DSD Performance Criteria

[MatrixS2000]

Of course the sarcasm.... I don't have the means like those available to ASR reviewers, otherwise I would not have posed the contextual question....

Well those people with the means are giving you answer - you refuse to believe it. So the ball is in your court not ours.

 

[Timcognito]

This is off-topic and out-of-context and belongs in some other thread regarding subjective comparative assessment of sound-quality of product encodings...

That's why it's FYI. Any topic on HiRes at ASR known is bunch of hooie. ASR measures well above any audible range and IMO is where things become a matter specmanship. One needs to find other another site, perhaps the blog on soundliason.com will have direction as they are all about DSD as well as some threads DYIAudio. Modifying audio playback devices with accuracy above far above audible range is waste of time and storage space.

 

[ahofer]

In fairness, ASR is still measuring DAC’s at thresholds far, far beyond audibility. Although I suppose that result is intended as a proxy for engineering quality.

It’s Amir’s site. I like what he does. He’s even free to be dogmatic if he likes, although I find just about any other audio site far more religiously dogmatic.

 

[Sokel]

The problem is how do you get that "1 bit signal"? Here are just the list of options of modulators in HQPlayer (and there are more options for other settings). Which one do you use to generate that 1-bit signal for the test? If you use a different one, are the test results still relevant?

The sampling theorem determines what the LPCM samples are. These "1-bit signals" are half processed intermediaries somewhere in the middle of one particular type of DAC (ΣΔ conversion) processing chain. Whatever happens before that point and after that point will both affect the final results.

View attachment 531947

From the practical side, here's how they look like (you can set the order at MTA's settings for the generator)
From 4th to 7th order modulator, DSD128 1kHz test signal, everything else the same.
(exact settings at the left of the chart) :


4th order


5th order


6th order


7th order


..and all together for easy visuals:

Above the 4th order which has couple of dB elevated noise, everything is pretty much the same.

If anyone wants further testing, let me now the parameters.

 

[Audicron]

From the practical side, here's how they look like (you can set the order at MTA's settings for the generator)
From 4th to 7th order modulator, DSD128 1kHz test signal, everything else the same.
(exact settings at the left of the chart) :

View attachment 532037
4th order

View attachment 532038
5th order

View attachment 532039
6th order

View attachment 532040
7th order


..and all together for easy visuals:


View attachment 532041

Above the 4th order which has couple of dB elevated noise, everything is pretty much the same.

If anyone wants further testing, let me now the parameters.

Am I correct in reading the input signal is a 96kHz PCM file being modulated to 1-bit 5.6MHz (DSD128) and the signal output device is limited to 96kHz with 24-bit dithering? It appears to me this is a measurement of the modulator(s) and not reflecting the dynamic-range of the source PCM file or a 32-bit truncation of the 64-bit DSP... The Shannon-Nyquist sampling theory is not in question... However, this does represent the noise-floor of the 24-bit output... Where is this measurement acquired... at the DAC output or in the computer?

My focus in the question is related to the measured output resolution and dynamic-range (ENOB) of the DAC platform when acting on 1-bit PDM signals, whether or not the signal has been filtered prior to D/A conversion. For example a direct input-to-output 1-bit PDM signal converted to analog via a simple low-pass filter versus a 1-bit PDM signal routed through the DSP architecture of the DAC platform, as is the case with ESS chipsets that decimate all DSD signals to PCM for Hyperstream processing or similarly in Chord DAC D.A.V.E. / WTA processing and D/A conversion of DSD signals.... (I know all digital signals are analog voltages.. except the bits residing in-or-on the storage medium before being read/interpolated...)

 

[Sokel]

Am I correct in reading the input signal is a16/44.14kHz PCM file being modulated to 1-bit 5.6MHz (DSD128) and the signal output device is limited to 96kHz with 24-bit dithering? It appears to me this is a measurement of the modulator(s) and not reflecting the dynamic-range of the source PCM file or a 32-bit truncation of the 64-bit DSP... The Shannon-Nyquist sampling theory is not in question... However, this does represent the noise-floor of the 24-bit output... Where is this measurement acquired... at the DAC output or in the computer?

My focus in the question is related to the measured output resolution and dynamic-range (ENOB) of the DAC platform when acting on 1-bit PDM signals, whether or not the signal has been filtered prior to D/A conversion. For example a direct input-to-output 1-bit PDM signal converted to analog via a simple low-pass filter versus a 1-bit PDM signal routed through the DSP architecture of the DAC platform, as is the case with ESS chipsets that decimate all DSD signals to PCM for Hyperstream processing or similarly in Chord DAC D.A.V.E. / WTA processing and D/A conversion of DSD signals.... (I know all digital signals are analog voltages.. except the bits residing in-or-on the storage medium before being read/interpolated...)

You're probably not familiar with measuring software and truth is that the ones generating DSD are rare.

So, Mutitone Analyzer has a DSD generator made from the ground up as test signals before that were rare and all over the place.
@pkane (the author) did a great job and delivered.

So, the test signal is 1kHz, pure, native DSD from the generator side, played through KTB using ASIO drivers playing DSD through and through, it's base freq family is 44.1k (could also be 48k but it's rare, I know of none such DSD recording) and the result is captured at 96kHz from the recording side.

To make it simple: Player plays native DSD128 and the recorder records it at 96kHz.
the bottom line of the chart has all the info you want for I/O .

Once again, this is a real chain, PC>MTA>DAC>ADC>MTA>PC.

Feel free to ask anything else.

 

[Audicron]

You're probably not familiar with measuring software and truth is that the ones generating DSD are rare.

So, Mutitone Analyzer has a DSD generator made from the ground up as test signals before that were rare and all over the place.
@pkane (the author) did a great job and delivered.

So, the test signal is 1kHz, pure, native DSD from the generator side, played through KTB using ASIO drivers playing DD through and through, it's base freq family is 44.1k (could also be 48k but it's rare, I know of none such DSD recording) and the rsult is captured at 96kHz from the recording side.

To make it simple: Player plays native DSD12 and the recorder records it at 96kHz.

Feel free to ask anything else.

So it is a measurement of the modulator(s) being interpolated by the 96kHz DAC output? Wouldn't this be the common denominator for an assessment of a PCM DAC and not an unfettered DSD signal produced by a simple-low pass filter of a DSD capable DAC? I am wondering how this applies to an evaluation of the 1-bit signal being handled by a DAC that is DSD capable? I see this applying to an evaluation of DAC architectures that convert the source 1-bit signal to PCM for output where the DSP interpolation may influence the output signal.

 

[Sokel]

So it is a measurement of the modulator(s) being interpolated by the 96kHz DAC output? Wouldn't this be the common denominator for an assessment of a PCM DAC and not an unfettered DSD signal produced by a simple-low pass filter of a DSD capable DAC? I am wondering how this applies to an evaluation of the 1-bit signal being handled by a DAC that is DSD capable? I see this applying to an evaluation of DAC architectures that convert the source 1-bit signal to PCM for output where the DSP modulation may influence the output signal.

Again, there's no PCM DAC output, DAC plays native DSD128 and that's it.
The 96kHz you see there is at the ADC side capturing the analog output of the DAC which (again) plays DSD128.
Look at the bottom, DSD128 out/96kHz in.

I could use lower or higher sample rate at the ADC side so to catch spectrum up to 48kHz or 96kHz range so to see what happens higher and the filtering beyond the audible 20Hz-20kHz range., that's easy to do.

 

[Audicron]

Again, there's no PCM DAC output, DAC plays native DSD128 and that's it.
The 96kHz you see there is at the ADC side capturing the analog output of the DAC which (again) plays DSD128.
Look at the bottom, DSD128 out/96kHz in.

I could use lower or higher sample rate at the ADC side so to catch spectrum up to 48kHz or 96kHz range so to see what happens higher and the filtering beyond the audible 20Hz-20kHz range., that's easy to do.

Okay... What is the nature of the DAC architecture handling the 1-bit PDM signal through to the D/A interpolation? This is an important fact... The obvious limiter is the LPCM ADC bit-depth and sample-rate.

 

[Sokel]

Okay... What is the nature of the DAC architecture handling the 1-bit PDM signal through to the D/A interpolation? This is an important fact.

You can see all about at the link I posted:

PCM vs DSD

Did a little comparison and I was wondering if any friend with better gear here has tested this. Results you're about to see is probably narrowed to the specific DAC but it's interesting to see how it handles things differently. I will not try to decipher the results other that DSD seems much...

www.audiosciencereview.com

You will see the differences between playing DSD and PCM through lots of measurements, don't stick to the first post only.
The particular is bad at PCM and better at DSD, could be the other way around as well.

You can do it yourself for your DACs, it's easy.

 

[Sokel]

Okay... What is the nature of the DAC architecture handling the 1-bit PDM signal through to the D/A interpolation? This is an important fact... The obvious limiter is the LPCM ADC bit-depth and sample-rate.

About the ADC, no limit there.
Here's an example with 192kHz from the ADC side capturing the analog output up to 96kHz (and all the garbage DSD is throwing away from the audible range) :

 

[Daverz]

And what do you think the DSD content recording engineers and content producers will say to this...? What a myopic definition of "Audiophile".

I don't need any testimonials from recording engineers, just a report from who anyone can pick out their preferred lossless format at least 70% of the time in a proper ABX test.

I didn't define "audiophile", but I cringe inside when I'm identified that way. Too much magical thinking and snobbery in that community.

 

[Sokel]

If the resolution sample-density of the ADC is 192kHz, it is decimating the 5.6MHz sample-density represented in the analog output signal of the DAC, I don't see this as being an accurate measurement of the 5.6MHz signal output of the DAC... what am I missing? Are telling me that the sample-density of a 1-bit 5.6MHz ADC is harmonically, equal to the sample-density of a 24/192kHz ADC? And are you telling me that a 24/192kHz D/A signal output from a DAC is harmonically equal to a 1-bit 5.6MHz D/A signal output where the 1-bit signal is a direct D/A conversion through a low-pass filter?

The ADC does not care about the sample rate density of the DAC, that's DAC's job.
The ADC only sees the analog waveform that comes of the output of the DAC which is then pure analog and that's trivial to capture.

As you see at the previous post the ADC has no problem capturing all the spectrum up to 96kHz and show DSD in all its glory, including the garbage it caries at its higher end.

Edit: want to make it super simple?
ADC can capture analog like from a microphone for example, that's what it does normally.
Now imagine the DAC playing whatever, a whole chain with speakers and all, etc and a mic capturing all this.

Now, scrap everything in the middle and hook the DAC straight to the ADC. Exact same result minus the room noise, and all.
It's that simple.

 

[Audicron]

About the ADC, no limit there.
Here's an example with 192kHz from the ADC side capturing the analog output up to 96kHz (and all the garbage DSD is throwing away from the audible range) :

View attachment 532059

(For clarification I deleted my post you referenced above in your last post and replaced it with this... sorry about the confusion.)
If the resolution sample-density of the ADC is 192kHz, it is decimating the 5.6MHz sample-density represented in the analog output signal of the DAC, I don't see this as being an accurate measurement of the 5.6MHz signal output of the DAC... what am I missing? Are you telling me that the sample-density of a 24/192kHz ADC is equal to the sample-density of a 1-bit 5.6MHz ADC and that the 24/192kHz D/A signal is harmonically equal to the 5.6MHz D/A signal? This is the ADC/DAC you are using?

https://media.digikey.com/pdf/Data%20Sheets/AKM%20Semiconductor%20Inc.%20PDFs/AK5385A.pdf

I fully understand the ADC has the encoded bit-depth and sample-rate of 24/192kHz, which limits the sample-density of the capture...

 

[Sokel]

al? This is the ADC/DAC you are using?

https://media.digikey.com/pdf/Data%20Sheets/AKM%20Semiconductor%20Inc.%20PDFs/AK5385A.pdf

That's the ADC, yes.
The DAC is ESS9038q2m, you can see both at the top of the chart.

 

[voodooless]

If the resolution sample-density of the ADC is 192kHz, it is decimating the 5.6MHz sample-density represented in the analog output signal of the DAC, I don't see this as being an accurate measurement of the 5.6MHz signal output of the DAC... what am I missing? Are you telling me that the sample-density of a 24/192kHz ADC is equal to the sample-density of a 1-bit 5.6MHz ADC and that the 24/192kHz D/A signal is harmonically equal to the 5.6MHz D/A signal? This is the ADC/DAC you are using?

https://media.digikey.com/pdf/Data%20Sheets/AKM%20Semiconductor%20Inc.%20PDFs/AK5385A.pdf

A 5.6 MHz 1-bit stream is not 5.6 MHz of useful audio detail. It’s mostly a high-rate noise-shaped encoding. What matters is the filtered analog output, and a 192 kHz ADC can measure that just fine in the audio band.

Didn’t you finish Sampling Theorem 101?

 

[voodooless]

That's the ADC, yes.
The DAC is ESS9038q2m, you can see both at the top of the chart.

I’m sure the comment will be that this DAC does not properly play DSD.

 

[Sokel]

I’m sure the comment will be that this DAC does not properly play DSD.

The particular implementation of this DAC plays DSD better than PCM (20dB lower IMD!)

The above problem is probably understanding the chain.

 

[Audicron]

A 5.6 MHz 1-bit stream is not 5.6 MHz of useful audio detail. It’s mostly a high-rate noise-shaped encoding. What matters is the filtered analog output, and a 192 kHz ADC can measure that just fine in the audio band.

Didn’t you finish Sampling Theorem 101?

It certainly contains more samples than a 192kHz signal... The Fc of the 5.6MHz signal will just define the noise-shaping and filter for playback. A 1-bit 5.6MHz recording is just that... whether or not you believe it captures contextually relevant audio energy is left to subjective opinion.

 

[Audicron]

The particular implementation of this DAC plays DSD better than PCM (20dB lower IMD!)

The above problem is probably understanding the chain.

I don't see where this device supports DSD... except in the data-sheet (if implemented) where I see there is no direct 1-bit PDM signal-path to a simple low-pass D/A circuit.

INTRODUCTION

Thanks for your purchase of the E-MU 0204 USB Audio Interface. The 0204

provides 2 analog inputs, 4 analog outputs and brings an unparalleled level of USB

audio quality to the Mac or PC, with pristine 24-bit/192kHz A/D and D/A

converters, ultra-low jitter clock, and Class-A, ultra-low noise mic/line/hi-Z

preamps. The signal-to-noise specs of the E-MU 0204 USB are unmatched by any

other USB interface on the market! From its plug-and-play functionality and

hands-on ergonomic design, to professional features like zero-latency direct

monitoring, the USB will forever change your expectations of USB audio.

Some of the other key features are detailed below:

• Record and playback support for a multitude of sample rates: 44.1k, 48k, 88.2k,

96k, 176.4k, 192k (176.4k &192k available on PC version only)

• Zero-latency direct hardware monitoring (disabled at 176.4k and 192k on the

Macintosh)

• Full 24-bit resolution and stereo-in/stereo-out at all sample rates

• Independent ground lift switches for both analog inputs help to solve potential

ground loop problems

• Headphone output can be used as separate stereo output from your software

with analog volume control

• Studio-grade headphone amplifier

• Anti-pop speaker protection minimizes noise during power on/off

• Ultra-low jitter clock subsystem: <100ps RMS

• Windows drivers: ASIO2 and WDM

• Macintosh driver: Apple CoreAudio

See the attached block diagram: