Choosing a DAC for DSD playback?

[Music1969]

I'm just talking about fidelity of the sound and probably we can't really hear that anyway.

I'm not sure. This site (ASR) likes to mainly talk about objective stuff, so I shared some measurements above and stayed away from sound quality.

 

[diegooo1972]

Until i'm sure DSD conversion is done properly that's just seems a good thing as I said.
I sure believe in that measurements. And they look better.
And it's not that hard to believe that DAC behave better in some way and worst in others.
I'm just pointing out that this could be related to DSD conversion.
Sometime software algorithm implementation can be tricky and not transparent as you can think.

 

[Music1969]

Sometime software algorithm implementation can be tricky and not transparent as you can think.

You mean something 'bad' is being done with the PCM to DSD conversion, that can result in better measured outputs (as shown in above examples)?

 

[diegooo1972]

Exactly.
Bad is too much but i'd like to know that conversion is not intrusive in that sense.
In that consideration you do 2 things.
1) convert PCM to DSD
2) feed the DAC with DSD
Both can be responsible for that better behaviour.
Maybe they already know that conversion can't be responsible for that behaviour but I don't know that.
Maybe it's my ignorance about DSD but I Just like to understand and learn more.

 

[Music1969]

Exactly.
Bad is too much but i'd like to know that conversion is not intrusive in that sense.
In that consideration you do 2 things.
1) convert PCM to DSD
2) feed the DAC with DSD
Both can be responsible for that better behaviour.
Maybe they already know that conversion can't be responsible for that behaviour but I don't know that.
Maybe it's my ignorance about DSD but I Just like to understand and learn more.

No problem. I can't help with that unfortunately.

I can only share the measurements that I have shared, to maybe show up-sampling to DSD with certain DACs may not be 'useless'...

 

[diegooo1972]

Considering one more thing.
Even if we have excellent results the conversion between PCM and DSD is destructive. Or Lossy if you prefer.
In the strict sense that you can't obtain back the original bitperfect PCM once converted in to DSD.
Not saying that this is necessarily awfull. We are not talking about MP3.
There is no direct correrlation between PCM and DSD formats that allow me to have lossless conversion.
DSD rappresent audio in Amplitude domain while PCM in Time domain.
Something like AM and FM.
The fidelity in this kind fo conversion is anyway great. Just not perfect.

 

[Eirikur]

Newbie here. I'd like to know the difference in sound quality between the "DSD direct" path (no digital volume control; supported by some AKM-based DACs, such as RME ADI-2 DAC) and the normal path (with volume control); but there seems to be nobody talking about this (neither here nor other forums).

I'm not a "believer" when it comes to DSD, the contents needs so much (quantization) noise filtering that for me the most likely explanation for a "nicer sound" is that we like the sound of those filters!

That said, a strong point for me is that much more effort seems to be put into the mastering process for SACD - almost none of it will ever play on the radio so loudness wars may simply not apply. You'll often find that they are (re)mastered in Japan where apparently there is a much better market for the medium as well, and generally a strong work ethic to deliver quality. None of these are scientific arguments for DSD though, in all likeliness an 88.2/24 PCM rendition of the same material would be just as good.

PS: while objectively 44.1/24 PCM should be enough to capture everything audible, the luxury of a higher signal bandwidth allows for gentler low-pass filtering and keep aliasing artifacts way above audible ranges.

 

[Music1969]

the contents needs so much (quantization) noise filtering

DSD64 yes, but DSD256?

Quantization noise is pushed quite high no? Easier filtering? Simpler analogue section design?

Like this:

the luxury of a higher signal bandwidth allows for gentler low-pass filtering and keep aliasing artifacts way above audible ranges.

 

[pkane]

DSD64 yes, but DSD256?

Quantization noise is pushed quite high no? Easier filtering? Simpler analogue section design?

Like this:

DSD makes the D/A converter simpler, easier to implement. The measurements posted earlier are from HQPlayer software conversion between PCM and DSD. The software does offer a number of filter types, as well as different modulators and noise shapers. Overall, HQP is well implemented, but too many choices leads to spending too much time comparing settings with tiny or non-existent differences.

For some DACs, DSD is easier to process, so it may be an option to reduce some measurable distortions. Whether this will make an audible difference depends on the DAC, but in my experience, it doesn’t.

 

[Music1969]

DSD makes the D/A converter simpler, easier to implement.

Noted, this is the same thing I hinted.

Overall, HQP is well implemented, but too many choices leads to spending too much time comparing settings with tiny or non-existent differences.

Not sure. I haven’t changed settings for a year... probably depends on the person...

For some DACs, DSD is easier to process, so it may be an option to reduce some measurable distortions.

I said the exact same thing... and limited my comments to the DACs measured above...

 

[pkane]

Noted, this is the same thing I hinted.



Not sure. I haven’t changed settings for a year... probably depends on the person...



I said the exact same thing... and limited my comments to the DACs measured above...

I’ve measured a few other DACs that support PCM and DSD. I also spent some time comparing their performance in the audible band, both listening and measuring — there’s very little to no difference. There’s a lot more difference in the ultrasonic range.

 

[Music1969]

I’ve measured a few other DACs that support PCM and DSD.

Which ones bypassed internal DSP for DSD?

We’ve discussed ESS chips in this thread but they don’t have a way to bypass DSP with DSD.

Not many AKM chip DACs support “DSD Direct”.

Just couple examples.

 

[pkane]

Which ones bypassed internal DSP for DSD?

We’ve discussed ESS chips in this thread but they don’t have a way to bypass DSP with DSD.

Not many AKM chip DACs support “DSD Direct”.

Just couple examples.

Holo Spring. The other was an ESS DAC.

 

[Music1969]

Holo Spring.

Of the two , only this has DSD DSP bypassed.

And Miska has shown it does measure better fed DSD256, than PCM44.1 with it’s internal OS.

Spring2 even better again, I’ve seen.

Note I haven’t made any claims about audibility anywhere... just talking improved measured performance.

 

[pkane]

Of the two , only this has DSD DSP bypassed.

And Miska has shown it does measure better fed DSD256, than PCM44.1 with it’s internal OS.

Spring2 even better again, I’ve seen.

Note I haven’t made any claims about audibility anywhere... just talking improved measured performance.

As I said, it measures better above audible frequencies. This is what I measured, and what Jussi shared, as well. If you believe that ultrasonic content is important (as Jussi does) then go for DSD with Holo Spring. I didn’t find any significant differences in the audible range.

 

[Music1969]

As I said, it measures better above audible frequencies.

As one example, for iFi micro iDSD IMD and THD figures (in audible range....)

 

[Music1969]

If you believe that ultrasonic content is important

Jussi has said:

"Because those high frequency high level directly correlated components easily generate intermodulation products in the audio band, down to 0 Hz..."


Amir recently hinted ultrasonic content can impact audible range distortion too (d50s comment):

 

[pkane]

Jussi has said:

"Because those high frequency high level directly correlated components easily generate intermodulation products in the audio band, down to 0 Hz..."


Amir recently hinted ultrasonic content can impact audible range distortion too (d50s comment):

View attachment 30642

Look, I took a complex music recording, played it back through the DAC as PCM and as DSD using HQPlayer. Captured both using a high resolution ADC at 24/192. I then created a null comparison of the two captures. The difference was below -100dB up to 22kHz. You can conjecture about high frequency IMD reflecting into the audio band all you want, but if the resulting distortion is below -100dB, I’m willing to ignore it.

 

[Music1969]

You can conjecture about high frequency IMD reflecting into the audio band all you want, but if the resulting distortion is below -100dB, I’m willing to ignore it.

Not conjecture from me... Amir raised it as a possibility, just a few days ago... not in the context of DSD but ultrasonics generally when discussing distortion in audio band of D50s...

 

[pkane]

Not conjecture from me... Amir raised it as a possibility too, just a few days ago... not in the context of DSD but ultrasonics generally when discussing distortion in audio band of D50s...

View attachment 30643

I’m giving you an actual, measured example where no audible high-frequency IMD exists in PCM or DSD case in the audible range, and you are quoting Amir saying that it could possibly happen in a completely unrelated case?

I’m not sure why you feel the need to argue. I said in the very first post that DSD might work better with some DACs, but not in the few cases I actually tested.