DAC listening test files - updated with new filters files

[pma]

Topping DX5 filters frequency response with Cosmos ADC at 96kHz/24bit, in high resolution on Y-axis. This is where I would be finding possible reason of eventual sound difference.

... and with Topping D10s added

Please note D10s highest passband ripple.

 

[pma]

Topping DX5 filters frequency response with Cosmos ADC at 96kHz/24bit, in high resolution on Y-axis.

.... and corresponding impulse responses measured at oscilloscope 12.5 MHz sample rate. (DAC in 96/24 of course)

 

[pjug]

Topping DX5 filters frequency response with Cosmos ADC at 96kHz/24bit, in high resolution on Y-axis. This is where I would be finding possible reason of eventual sound difference.

View attachment 555084

... and with Topping D10s added
View attachment 555085

Please note D10s highest passband ripple.

So if passband ripple is suspected, how does that present itself so that it can be heard? We know that ripple in the frequency domain is the same as echoes in the time domain. A summary of the math is below (from AI)

Suppose you observe ripple approximately like

H(f)=1+2acos⁡(2πfτ).

Its inverse Fourier transform is

h(t)=δ(t)+aδ(t−τ)+aδ(t+τ).

So you see in the time domain you have the main signal and a pre echo and post echo with τ delay before and after the main signal. We don't hear those as echoes because if the period of the passband ripple is 3KHz like in the D10s then τ is 333us. But it may audible in other ways. @pma are you able to describe what difference you are hearing in the cymbal crash around 10.5 seconds in the Test 2 file?

 

[pma]

So if passband ripple is suspected, how does that present itself so that it can be heard? We know that ripple in the frequency domain is the same as echoes in the time domain. A summary of the math is below (from AI)

Is there any scientific study to approve that the passband ripple of D10s as shown in this post in red would be inaudible? Thank you for the math generated by AI, but AI outputs are not any proof. You may create virtually anything in digital domain filters.

 

[pjug]

Is there any scientific study to approve that the passband ripple of D10s as shown in this post in red would be inaudible? Thank you for the math generated by AI, but AI outputs are not any proof. You may create virtually anything in digital domain filters.

I was lazy so used AI but those equations are well known. Do you want me to dig up a reference?

edit: I am trying to show that the passband ripple could be audible, not that it is inaudible. In these discussions, the corresponding echoes in time domain are usually suspected. The paper usually cited with the math is this. I don't know if there is a free link somewhere.
R. Lagadec and T. G. Stockham, "Dispersive models for A-to-D and D-to-A conversion systems", Audio Engineering Society preprint 2097, presented at the 75th convention, March 1984

edit: the equation is also given in Appendix B here

https://linearaudio.net/sites/linearaudio.net/files/03%20Didden%20LA%20V13%20mvdg.pdf

 

[pjug]

Something I am wondering is whether people like me who cannot pass ABX for Test 2 can train ourselves to hear the passband ripple and then become able to pass the ABX. So I was hoping you could describe what difference you hear to help with that.

 

[pjug]

Interesting that the apodizing fast rolloff filter is the default for the ES9038Q2M. That has the poorest passband flatness spec of all the filtyer options on that ESS
device. So probably most DACs like the D10s that use that IC go with the default and have the same passband ripple issue?

 

[pma]

@jabbr :

If even a single person can detect differences that their measurement array doesn't show, the array is incomplete. That is an objective fault

Measurements ALWAYS show differences. Always, in case your are competent to make them and understand them. The reality is opposite, not every measured difference results in audible difference. So your pressumption is wrong.

 

[pjug]

@jabbr :



Measurements ALWAYS show differences. Always, in case your are competent to make them and understand them. The reality is opposite, not every measured difference results in audible difference. So your pressumption is wrong.

He did say "measurement array", though, meaning a subset of possible measurements. I don't think what @jabbr said there was wrong about the measurement array being incomplete. I don't see this as an objective fault. @amirm cannot do all the measurements possible for every DAC he measures. Measuring every parameter of every filter of every DAC he is reviewing would be too much. Also, it should be repeated that it isn't the opinion of everyone here that differences in DACs can never be heard. Amir was clear that he did not hold this opinion. @pma appears to be able to hear differences in one of the DAC recording pairs he made himself.

 

[pma]

He did say "measurement array", though, meaning a subset of possible measurements. I don't think what @jabbr said there was wrong about the measurement array being incomplete. I don't see this as an objective fault. @amirm cannot do all the measurements possible for every DAC he measures. Measuring every parameter of every filter of every DAC he is reviewing would be too much. Also, it should be repeated that it isn't the opinion of everyone here that differences in DACs can never be heard. Amir was clear that he did not hold this opinion. @pma appears to be able to hear differences in one of the DAC recording pairs he made himself.

Again, there is a lot of words and no sensible suggestion. What him or you mean be "subset of possible measurements"? Do you speak about fast time domain measurements, VHF spectrum at the DAC output, susceptibility to external EMI interferences, or something else? It is all doable, however, an interpretation needs qualified approach and knowledge, and not only lay audience approach and mere speculations. I have not seen any sensible approach or real suggestions at AS, other than undefined common audiophile style speculations.

• time domain responses - you or anyone else please post the test that would verify audibility of the DAC/ADC system time response, other than it is covered by amplitude and phase response measurements.
• VHF spectrum at DAC output - please explain how it matters if we use headphones in the DAC tests.
• susceptibility to EMI - please you or anyone else show the measured data and how they affect a listening test result. And they will be unique for the user's system configuration, place of living conditions.
• susceptibility to system configuration - regarding signal ground loops, power ground loops, class of operation - yes it exists but is related to reader's own system, so not transferrable. However easily measurable, if one is technically competent

Just a note, I do all of the above mentioned except for professional susceptibility to EMI measurement.

My main objection to tests similar to that one at AS is a lack of real data that would be available for independent evaluation. Self acclaimed listening impressions do not make much sense. My main objection to those speaking about insufficient set of measurements is that they speak in vague terms and usually even do not know what they speak about, do not understand terms they are using in their arguing.

"@pma appears to be able to hear differences in one of the DAC recording pairs he made himself."
Yes, appears to be, but also he has posted easily measurable difference in the 20Hz - 20kHz audio band (when using enough resolution on Y-axis in dB) and easily measurable difference in impulse response (measured with 12.5MHz sampling rate) and it is the impulse response that completely defines the system transfer function. And all of this was measured in a complete test chain.

 

[pjug]

Again, there is a lot of words and no sensible suggestion.

I didn't want to make a suggestion. I wanted to almost do the opposite and say I wouldn't want to ask Amir to include more measurements (in reviewing DACs). At least not unless we can find some missing measurement in particular related to what Josh is hearing that is important. Unfortunately many at AS seem happier to let it all be a mystery.

 

[pma]

At least not unless we can find some missing measurement in particular related to what Josh is hearing that is important.

We do not know what Josh "is hearing". We do not have any valid data from him that would allow us to repeat his tests (which I however think is of not much interest). We have only his "audiophile style" review with his impressions and some uncontrolled test output. It is a vague basis that excludes any more sophisticated investigation. It just attracts attention, nothing more.

 

[welwynnick]

• o ttime domain responses - you or anyone else please post the test that would verify audibility of the DAC/ADC system time response, other than it is covered by amplitude and phase response measurements.

Yes, they are covered by amplitude and phase response measurements, but apart from SoundStageNetwork, phase measurements are rarely published.
That's a shame, as I think there may be significant differences in time domain response between DACs and filters, that are not captured by amplitude vs frequency alone.
In the last few days there have been several references to audible differences with high frequency transient sounds like some percussion, that DACs are not playing the same.

If one were to search for objective aspects that indicated audible differences, by measuring dynamic range, or SINAD at 1kHz, I'd say good luck.
If you measure everything in the middle of the envelope ( always the best case) then you will quickly convince yourself that everything is brilliant, and must sound the same.
I'm firmly convinced that the audible differences between equipment become apparent at the edges of the audio envelope.

• susceptibility to EMI - please you or anyone else show the measured data and how they affect a listening test result. And they will be unique for the user's system configuration, place of living conditions.

As a seasoned EMC R&D engineer, I applied what I knew about EMI and did a lot of experiments to lots of audio equipment, over the years.
I learned a lot, some of it unexpected, but the short of it is that EMI definitely makes a significant difference to measurements and to audible sound quality.
Unfortunately it's subject to confidentiality aspects of a patent application. But EMI is definitely a thing.

 

[welwynnick]

We do not know what Josh "is hearing". We do not have any valid data from him that would allow us to repeat his tests (which I however think is of not much interest). We have only his "audiophile style" review with his impressions and some uncontrolled test output. It is a vague basis that excludes any more sophisticated investigation. It just attracts attention, nothing more.

The internet is wonderful, but you can't believe everything you read. Learning who to trust is very valuable. Both you and Josh are people that I trust, and I listen carefully. I think it's unfair to describe his contribution to this particular episode in those terms. Coming from a subjective perspective, I think his efforts to bridge gap with objectivity are unusually forthright, and he went to great lengths to control the tests. Yes, many round here would like him to make greater efforts, but they're not paying him, they're typing on their keyboards.

This story is far from over.

 

[Plcamp]

RFSusceptibility does have a defined test regimen, in three major categories…

Conducted : Through a line impedance stabilization network, RF is injected onto cables. While the RF spectrum is injected, relevant outputs are monitored for noise margin…we used to test ‘idle channel noise’ on earpiece of local phone and connected phone to assure the noise levels remained below tolerable levels. (Aliasing effects were very common and difficult to overcome.)

Broadband : AKA ESD. We tested mostly for logical disruption of unit functionality, not analogue effects.

Radiated: Using a transverse electromagnetic wave cell (basically an exploded coaxial cable with the unit under test in the middle) we tested for audible effects up to a field strength of 5V/m. Could test with any modulated field, not just static.

All of these are localized effects on electronic equipment, and I would be very surprised if any audio equipment manufacturer paid any attention to them.

The effects can be very dramatic…ie easily audible.

 

[pma]

RFSusceptibility does have a defined test regimen, in three major categories…

Yes. I only do not speak much about it as I do not have professional, standardized measurements. I do have quite a lot of intuitive measurements, including those up to several GHz range. But I am not a fan of speculations. One of the reasons is my former work on this system, which included measurements in huge EMC chambers. So I am sorry to be underestimating the audiophile tales about EMC impact.

 

[amirm]

@pma, can you re-link what you want people to listen to from time to time? Thread is moving fast and in casual look, I have no idea which clips are being asked to be compared.

 

[pma]

@pma, can you re-link what you want people to listen to from time to time? Thread is moving fast and in casual look, I have no idea which clips are being asked to be compared.

I know. That is why the post #1 has been permanently updated, with links included, and also a PDF file attached with all the 8 tests, including the on-line ones. So please visit the post #1. I feel it is the only way how to keep actual updated info. It also contains updated chart with the test results of our members.

 

[pma]

To lighten the mood, anyone knows what is this?

 

[Iving]

To lighten the mood, anyone knows what is this?

View attachment 556434

Heartbeat