PC+SW+DSD direct VS Any DAC

[widemediaphotography]

Rather than argue about your definition of efficiency, the real question is what do you gain with all this extra processing? Where’s the evidence that this is really needed for listening to audio?

Could you read again my post #1 and #35, please?

 

[pkane]

Could you read again my post #1 and #35, please?

Can you read my posts here, on ASR, to see if maybe I do know something about DSD conversion and filtering?

You asked if DSD conversion is objectively better than simple PCM. It’s not. Not in most cases, and any minor improvements claimed by the HQP author are well below any known audibility levels.

Certainly some DACs can be broken or have very poorly designed filters. These are the exceptions, not the rule, and while an off-board filter might help in those cases, I’d recommend getting a properly designed DAC rather than go down the rabbit hole of converting everything to DSD.

 

[MaxwellsEq]

Use the word you think is most appropriate instead of "efficient", but the PC and the software do the job for us and you still haven't understood that if it is true that we input the audio flow into the same DAC, but this works in a completely different way, because it is specifically designed to accept DSD files without operating digital filtering on them. You don't have the same result at all. This is what we are discussing, nothing else!

I really have understood. You continue to demonstrate that you have an opinion.

Fine, but every effort you make to demonstrate a factual basis behind your opinion, falls short of a scientific or engineering proof. You must soon begin to notice that you have been wrong, then wrong, then wrong again.

 

[MaxwellsEq]

Hi @widemediaphotography please can you do some background reading on the work @pkane has done. You need to realize you are debating with some brilliant experts, and are not delivering solid, well-reasoned engineering proofs for your opinion.

 

[NTK]

Everyone can verify, for example, that it is simple to eliminate ringing present in PCM streams that was not present before master acquisition, this is an improvement, not an artifact!

The "ringing" you refer to is exactly what the digitized samples represent when converted back to continuous time signal, per the Nyquist–Shannon sampling theorem. Have you asked where those digitized samples that give you the "ringing" came from?

Hint: They are synthesized samples where their equivalent band limited continuous time version is not shown to you.
Additional hint: See if you can find some properly A/D converted samples that show those "ringing", when the "ringing" is absent from the original continuous time analog signal.

 

[Verig]

@widemediaphotography DSD was marketing to begin with. And now HQP crowd is taking this format, converting absolutely fine PCM files and feeding the bloated files into obscure dacs.
When the premise is wrong the whole chain becomes weirder and weirder. And when you start to feel there is no benefit... perhaps you should sample a bit more, why not even 1024DSD, yes that'll do it for sure, I mean, it can't be that this is just nonsense, you just need more...

The number of filters in dacs is marketing too. They should select the best and give just that. But people seem to want these things so they are given. There are no good reasons for it. When time passes people start to think there sometime was a good reason - why would they do if there was none.

Just move your speakers 2cm and hang a painting onto your wall, or buy thicker carpet and curtains, that will give you audible benefit, not tinkering with software conversion which has no basis to make any improvement.

 

[widemediaphotography]

You asked if DSD conversion is objectively better than simple PCM. It’s not. Not in most cases, and any minor improvements claimed by the HQP author are well below any known audibility levels.

Certainly some DACs can be broken or have very poorly designed filters. These are the exceptions, not the rule, and while an off-board filter might help in those cases, I’d recommend getting a properly designed DAC rather than go down the rabbit hole of converting everything to DSD.

This is the first point of convergence in the discussion that I encounter and which I appreciate very much. I don't believe at all that converting a PCM to DSD, as an end in itself, constitutes an improvement, never I wrote that. The apparent increase in resolution would be due to the introduction of artifacts. I would never execute of upsampling my collection of digital photographs and preserving them, given the complete uselessness of the operation, it just cause huge introducing of artifacts withount indreasing the quality of the images.

When you say that the performance increase made by HQplayer does not enter the audible field, it represents an important point to make.
Contrary to what people might think reading these 3 pages, I am by no means an admirer of HQplayer or a detractor, and I have never claimed that its application brings benefits (posts #1 and #35), but I had a lot of trouble to make it clear that sending a DSD to a DSD-direct DAC is very different from sending it to a common DAC and that the conversion to DSD only serves to bypass the filtering and modulation stage present in any D-S DAC that in each case done trough oversamplig operations.
Therefore, if you can share measurements or references that can confirm that what HQplayer would do is just collective delirium or a flimsy attempt, I would be very grateful.

 

[Verig]

but I had a lot of trouble to make it clear that sending a DSD to a DSD-direct DAC is very different from sending it to a common DAC and that the conversion to DSD only serves to bypass the filtering and modulation stage present in any D-S DAC that in each case done trough oversamplig operations.

I noticed. And I guess I could have been more eloquent in the last post.
My problem with HQP is that it's based on false assumptions about Delta-Sigma dacs. I can't really trust anything based on misunderstandings. Having a goal of bypassing D-S dac filtering makes an assumption that DSD is better and even more, it assumes that simply a DSD dac is better no matter what the recording was to begin with. This does not really make any sense. I know it sounds good and kind of logical that "no filter" is better but it's not logically or technically right.

 

[widemediaphotography]

I noticed. And I guess I could have been more eloquent in the last post.
My problem with HQP is that it's based on false assumptions about Delta-Sigma dacs. I can't really trust anything based on misunderstandings. Having a goal of bypassing D-S dac filtering makes an assumption that DSD is better and even more, it assumes that simply a DSD dac is better no matter what the recording was to begin with. This does not really make any sense. I know it sounds good and kind of logical that "no filter" is better but it's not logically or technically right.

Yours is a shared thought, I too, when trying HQplayer, realized that it's not all roses and that there are many things to try and above all that the conditions vary from DAC to DAC.
The theory behind HQplayer and DSD-direct seems correct, but I intend to find out how it was implemented and how it actually works. Never more than in this case should we not trust our ears, because when we talk about filters, DSP, an artifact can easily be mistaken for improvements

 

[Verig]

Audiophile Myth #260: The Detestable Digital Filter Ringing and Real Music...

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Here's a nice collection of history with extensive links. If you take a look at the steps through years do all those special filters and processing power needed... sound familiar?
There's a good bit about ringing too but I find the history more interesting.

 

[pkane]

I had a lot of trouble to make it clear that sending a DSD to a DSD-direct DAC is very different from sending it to a common DAC and that the conversion to DSD only serves to bypass the filtering and modulation stage present in any D-S DAC that in each case done trough oversamplig operations.

Right, and that's true. I even have a few of these "direct-DSD" DACs. Here is some reading that may help you:

https://www.audiosciencereview.com/forum/index.php?threads/pcm-vs-dsd.41092/post-1810049
https://www.audiosciencereview.com/forum/index.php?threads/hqplayer-do-i-need-it.13720/post-1363600
https://www.audiosciencereview.com/...sic-analysis-video.31620/page-56#post-1138889

HQP author's comparison, showing same distortions but higher noise floor for DSD:

Therefore, if you can share measurements or references that can confirm that what HQplayer would do is just collective delirium or a flimsy attempt, I would be very grateful.

There were plenty more discussions here about DSD vs PCM, including those with Jussi (Miska) the HQPlayer author. Just do some digging.

Note that I consider HQPlayer to be well-written and powerful software with many good features well beyond PCM to DSD conversion, but I do find it strange that Jussi focuses so much (marketing) energy in trying to prove that PCM converted to DSD is somehow better.

 

[Keith_W]

Therefore, if you can share measurements or references that can confirm that what HQplayer would do is just collective delirium or a flimsy attempt, I would be very grateful.

I seem to have misplaced my measurements, and I can't seem to find the old ASR thread where I posted these measurements. I have a Merging DAC that can decode anything from PCM 44 to DSD256. When it was measured with PCM44, the SINAD is -111dB. With DSD256, it is -118dB.

So yes, there is an improvement. But it's really a "so what" improvement. If HQPlayer didn't cost so much money and I didn't have to pay such a heavy penalty in CPU power to run it, of course I would grab that improvement even though I know it's not audible. But let's be honest here, it would be a fetish for nicer numbers rather than chasing an audible improvement.

 

[Snoopy]

I noticed. And I guess I could have been more eloquent in the last post.
My problem with HQP is that it's based on false assumptions about Delta-Sigma dacs. I can't really trust anything based on misunderstandings. Having a goal of bypassing D-S dac filtering makes an assumption that DSD is better and even more, it assumes that simply a DSD dac is better no matter what the recording was to begin with. This does not really make any sense. I know it sounds good and kind of logical that "no filter" is better but it's not logically or technically right.

HQplayer uses different delta sigma modulators that offer a different result. For that you need a DSD direct that. Otherwise you are still going through the dacs SDM .. it’s not that wild of a claim.

That’s the signal that ends up in the analog output stage. If the improvements are audible, better , or worth the costs (software and a computer To run it ) is easy to find out.

Try the software, it works for 20minutes in trial mode, after that you need to restart it.


I stopped using it because usually my MacBook m1 idles with roon at a few % cpu usage , and with HQplayer to get DSD256, it’s like running a benchmark the whole time

 

[Blumlein 88]

HQplayer uses different delta sigma modulators that offer a different result. For that you need a DSD direct that. Otherwise you are still going through the dacs SDM .. it’s not that wild of a claim.

That’s the signal that ends up in the analog output stage. If the improvements are audible, better , or worth the costs (software and a computer To run it ) is easy to find out.

Try the software, it works for 20minutes in trial mode, after that you need to restart it.


I stopped using it because usually my MacBook m1 idles with roon at a few % cpu usage , and with HQplayer to get DSD256, it’s like running a benchmark the whole time

It gobbles up the M1 CPU because it is efficient. /sarc

 

[phofman]

that sending a DSD to a DSD-direct DAC is very different from sending it to a common DAC and that the conversion to DSD only serves to bypass the filtering and modulation stage present in any D-S DAC that in each case done trough oversamplig operations.

I am no specialist in DSD, but this seems quite similar to what the PCM processing in the DAC does in case of a one-bit SDM:

https://www.audiopraise.com/services/fpga-cores/pcm-to-dsd-conversion/:

A DSD stream is usually produced in two steps. The first one is up-sampling of the PCM signal to a sufficiently high sampling rate. The second stage is a noise-shaper with a 1-bit quantizer, which actually converts multi-bit PCM into 1-bit stream.

So IIUC basically you move the PCM processing stage from the DAC to the PC and transfer the resulting high-speed bitstream instead.

 

[gwing]

I certainly won't be the one to establish the meaning of efficiency, but among the various ways of achieving a result, I would define as more efficient the one that does it more precisely, the one closest to reality, the one most quickly, etc. etc. There are many parameters that contribute to greater efficiency

I'm not asking for a definition of efficiency but just asking for clarification of what you mean by it in this context.

You say "That a CPU is more efficient than any filtering implemented using circuitry is a truism that does not need to be proven". Unsupported claims of absolute truth not needing proof always raise my eyebrows but in this case I don't even know what you mean so I've just asked for clarification of what you mean before questioning the claim.

 

[phofman]

IMO the word "efficiency" was meant as "precision" or "quality" instead.

 

[gwing]

I would never execute of upsampling my collection of digital photographs and preserving them, given the complete uselessness of the operation, it just cause huge introducing of artifacts withount indreasing the quality of the images.

Be careful when introducing analogies. Whereas upsampling your images and increasing storage costs might be a bit daft that doesn't mean you can't get image improvement. As an example, even going back many years, I had an Epson printer that when asked to print a low resolution image at large size would (optionally) upscale it in real time on its tiny processor when printing giving a much better and less pixelated image and good modern phot processing software can do better. I do agree though that upscaling an image for its own sake isn't going to improve quality - generally only useful when the required output size is larger than you would get natively on the display for your pixel count.

 

[MaxwellsEq]

Be careful when introducing analogies. Whereas upsampling your images and increasing storage costs might be a bit daft that doesn't mean you can't get image improvement. As an example, even going back many years, I had an Epson printer that when asked to print a low resolution image at large size would (optionally) upscale it in real time on its tiny processor when printing giving a much better and less pixelated image and good modern phot processing software can do better. I do agree though that upscaling an image for its own sake isn't going to improve quality - generally only useful when the required output size is larger than you would get natively on the display for your pixel count.

I agree. To a first approximation, upscaling/sharpening pictures and upscaling sound look like sensible analogies. But a bit of knowledge about the finite MTF limit of lenses, measured across sensor distance, not over time/frequency, shows that our "Nyquist" point with modern sensors exceeds what physically can enter the sensor. This is not the case with sound pressure waves where our Nyquist point can easily be lower than the bandwidth entering a microphone capsule. They are not good analogies of each other.

 

[widemediaphotography]

I will briefly report my young experience with HQplayer, clarifying and apologizing that I am not a English mother tounge , and that for example the term "efficiency" in the Italian language probably takes on a different meaning from an engineering point of view. I am an engineer and owner of a Research and Development company, which operates in the Automotive sector, and where any result obtained must be measured and certified by an external Autorithy. I can't sell my clients feelings or suggestions, but the facts and my approach to my passions remain the same.

As you can see, on my unoptimized laptop (i7 12700 and Nvidia GPU) with all my job software (189 active processes) I only use 15% CPU in DSD512 (limit of my DAC), many users stop it to DSD256 is alrady enough.

On my audio PC I use just Foobar2000 I have about 78 active processes and it has 27% of the Celeron N5100 CPU

I don't have to convince anyone but myself, but I want to do it with a scientific approach and not because a phenomenon like mass delirium exists or reject it because someone simply says it is useless.
We are here to discuss and every comment or consideration always does me good.

Thanks