A/B test of two DACs - do they sound different?

[pma]

A/B test of two DACs – do they sound different?

Doing some measurements and changes in my audio chain, once again I had a feeling that the DAC that I have just connected into the system sounds different. Wow, better clarity and tighter, well defined bass! Not the first time such situation, right? Hopefully, my current system allows for doing A/B test without any hardware change, just switching the input selector switch on the preamp. So, I connected the newly tested DAC and my usual DAC and started to make a listening test, and also collected some measurements and files to be archived and for listening test.

Test setup

This is the schematic diagram of my DAC test setup.

PC notebook is connected, via USB-ISO isolator (this is crucial to eliminate the signal groundloop between PC, both DACs, preamp and ADC back to PC), to Topping D10s DAC. Analog signal from Topping output goes to INPUT2 of my ULTIM2 preamplifier:

ULTIM2 preamplifier

RightMark Audio Analyzer test : [ASIO] ULTIM2 96

Topping D10s also works as a digital bridge, sending input digital data to coaxial and optical Toslink output. I used the Toslink (again to eliminate groundloop) to send the data to my old Cambridge Audio DacMagicPlus (DM+). Analog output from DM+ is again send to the ULTIM2 preamp, INPUT1. Inputs are selectable by the input selector as seen in the ULTIM2 photo. Hopefully, the level difference at the ULTIM2 output, after switching inputs with both DACs, was less than 0.4dB.

For listening, ULTIM2 output was connected to my PMA-NC252MP power amplifier. For measuring and capturing files, ULTIM2 output was connected to E1DA Cosmos ADC stereo XLR inputs, at 4.5V input range selected.

Measurements

This is the result of THD distortion vs. output voltage, at 1kHz.

This is the result of THD+N distortion+noise vs. output voltage, at 1kHz.

We can see that the audio path with D10s has had about 4dB lower noise than the path with DM+. THD (w/o noise) is very low in both cases. Traditional multitone measurement has had dynamic range of >= 120dB in both cases. So, except for a slight noise difference, very similar.

Distortion data, DM+ and ULTIM2:
131072-point spectrum using Blackman-Harris 7 window and 4 averages
Input RMS 3.631 V
3.631 VC, 3.631 VA
3.631 V 22 - 22k UNW
Distortion at 999.9 Hz, -1.86 dBFS:
THD: -104.0 dB based on 21 harmonics [22..22000 Hz]
HHD: -130.2 dB [10 .. 22]
N: -105.1 dB [22..22000 Hz]
N+D: 31.22 uV A
THD+N: -101.5 dB [22..22000 Hz]

Distortion data, D10s + ULTIM2:
131072-point spectrum using Blackman-Harris 7 window and 4 averages
Input RMS 3.432 V
3.432 VC, 3.432 VA
3.432 V 22 - 22k UNW
Distortion at 999.9 Hz, -2.35 dBFS:
THD: -110.3 dB based on 21 harmonics [22..22000 Hz]
HHD: -132.1 dB [10 .. 22]
N: -110.1 dB [22..22000 Hz]
N+D: 14.64 uV A
THD+N: -107.2 dB [22..22000 Hz]

Listening test file
Music files were captured during testing and you may download them from

https://pmacura.cz/2_DAC_test.zip

and listen to them, after unzipping. They are in 44.1kHz/24bit flac format. You may try if you hear a difference or not. In case you thing you can hear a difference, please make an ABX test and post the ABX test report (like from Foobar ABX plugin).

Deltawave results
The captured files were tested by Paul @pkane excellent Deltawave software and here are the results:

Summary

I forgot to mention that I have not heard any sound difference in a level matched A/B test (not ABX, please note), so I did not bother to make ABX test. I made the A/B test after I had a feeling I heard the difference in a sighted “test” after DAC exchange in the chain. An illusion, as usually and so many times.

Have fun with audio!

 

[solderdude]

The input voltage of the DM+ = 3.631V
The D10S = 3.432 V
That is 0.49dB in favor of the DM+ on direct comparison (switching).
Based on Amir's reviews there would be a 0.4dB difference.

It appears that for direct comparison the 0.5dB difference was not detected which is interesting.
No problem for DeltaWave. It compensates for this (afaik) for creating a null.

What did the null sound like (amplified) ?
Can you post the null (audio) file ?

 

[Sokel]

Oh,I love this tests!
And I love more that stuff are measured as a chain,more real life-like.
As long as an 101db SINAD chain shows such small differences with an 107 SINAD chain,and as long as I have tested myself as to where is my absolute limit of noise and distortion I have no interest in chasing numbers other than the fun of measurements and the knowledge that brings to one.

There are other areas to improve things,as for example which device is more reliable or handles things better (like intersample overloads,matched volume level,etc).

 

[pma]

The input voltage of the DM+ = 3.631V
The D10S = 3.432 V
That is 0.49dB in favor of the DM+ on direct comparison (switching).
Based on Amir's reviews there would be a 0.4dB difference.

It appears that for direct comparison the 0.5dB difference was not detected which is interesting.
No problem for DeltaWave. It compensates for this (afaik) for creating a null.

What did the null sound like (amplified) ?
Can you post the null (audio) file ?

Hi, sure, here is the link to the null:

https://pmacura.cz/dactest_delta.zip

In my opinion, the resulting difference is in filters used. D10s has sharp, steep filter linear phase. DM+ has also the filter steep enough, but starting at a bit higher FC, and I use a minimum phase filter with DM+. This can be also seen in Deltawave difference phase plot. I do not use phase EQ with the Deltawave.

 

[solderdude]

The difference is what could be expected. Sounds like there is absolutely no difference in the bass and mids and that what is different is inaudibly low, above upper mids and treble and does not sound 'distorted'.

Thanks for posting

 

[pma]

The difference is what could be expected. Sounds like there is absolutely no difference in the bass and mids and that what is different is inaudibly low, above upper mids and treble and does not sound 'distorted'.

Thanks for posting

Thank you. The small problem is that we would not get much subjective opinion supported by technically usable data, from our members. I have no doubts that you and many others here understand what is it all about, but there is also a big camp of those who do not. I understand that the orientation is not easy and needs some knowledge and efforts. I have not much expectation that the test like this would make any change, but even if we help to one or two persons to get some insight, it would be a big success.

 

[pma]

Testing the same setup with a multitone. Again we can see that the difference is mostly in the phase response, due to difference in digital filters.

 

[Purité Audio]

Why is the Topping in their twice why not connect both dacs directly to the PC ( two USB isolators)
Keith

 

[pma]

Why is the Topping in their twice why not connect both dacs directly to the PC ( two USB isolators)
Keith

- it would make no difference,
- if DM+ would work with OPTO-ISO (I do not know, some soundcards do not, like Focusrite 2i2. Cosmos also does not work with USB-ISO)
- more important, I have only 2 USB lines in my notebook, one goes for DAC and the other for ADC.

Anyway, I think it is unimportant. Topping makes 1:1 bridge and DM+ works perfect from Toslink.

 

[Purité Audio]

For me that invalidates the whole test, especially as it would be simplicity itself to connect both dacs to the PC.
Keith

 

[pma]

For me that invalidates the whole test, especially as it would be simplicity itself to connect both dacs to the PC.
Keith

For me, your reasoning is a technical nonsense. For me, the way it was done is more pure than using 2 different USB outputs. How would you assign 2 DACs to one music file at the same moment?? 2 ASIO drivers at the same moment? It was the A/B listening test, with immediate switching between simultaneously playing DACs, by input selector.

D10s digital bridge is bit-to-bit precise and works ALWAYS and is independent of rest of the DAC circuit. Sir, you are a salesman, or engineer?

 

[Purité Audio]

I just connect two dacs via two USB ports and switch between the two via Roon, before that I used Audio Midi, I would never run one device through another.
Keith

 

[xeizo]

I just connect two dacs via two USB ports and switch between the two via Roon, before that I used Audio Midi, I would never run one device through another.
Keith

He's using it as a digital bridge, it can't affect SQ any more than your ISP providers switches

 

[Purité Audio]

For the cost of another USB isolator…
Keith

 

[pma]

Topping D10s as a digital bridge, USB to Toslink, 24bit, to Toslink digital receiver

 

[Blumlein 88]

I just connect two dacs via two USB ports and switch between the two via Roon, before that I used Audio Midi, I would never run one device through another.
Keith

The output of the Topping is excellent and bit perfect. In this case optical prevents any direct noise over power supplies. The only way it might matter is the clock of the Topping is what the DacMagic is using as it will be slaved to it. That if anything might make them sound more similar than if it were done separately. As both will have essentially identical clocking. Or if the DacMagic has a poor Toslink input and doesn't do a good job locking on cleanly. The test results PMA has shown indicate that isn't an issue.

I think it highly highly unlikely that results would be differenct if each DAC is fed off a different USB port if the USB connection is buffered and isolated.

 

[Purité Audio]

Is that how you would set up for that comparison\test?
Keith

 

[RayDunzl]

My dumb question:

How well matched are the two ADC channels?

 

[Blumlein 88]

Is that how you would set up for that comparison\test?
Keith

With the gear he had I very well might yes. I don't see a good reason it would cause issues.

I've done similar things like you are saying two different USB ports, but I had no USB isolation. Ran the PC off battery and had no ground loop issues. I didn't have a way for instant switching however. So I recorded with the ADC and used Foobar for the listening afterwards. Having done multi-loop tests I wasn't concerned with the ADC causing problems, but it would have been more 'pure' the way PMA has done it. Or if I had someone to switch it quickly for me as I listened.

 

[solderdude]

My dumb question:

How well matched are the two ADC channels?

Between L and R ?
pma does not compare by having DAC A on L and DAC B on R though.
The pre-amp switches between DAC A and B so there is only L and R on the ADC.
L-A + R-A or L-B + R-B (but balanced in) for the ADC.