PC+SW+DSD direct VS Any DAC

[widemediaphotography]

The point is not whether PCM 16/44.1 is a very comfortable and good audio stream. We cannot take for granted that all DACs are the same and therefore the quality of the 16/44.1 is not conditioned by the quality of the DAC (even if today the quality DACs are very similar to each other). The question is whether a PC does a better job than a good DAC in DSD direct mode

 

[MaxwellsEq]

The point is not whether PCM 16/44.1 is a very comfortable and good audio stream. We cannot take for granted that all DACs are the same and therefore the quality of the 16/44.1 is not conditioned by the quality of the DAC (even if today the quality DACs are very similar to each other). The question is whether a PC does a better job than a good DAC in DSD direct mode

But, as I keep pointing out, you need to have insider knowledge of the internal architecture of every DAC to be able to state that PC-enabled PCM-to-DSD conversation passed through the DSD chain in the DAC is better than the native PCM chain. It may differ from vendor to vendor and generation to generation.

 

[widemediaphotography]

But, as I keep pointing out, you need to have insider knowledge of the internal architecture of every DAC to be able to state that PC-enabled PCM-to-DSD conversation passed through the DSD chain in the DAC is better than the native PCM chain. It may differ from vendor to vendor and generation to generation.

I'll try to ask the same question another way... It's absolutely normal that the result may depend from DAC to DAC, as you say. However, the answer to my question ends if you can find at least one D-S DAC that performs better conversion than the competing system (PC based) as proposed in this Thread. Even better if the same DAC identified by you is DSD Direct so we can compare/measure results that we know share the same analog stage. Is transforming a PCM into DSD (with the possible addition of specialized filters) a better job than what happens in the digital stage of a good D-S DAC? In theory it would seem absolutely yes, in practice I don't know the answer yet.

 

[LGD_]

In audio enthusiast communities, the practice of performing analog filtering using a native DSD DAC with DSD direct functions is gaining increasing support. Even in the case of simple PCM 16/44.1, any Delta-Sigma DAC performs upsampling/oversampling for filtering and noise shaping. Therefore, by inputting directly with an improved equivalent in DSD 256/512, for example, to one DSD direct DAC performs the conversion directly without performing digital oversampling. The principle is based on the fact that a PC's CPU has a computational capacity far superior to any DAC chip and can therefore do a better job using more efficient real-time filters.

Using foobar's DSD Processor (part of the SACD add-on), I Up-convert all of my FLACs to native DSD 256 in real time on a Win 10 PC..
This is fed via USB to my DAC (SMSL D6) set to the direct DSD mode.
I do this because it gives a clearly audible SQ improvement over feeding PCM to the DAC .

I posted a explanation of my reasons for doing this to the D6 DAC Review thread- and received a lot of flac over it.
Some didn't understand what or why I was doing it, others just wouldn't accept the idea.
But, at least in my case on this DAC, it is a very clear improvement.

SMSL D-6 DAC Review

D-6 'PrD' is DSD pass through (as per D-6 manual) and this activates DSDD mode of the DAC chip, shown below from AK4493S block diagram. DSD Direct mode passes the DAC chip DSP (volume control and modulator) and goes straight to switched capacitor filters and through to analogue output. I...

www.audiosciencereview.com

(see post 374)

 

[MaxwellsEq]

Is transforming a PCM into DSD (with the possible addition of specialized filters) a better job than what happens in the digital stage of a good D-S DAC? In theory it would seem absolutely yes

Brilliant - we have a theory that can be mathematically modelled. Please describe the theory in sold engineering terms, starting with a stored piece of PCM encoded music.

 

[Sokel]

I'll try to ask the same question another way... It's absolutely normal that the result may depend from DAC to DAC, as you say. However, the answer to my question ends if you can find at least one D-S DAC that performs better conversion than the competing system (PC based) as proposed in this Thread. Even better if the same DAC identified by you is DSD Direct so we can compare/measure results that we know share the same analog stage. Is transforming a PCM into DSD (with the possible addition of specialized filters) a better job than what happens in the digital stage of a good D-S DAC? In theory it would seem absolutely yes, in practice I don't know the answer yet.

Will it be enough if I saw you how a DSD signal performs through a normal dac and how it performs through real time conversion PCM>DSD?

 

[widemediaphotography]

Will it be enough if I saw you how a DSD signal performs through a normal dac and how it performs through real time conversion PCM>DSD?

NO, because in this way I would see a DSD that will undergo manipulation in the digital stage of a DAC, while a PCM is needed, transformed into DSD that enters directly into the analog stage of the DAC. It is completely different!

 

[Sokel]

NO, because in this way I would see a DSD that will undergo manipulation in the digital stage of a DAC, while a PCM is needed, transformed into DSD that enters directly into the analog stage of the DAC. It is completely different!

DACs don't do conversions,this must be done upstream.
DAC only sees the signal that is fed,can't convert it of it's own.

Can you provide the scheme you're thinking?

 

[widemediaphotography]

DACs don't do conversions,this must be done upstream.
DAC only sees the signal that is fed,can't convert it of it's own.

Can you provide the scheme you're thinking?

Do we need a theory or schematics to prove that a CPU, in the digital domain, performs any type of filtering and modulation more quickly and efficiently than implemented using circuits or chips with more modest processing capabilities? There are SW created for this and HW adequate enough to support them. We are not arguing whether the measurable improvements are actually audible, maybe yes.
All D-S DACs in the digital stage oversample to filter out noise, this happens at high frequencies otherwise you would get phase rotation beyond noise reduction. Then there are the NOS DACs, which however still perform the filtering in the analog stage by continuing to oversample the signal (otherwise the filtering would be impossible).
If instead of oversampling in the digital stage, you transform the PCM into DSD and enter directly into the analogue stage... this is what you do for example with HQplayer and a direct DSD DAC like the Holo or others like the T+A. Just Google it

 

[Sokel]

Do we need a theory or schematics to prove that a CPU, in the digital domain, performs any type of filtering and modulation more quickly and efficiently than implemented using circuits or chips with more modest processing capabilities? There are SW created for this and HW adequate enough to support them. We are not arguing whether the measurable improvements are actually audible, maybe yes.
All D-S DACs in the digital stage oversample to filter out noise, this happens at high frequencies otherwise you would get phase rotation beyond noise reduction. Then there are the NOS DACs, which however still perform the filtering in the analog stage by continuing to oversample the signal (otherwise the filtering would be impossible).
If instead of oversampling in the digital stage, you transform the PCM into DSD and enter directly into the analogue stage... this is what you do for example with HQplayer and a direct DSD DAC like the Holo or others like the T+A. Just Google it

My argue is more on the format type.
Why do all this if the music you want is already on DSD format?

You just feed it to the DAC and you're done.
If not you just convert it upstream (with HQ,foobar,etc) and feed it as well.
No matter what DAC you feed (even a no-DAC,a straight resistor array) will see the same thing.

As for the conversion,I have done a gazillion tests (DSD to upsampled DSD,PCM to DSD,DSD to PCM,etc) .
What appears at the input of the DAC is of the same quality and specifically for DSD,the same signal.

So the DAC does the same thing all the time,it's only up to the upstream gear what to feed it,whether it's a PC,a SACD drive,etc.
DAC just don't care as it sees the same DSD stream.

 

[widemediaphotography]

My argue is more on the format type.
Why do all this if the music you want is already on DSD format?

You just feed it to the DAC and you're done.
If not you just convert it upstream (with HQ,foobar,etc) and feed it as well.
No matter what DAC you feed (even a no-DAC,a straight resistor array) will see the same thing.

As for the conversion,I have done a gazillion tests (DSD to upsampled DSD,PCM to DSD,DSD to PCM,etc) .
What appears at the input of the DAC is of the same quality and specifically for DSD,the same signal.

So the DAC does the same thing all the time,it's only up to the upstream gear what to feed it,whether it's a PC,a SACD drive,etc.
DAC just don't care as it sees the same DSD stream.

IMHO, from what you write you are wrong in at least 2 points.

1) Not all music is available in DSD (you can check on TIDAL or Quboz) and I never wrote that I was interested in the DSD format . Instead I say to upsample a 16/44.1 PCM file, transforming it into DSD256/512 and then sending it to the DAC, but only for the D-A conversion part excluding any digital filtering.

2) NOT all DACs are DSD-direct, your tests are completely useless if you haven't sent the DSD-transformed file to a suitable DAC (DSD-direct, like HOLO and others).

 

[Chrispy]

I believe you already know about that Forum with thousands of enthusiasts ready to swear, also supported by measurements, that having the digital filtering performed by a PC is much more efficient than having it performed by the DAC.

Mass delusions aren't particularly meaningful.

 

[widemediaphotography]

What arguments can you present to specifically state that what is being discussed is just a mass delusion? Do you know how a DAC works and how filtering and modulation take place in the digital stage?

 

[Chrispy]

What arguments can you present to specifically state that what is being discussed is just a mass delusion? Do you know how a DAC works and how filtering and modulation take place in the digital stage?

More importantly, rather than base your assumption on wildly varying subjective impressions without any standards, what if you apply some basic ones? Why are you here?

 

[widemediaphotography]

I'm here to discuss with those who know more than me and my discussion certainly doesn't stop at just reading that it's a mass delirium. I'm looking for someone who has done some disinterested measurements, other than what they do over there, who can confirm or deny that a DSD-transformed PCM sent to a DSD direct DAC is better than any simple PCM-to-DAC stream. I'm here because the theory behind their talk is correct, but what is the practice?
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!

 

[MaxwellsEq]

here because the theory behind their talk is correct

That's simply not proven so far.

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

Please unpack this.

Let's suppose I have a 30ips tape recording with musical content (harmonics) above, say 23kHz; OR, I have a live microphone feed with content above 23kHz which I've used a state of the art ADC to convert into PCM 16/44.1. Are you proposing that any artefacts introduced by the ADC's anti-aliasing filter can be fixed by converting the PCM into DSD? I find that hard to believe - I think that ADC "constraints" such as imperfect clocks and filters are permanently baked into the PCM content.

 

[widemediaphotography]

No, I'm stating two different things.

A) That a CPU is more efficient than any filtering implemented using circuitry is a truism that does not need to be proven. At most we should demonstrate that the improvement is perceptible to our hearing, but it is not the subject of the dispute. We are interested in measuring the improvement, a necessary condition for it to be possible to say that X is better than Y, just to remain within the scope of science.
B) I am not stating that any artifact can be resolved with a filter that can be implemented via SW or via HW, but that for example the ringing, always present and essentially due to the ADCs used for audio acquisition, can be resolved and this type of filtering it should not be confused with an artifact, because it is not present before the ADC.

 

[Blumlein 88]

Oh the ringing we can't hear (and which is not ringing rather a result of bandwidth limiting). The efficiency of using a larger file to accomplish an inaudible difference. The efficiency of a complex device to accomplish what can be done in hardware made once and then we don't worry about. At a very minimum you need a different word in place of efficiency.

 

[Snoopy]

I also do this in roon and did with HQplayer for a while. But it’s more for simplicity reasons. I have a whole bunch of DSD files (SACD rips, dsd purchased on various sites).. and switching to pcm with multiple samplerates will cause a mild „pop“ sound sometimes.
So by upsampling everything to DSD256 I bypass that problem and I also bypass the SDM in my dac (SMSL D300).

Since I use a computer for roon anyway, it doesn’t make a difference really if I’m streaming pcm or DSD. CPU and my network has no issue with DSD.

HQplayer actually does make a difference with the correct filters in place but it’s currently to much for my M1 MacBook.

But there are also interesting dacs that do DSD upsampling (Sony TA-ZH1ES, marantz SACD players , Teac has that in some devices), but it’s not really about „DSD“ but the filters implemented here.

Might not be a purist approach and it might even measure worse but so does Vinyl and R2R dacs.. as long as it sounds good, why not ? I just wouldn’t claim it is better because it’s this or that..

 

[gwing]

A) That a CPU is more efficient than any filtering implemented using circuitry is a truism that does not need to be proven.

That's twice you have mentioned 'efficient' now and I'm really not at all clear what you are referring to , could you explain what you mean by efficiency in this context please?