM -> PCM -> DSP -> SDM -> deliver content as SDM -> DAC

 

So, once you assume DSP as part of creation (which it is), SDM isn't so purist after all.

PCM delivery would actually do fewer conversions, with better technical quality.

That said, DSD is the only way you can find a lot of hi-res content.

Forgoing any DSP to avoid the conversion back to PCM seems silly when you consider that it has already been converted back and forth during creation.

 

24/96 is sufficient for SDM conversion back to hires PCM, and 24/192 is overkill, if you're worried about getting every last drop.

Link to comment
Hi,

 

Any of you know of a DSP software that can be used with a player such as A+ that supports both PCM @384kHz, aka DXD, and SDM?

 

Seems like the most advanced one is DIRAC that supports now up to 192kHz...

For the above request the best option without interim PCM conversion, as dallasjustice suggested, is to use the Win-only application Acourate (or the open source DRC) to generate a filter @4x and then use such filter with HQPlayer for Mac for both PCM > DSP > DAC and SDM > DSP > DAC. HQPlayer has two separate DSP engines and uniquely it always mantains the input (PCM or SDM) in its original format. Bare in mind that to apply DSP, like room correction, to SDM signal you need a medium power computer.

Link to comment
Realtime PCM-to-DSD conversion does not seem to be that good. Roundtrip will probably be even worse.

 

Archimago's Musings: MEASUREMENTS: PCM to DSD Upsampling Effects (JRiver MC19 Beta).

Personally I do not take much notice of Archimago's measurements. However, the algorythm used by JRiver for PCM-to-DSD conversion is not that good and it measures poorly even compared to Foobar's AsioProxy, as demonstrated by AMR/IFI Thorsten Loesch.

Link to comment
Realtime PCM-to-DSD conversion does not seem to be that good. Roundtrip will probably be even worse.

 

Archimago's Musings: MEASUREMENTS: PCM to DSD Upsampling Effects (JRiver MC19 Beta).

 

Beg, borrow or steal a DSD 128 or better still a DSD256 DAC, try out HQ Player 3.4.1 PCM to DSD conversion with Polysic Filter and DSD7 settings, and then tell us what you think about the sound quality of PCM converted to DSD128/256

Sound Test, Monaco

Consultant to Sound Galleries Monaco, and Taiko Audio Holland

e-mail [email protected]

Link to comment

Any conversion PCM to PCM or PCM to DSD or DSD to PCM depend on precision of used algorithms.

 

If say roughly for playback DSD almost equal 1-bit PCM with super high sampling rate.

 

Allowing of hard algorithms depend on performance of hardware.

 

For conversion first criterias is noise and ringing artefacts.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment
When you say "SDM" I think you are referring to DSD files like .dff and .dsf? If so, it is mathematically impossible to apply DSP to a one bit file. It must be converted to some kind of multi-bit file.

 

DSP with delta sigma modulated 1-bit stream possible without conversion to multibit PCM. As I said before 1-bit DSD almost 1-bit PCM.

 

Before any DSP I, as example, input both streams (1-bit and PCM) as 64-bit floating point (double precission in C++ programming language). Noise floor of this format about -200 dB.

 

Further we can do DSP and convert it to multibit PCM or 1-bit sigma delta modulation.

 

If we right convert integer PCM -> 64-bit float wo DSP -> PCM, we get input and output PCM bit-to-bit.

 

For DSD -> 64-bit float wo DSP -> DSD same result.

 

If we do any DSP we loss and add info anyway. Mathematically we input 1-bit stream as float point stream, do any DSP with it.

 

For delta sigma modulation: signal pass by through delta sigma modulator. During it appear noise of quantization. Be we shift most energy of the noise out of audible range (20 ... 20 000 Hz).

 

But "useful" signal we get at output delta sigma modulator almost as is.

 

Thus we get 1-bit PCM with shifted noise.

 

Unlike PCM to analog conversion we must filter this shifted noise (all above 20 ... 30 kHz).

 

If you use qualitative "hard" algorithms of conversion and DSP it is not most "narrow place" in whole audio system :)

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment
Beg, borrow or steal a DSD 128 or better still a DSD256 DAC, try out HQ Player 3.4.1 PCM to DSD conversion with Polysic Filter and DSD7 settings, and then tell us what you think about the sound quality of PCM converted to DSD128/256

 

I already have a DAC that supports DSD128, yet, there is no music on DSD that I like (not interested in audiophile no-name bands/etc). I would not touch that HQPlayer with a 10 foot pole. If the developer could not even bring it to a decent user interface, who knows whats going on behind the scenes there.

Link to comment
I would not touch that HQPlayer with a 10 foot pole. If the developer could not even bring it to a decent user interface, who knows whats going on behind the scenes there.

 

Could you elaborate a bit more what you mean by a decent user interface and why you think HQPlayer doesn't have one?

 

I'm a software engineer, not a graphics designer. My priorities are on audio quality and not on shiny looks...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

I have a hard time accepting that DSD is inherently better than PCM. There's just the sad situation that some hires content has been converted to DSD as part of SACD distribution. Given that, something has to perform aggressive low-pass filtering on it at some point. That could happen in software or in a DSD DAC.

 

I assume that anything that does FIR DSD "directly on DSD data" must be converting 1-bit DSD to float DSD, doing a CPU-intensive convolution on high sample rate float data, then re-dithering back to 1-bit. At that point, the DAC would still need to perform aggressive low-pass filtering to convert to analog without high frequency noise.

 

Why is this any better than pre-filtering the data to PCM prior to the FIR convolution at a reasonable sample rate (like 24/96)? I assume most modern DACs do a multi-bit digital to analog conversion and/or do a higher than DSD sample rate, in order to avoid aggressive low pass filtering. With that in mind, this doesn't seem like it would be lower quality. It's certainly much more practical. Also, it avoids redithering back to 1xDSD.

 

Of course, any differences in this kind of conversion path could create very subtle differences in sound. My personal experience is that the flexibility of fine-tuning DSP EQ dwarfs the subtle differences in modern DAC filtering. If you like a slightly smoother treble rolloff, just shape the EQ that way.

 

Anyway, this is why I think doing nice quality conversion (whether online or offline) from DSD to PCM is the most practical approach. Most of my content is PCM, anyway, and my playback software and DSP is optimized around that.

Link to comment
Given that, something has to perform aggressive low-pass filtering on it at some point.

 

PCM needs even more aggressive low-pass filtering to remove all the images... (correct reconstruction)

 

Why is this any better than pre-filtering the data to PCM prior to the FIR convolution at a reasonable sample rate (like 24/96)? I assume most modern DACs do a multi-bit digital to analog conversion and/or do a higher than DSD sample rate, in order to avoid aggressive low pass filtering. With that in mind, this doesn't seem like it would be lower quality. It's certainly much more practical. Also, it avoids redithering back to 1xDSD.

 

So you would like to go to 96 kHz sampling rate, do the convolution, then let the DAC oversample back to 384 kHz and from there on using the usual poor methods to 5.6 MHz and then convert to typical 2.5-bit SDM? And then you need to do the same noise filtering...

 

While doing the convolution in software you could also choose to upsample to higher rate DSD, for example 24.576 MHz... And never go down in sampling rate.

 

Anyway, this is why I think doing nice quality conversion (whether online or offline) from DSD to PCM is the most practical approach. Most of my content is PCM, anyway, and my playback software and DSP is optimized around that.

 

My playback software and DSP is optimized for both, because I wanted no-compromises solution for both. For myself.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

In my opinion the debate about PCM vs. DSD can go on forever but both produce excellent results and differences that show in measurements are not necessarily meaningful in listening tests.

 

A now rather old (ten years) paper presented to the AES and comparing PCM vs. DSD in listening tests found a draw and summarized that as follows:

"These listening tests indicate that as a rule, no significant differences could be heard between DSD and high-resolution PCM (24-bit /176.4 kHz) even with the best equipment, under optimal listening conditions, and with test subjects who had varied listening experience and various ways of focusing on what they hear.

Consequently it could be proposed that neither of these systems has a scientific basis for claiming audible superiority over the other.

This reality should put a halt to the disputation being carried on by the various PR departments concerned"

 

http://old.hfm-detmold.de/eti/projekte/diplomarbeiten/dsdvspcm/aes_paper_6086.pdf

 

Now as mentioned before that paper is ten years old... would someone point me to more recent double blind reliable ABX listening tests that back the superiority of one or the other?

Miska's solution of optimizing for both is an optimal one but if I had to choose one or the other I would privilege the one which has the wider choice of music material in case reliable blind listening tests do not show a meaningful difference in listening correlated to measurements.

 

Again... in my opinion, which does not necessarily represent Dirac's one

Flavio

Warning: My posts may be biased even if in good faith, I work for Dirac Research :-)

Link to comment

That's consistent with what I have read, that 24/176-24/192 can't be discriminated between DSD, and that 24/96 was very slightly marginal and could comfortably be considered good enough at a much lower cost.

 

Doubling the sampling rate means that equivalent length FIR filters require 4x the processing power. Rendering a computer useless for doing anything else or using shorter FIR filters has more meaningful consequences than something I can't hear.

 

At an equal sampling rate, 1-bit requires heavier filtering than multi-bit.

If you have source music at 2.8mhz 1-bit, that determines the most aggressive filtering that will need to be used (whether analog or digital filtering) - no way around that.

I can see your point, though, that secondary filtering can be controlled, and 25mhz is higher than 96khz.

But, while there may be theoretical advantages, 25mhz has no audible advantage over 24/176-192.

On the other hand, it is much more CPU intensive, which does create real limitations in how much processing you can do, as well as limiting which software you can use.

I would argue that makes the solution worse, not better.

 

It would certainly be neat to see more content become available in hi-res, and as a hi-res format, SACD has gotten a bit of traction, which is better than nothing.

 

Personally, I'm more excited about DTS-HD, TrueHD, and Atmos. :)

Link to comment

http://old.hfm-detmold.de/eti/projekte/diplomarbeiten/dsdvspcm/aes_paper_6086.pdf

 

Now as mentioned before that paper is ten years old... would someone point me to more recent double blind reliable ABX listening tests that back the superiority of one or the other?

I've seen a couple others, but they came to the same conclusion.

I saw one that compared DSD to 24/96, and found that it could only be discriminated something like 2% of the time, but 24/192 could never be discriminated.

 

BTW, this was dithered 24/96 and dithered 24/192. I think there was a bit higher margin for non-dithered multi-bit.

Link to comment

If you have source music at 2.8mhz 1-bit, that determines the most aggressive filtering that will need to be used (whether analog or digital filtering) - no way around that.

 

Most agressive filtering for last (SRC or PCM to DSD) stage is not most distorting link in full audio chain. Almost any effect/DSP used for music production give more distortion than the last stage. I says about pro algorithms (last stage) of course.

 

For agressive filtration used optimized algorithms.

 

What optimized:

 

brick wall - minimal riging

 

If used minimal phase filter:

 

brick wall - no-pre-ringing - minimal post-riging - minimal phase distortions

 

For online (on fly, real-time) algorithms need optimize performance also.

 

All these things is compromis, but is not most narrow link in whole system.

AuI ConverteR 48x44 - HD audio converter/optimizer for DAC of high resolution files

ISO, DSF, DFF (1-bit/D64/128/256/512/1024), wav, flac, aiff, alac,  safe CD ripper to PCM/DSF,

Seamless Album Conversion, AIFF, WAV, FLAC, DSF metadata editor, Mac & Windows
Offline conversion save energy and nature

Link to comment

From my perspective, high fidelity audio includes DSP for both mastering and playback (regardless if whether it's considered "distorting"). On the playback side, that may include things like EQ, bass management, and active crossovers. Aside from most source coming in multi-bit, if you assume that significant DSP needs to take place, things should be optimized towards that.

 

Of course, if Miska prefers to be more purist, that's fine, and it's great that there's a DSP product out there for him.

 

Personally, though, I want to pretend that DSD doesn't exist, aside from a weird lossy compression format for hi-res distribution, and converting it to 24/96 or 24/172-192 is a convenient way to "decompress" it and get on with the normal signal chain and my favorite playback and DSP software.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...