DSD Direct vs non-DSD Direct: Myth or Turth?

[sunjam]

Nowadays, virutally all DAC supports "Native DSD" but not all support DSD Direct mode.

Based on my understanding, ESS does not support DSD Direct mode at all (even they are problaby no.1 in terms of market share).

AKM does support it; CS41398 supports it but not their CS43131.

Based on ESS, looks like that DSD Direct mode is not important at all (as they don't care). Is it because that they are so big that they can dirve the market's need, or DSD Direct mode is indeed a myth; just some purists want their DSD music to be "bit perfect" down to the last digital-to-analogy step (even the difference in terms of the sound produced is virutally close to none)?

What's your experience in terms of switching between DSD Direct on/off? Does it really make a difference to you?

I have no experience personally as my DACs do not support it. I think it would help some of us (who have no experinece with it) in making the next DAC purchase decision.

Cheers

 

[ppataki]

I was just about to start a very similar thread
If you search ASR (and Google in general) you will find that people prefer DSD Direct vs Non-Direct in terms of sound quality but of course that is totally subjective - I have not seen any measurements out there to prove it

In two weeks' time I am going to get an SMSL D-6 DAC that does support DSD Direct and I will post here my subjective experience and also measurements comparing:
- PCM
- PCM converted to DSD with Jriver in realtime using Non-Direct DSD mode
- PCM converted to DSD with Jriver in realtime using Direct DSD mode

Stay tuned!

 

[sunjam]

I was just about to start a very similar thread
If you search ASR (and Google in general) you will find that people prefer DSD Direct vs Non-Direct in terms of sound quality but of course that is totally subjective - I have not seen any measurements out there to prove it

In two weeks' time I am going to get an SMSL D-6 DAC that does support DSD Direct and I will post here my subjective experience and also measurements comparing:
- PCM
- PCM converted to DSD with Jriver in realtime using Non-Direct DSD mode
- PCM converted to DSD with Jriver in realtime using Direct DSD mode

Stay tuned!

Thanks a lot for your upcoming tests. I'm looking forward for that.

By the way, could you also try your test with HQPlayer for DSD upsampling? If you don't have the license, they do give you free trial: it is free 30 minutes usage per session. After that you need to restart the HQPlayer again for another 30 minutes free trial. You can keep on doing it. Based on my experience they do provide the best DSD upsampling.

 

[Scytales]

Nowadays, virutally all DAC supports "Native DSD" but not all support DSD Direct mode.

Based on my understanding, ESS does not support DSD Direct mode at all (even they are problaby no.1 in terms of market share).

AKM does support it; CS41398 supports it but not their CS43131.

Based on ESS, looks like that DSD Direct mode is not important at all (as they don't care). Is it because that they are so big that they can dirve the market's need, or DSD Direct mode is indeed a myth; just some purists want their DSD music to be "bit perfect" down to the last digital-to-analogy step (even the difference in terms of the sound produced is virutally close to none)?

What's your experience in terms of switching between DSD Direct on/off? Does it really make a difference to you?

I have no experience personally as my DACs do not support it. I think it would help some of us (who have no experinece with it) in making the next DAC purchase decision.

Cheers

To my understanding, what you referred to as "Direct DSD" is just the possibility built in some digital to analogue converters (DAC) chips based of one particular technique, that of switched capacitor output, to drive said switched capacitor output stage with bitstreams produced externally to the DAC chip instead of using the internal modulator or signal processing stage of said DAC.

There is nothing special about that. It is intended to give application engineers more freedom of design. That's it.

To my knowledge, ESS DACs are not of the switched capacitor family, but of the PWM family. Hence no provision for a direct drive of the actual D/A converter output stage, because the noise shaped digital data must be mapped in some way in order for the converter stage to put out the correct pulse width corresponding to each digital input word.

I could be wrong, but I think that mostly addresses the issue.

 

[Keith_W]

I have a Merging NADAC. This has an ESS 9008 DAC. I believe that ESS DAC's are natively DSD, and that any PCM is converted to DSD prior to conversion to analog. The premise of HQPlayer is that the PC is better at PCM to DSD conversion than the internal converter on ESS chips. So let's test that.

EDIT: @Sokel pointed out that the charts were reversed. I have edited this post with the correct remarks. Sorry about that!

This is the output of my Merging NADAC using a 1kHz test tone at DSD256. The test tone was generated with REW, recorded as a .WAV file, and played back in HQPlayer. The output of the NADAC was captured with an E1DA and analysed with REW. The THD+N (SINAD) is -111.6dB.

This is the output of my NADAC with PCM44.1. The THD+N is -114.2dB.

What I see:

- There is more jitter on the PCM44.1 chart (look at the symmetrical peaks on either side of the 1kHz test tone)
- The overall THD remains the same, but the distortion components are different.
- The improvement in SINAD is mostly due to the lower noise floor in the PCM44.1 measurement. A 3dB improvement.

So there you have it. Slightly worse SINAD with DSD256, mostly due to higher noise floor, but it's only 3dB. Slightly better jitter.

I should note that a HQPlayer license is expensive, and the CPU requirements for convolving 8 channels of DSD is intensive. Doing 8 channels at DSD256 puts my CPU at 80% usage, and if Windows wanted to do anything else, there are audible interruptions to the music.

I probably can't hear a 3dB improvement in SINAD. But I can easily hear audio stuttering. This is why I use PCM44.

 

[Music1969]

In two weeks' time I am going to get an SMSL D-6 DAC that does support DSD Direct and I will post here my subjective experience and also measurements comparing:
- PCM
- PCM converted to DSD with Jriver in realtime using Non-Direct DSD mode
- PCM converted to DSD with Jriver in realtime using Direct DSD mode

Stay tuned!

how u will measure it pls? what equipments u will use

 

[ppataki]

how u will measure it pls? what equipments u will use

I will just use my UMIK-1 microphone in the main listening position and will perform sweep measurements in REW
That way not only the frequency response but everything else will be visible and comparable

 

[ppataki]

By the way, could you also try your test with HQPlayer for DSD upsampling? If you don't have the license, they do give you free trial: it is free 30 minutes usage per session. After that you need to restart the HQPlayer again for another 30 minutes free trial. You can keep on doing it. Based on my experience they do provide the best DSD upsampling.

I am afraid I can't promise that but I think using Jriver will suffice just to see if there is any measurable/audible difference

 

[ppataki]

btw last year I reached out to Topping to ask them about DSD Direct and this is what they told me:

 

[Geert]

I will just use my UMIK-1 microphone in the main listening position and will perform sweep measurements in REW
That way not only the frequency response but everything else will be visible and comparable

I appreciate the initiative, but did you ever see someone using a mic to measure differences between DAC's? Something with the S/N ratio of microphones and listening rooms.

 

[ppataki]

I appreciate the initiative, but did you ever see someone using a mic to measure differences between DAC's? Something with the S/N ratio of microphones and listening rooms.

Well, that is the best I can do - at least we will be able to see if the microphone 'hears' any difference
If others here have other ways (like a Cosmos ADC) then I would be happy to see those measurements too
btw I will not measure different DACs, I will measure the same DAC but with different signal type and settings

 

[Geert]

at least we will be able to see if the microphone 'hears' any difference

What you will see is differences in acoustic noise and other environmental changes.

 

[Sokel]

I have a Merging NADAC. This has an ESS 9008 DAC. I believe that ESS DAC's are natively DSD, and that any PCM is converted to DSD prior to conversion to analog. The premise of HQPlayer is that the PC is better at PCM to DSD conversion than the internal converter on ESS chips. So let's test that.

This is the output of my Merging NADAC using a 1kHz test tone at 44.1kHz PCM. The test tone was generated with REW, recorded as a .WAV file, and played back in HQPlayer. The output of the NADAC was captured with an E1DA and analysed with REW. The THD+N (SINAD) is -111.6dB.

View attachment 359559

This is the output of my NADAC with DSD256. The THD+N is -114.2dB.

View attachment 359560

What I see:

- There is more jitter on the DSD256 chart (look at the symmetrical peaks on either side of the 1kHz test tone)
- The overall THD remains the same, but the distortion components are different.
- The improvement in SINAD is mostly due to the lower noise floor in the DSD256 measurement. A 4dB improvement.

So there you have it. Proof that DSD is superior to PCM on my DAC. However, this 4dB improvement comes at a cost - a HQPlayer license is expensive, and the CPU requirements for convolving 8 channels of DSD is intensive. Doing 8 channels at DSD256 puts my CPU at 80% usage, and if Windows wanted to do anything else, there are audible interruptions to the music.

I probably can't hear a 4dB improvement in SINAD. But I can easily hear audio stuttering. This is why I use PCM44.

You probably have the charts reversed,the first one is DSD (we can tell by the out of band noise) and by the fact that it's usually DSD that gives a cleaner spectrum but at the expense of couple of db elevated noise (you can see it in the thread I did about it) .

You should also use some averaging (4-8)

 

[Roland68]

Well, that is the best I can do - at least we will be able to see if the microphone 'hears' any difference
If others here have other ways (like a Cosmos ADC) then I would be happy to see those measurements too
btw I will not measure different DACs, I will measure the same DAC but with different signal type and settings

And with such a measurement, what would be the point of having an amplifier, loudspeaker and measuring microphone in the signal path? It distorts the measurement and makes it nonsensical.
It would make more sense to measure directly at the analog outputs of the DAC.

 

[ppataki]

And with such a measurement, what would be the point of having an amplifier, loudspeaker and measuring microphone in the signal path? It distorts the measurement and makes it nonsensical.
It would make more sense to measure directly at the analog outputs of the DAC.

As I already mentioned above, anybody here with the right equipment is more than welcome to make it happen

 

[Keith_W]

You probably have the charts reversed,the first one is DSD (we can tell by the out of band noise) and by the fact that it's usually DSD that gives a cleaner spectrum but at the expense of couple of db elevated noise (you can see it in the thread I did about it) .

You should also use some averaging (4-8)

There are some really clever people on ASR! You are right, the charts were reversed and I wrote my entire post interpreting the wrong information. Sorry about that. I will go edit.

 

[Scytales]

I have a Merging NADAC. This has an ESS 9008 DAC. I believe that ESS DAC's are natively DSD, and that any PCM is converted to DSD prior to conversion to analog. The premise of HQPlayer is that the PC is better at PCM to DSD conversion than the internal converter on ESS chips. So let's test that.

Although ESS are always been sketchy about the internals of their DACs, we can say with confidence that what I have stressed in bold type is not correct. ESS DAC remodulate PCM in sigma-delta for D/A conversion, as almost every modern audio DACs on our planet do for more than 30 years. Sigma-delta modulation is not per se the same thing as DSD. DSD is a by-product of sigma-delta modulation which has been standardised to create a production chain and to allow its commercial exploitation, just as Sony and Philips defined the CD Audio standard to be in PCM 16/44.1 kHz and not in 18 bit/32 kHz or whatever else.

 

[Sal1950]

DSD Direct

DSD Direct, is this yet another MQA debluring BS myth to sell "special" DAC's ?

 

[ppataki]

DSD Direct, is this yet another MQA debluring BS myth to sell "special" DAC's ?

it is a way for the DSD signal to bypass certain elements (DATT Soft Mute and the Delta-Sigma modulator) in the DAC
see below the block diagram for the AKM AK4497 chip

 

[sunjam]

Although ESS are always been sketchy about the internals of their DACs, we can say with confidence that what I have stressed in bold type is not correct. ESS DAC remodulate PCM in sigma-delta for D/A conversion, as almost every modern audio DACs on our planet do for more than 30 years. Sigma-delta modulation is not per se the same thing as DSD. DSD is a by-product of sigma-delta modulation which has been standardised to create a production chain and to allow its commercial exploitation, just as Sony and Philips defined the CD Audio standard to be in PCM 16/44.1 kHz and not in 18 bit/32 kHz or whatever else.

Is it fair to say DSD is a subset of sigma-delta modulation (as SDM can be mutli-bit, or someone called it multi-level). For DSD, it just happens to be 1-bit SDM.

For what I learnt from the intenet, ESS is using 6-bit SDM internally (I could be wrong).