Can you hear a difference between dac+amp path with 92dB vs. 78dB SINAD? (test)

[pma]

Can you hear a difference between dac+amp path with 92dB vs. 78dB SINAD?

Quite a lot has been said here about “audibility” of SINAD, or existing/non-existing sound differences between amplifiers or DACs. Quite often we forget to evaluate SINAD of the whole audio chain consisting of a DAC, power amplifier, and eventually a preamplifier. It is the gain structure of the whole chain and individual noises of the components used that usually define the resulting SINAD of the complete audio chain, in case we have no ground loop issues. The gain structure and noise is a key, as will be explained.

Let's investigate 2 complete audio chains, one of them consisting of a DAC1, analog preamplifier and a power amplifier, the other one just from a DAC2 with digital volume control and the same power amplifier.

Chain1: Topping D10s DAC, OPFET preamplifier with balanced output (my design), Hypex NC252MP power amplifier.

Chain2: iConnectAudio4 DAC with digital volume control, Hypex NC252MP power amplifier.

Output signal is recorded from NC252MP outputs. Output voltage from NC252MP was 0 – 4V, to match common listening level at home listening conditions. NC252MP has 26dB gain (20x), so the DAC or preamp output would be only 0 – 200mV, about 20dB below the DAC full scale output.
-------------------

Chain1 with Topping D10s and the preamp shows THD = 0.002%, THD+N = 0.0025% and SINAD = 92.04dB at 4Vrms measured at the NC252MP output.

-------------------------

Chain2 with iConnectAudio4 shows THD = 0.00098%, THD+N = 0.013% and SINAD = 77.72dB at 4Vrms measured at NC252MP output. SINAD is now of 14dB worse, as a result of higher permanent voltage noise at the output of the iConnectAudio4 DAC analog output. Though there is one componet in the chain missing (the preamp), SINAD is worse due to higher noise of the DAC analog output.

--------------------------
THD vs. frequency comparison

The same track was recorded through Chain1 and Chain2 and you can download these 2 files from

https://pmacura.cz/dacamp_test.zip

The files are in 96/24 flac format. You can try if you hear a difference and if you do, please let us know and also post a foobar2000 ABX protocol here.

The files were also compared by Paul's @pkane excellent Deltawe software and these are the results:



------------

And this is my ABX report

foo_abx 2.1 report
foobar2000 v2.0
2023-11-23 15:29:47

File A: dacamp_test1.flac
SHA1: d18e80d55835f5f2db9e29c2246e9bb67e5e36e9
File B: dacamp_test2.flac
SHA1: fbab6bd34063f58dfb5962479bbe62c88810d35e

Output:
ASIO : Topping USB Audio Device
Crossfading: NO

15:29:47 : Test started.
15:31:05 : 00/01
15:31:25 : 00/02
15:31:38 : 00/03
15:31:54 : 01/04
15:32:24 : 01/05
15:32:40 : 02/06
15:33:07 : 02/07
15:33:18 : 03/08
15:33:47 : 03/09
15:34:05 : 04/10
15:34:25 : 04/11
15:34:46 : 05/12
15:35:05 : 05/13
15:35:33 : 05/14
15:35:55 : 06/15
15:36:16 : 07/16
15:36:16 : Test finished.

 ----------
Total: 7/16
p-value: 0.7728 (77.28%)

 -- signature --
4f3a6e8052276d96850c23e7c0f8ee297ffa2ab0

______________________________________
Edit
Original file

I am posting the original file as well:

https://pmacura.cz/dacamp_orig.zip

Please feel free to use it in your ABX comparisons as well.

 

[pma]

I forgot to show the difference in frequency responses between Chain1 and Chain2, so here it goes:

 

[pma]

Now, what is the subjective "sound difference" between the two chains? When I listen to them in my desktop arrangement, with small 2-way speakers each about 80cm from my head, the Chain1 is deathly quiet, though with Chain2 I can hear a slight hiss, if the ambient room is silent. Subjectively, it seems there MUST be a difference in sound. But, the ABX says there is no difference, it is just a technocratic bias based on knowledge of parameters difference. No difference at the listening level used, 0 - 4V from power amp. Still, there IS a possibility, that at very very low listening level, like in the night as a whispering sound, there might be a difference due to very different residual noise level of both chains under test.

 

[Warlequin]

I have a lot of experience with A/B testing, have tested the two files thoroughly and whenever I thought I heard an (extremely) small difference I was wrong on my random test. So I clearly vote: I hear no difference.

 

[Deleted member 48726]

Thank you. I can't hear any difference. -Bookmarked for reference. Might even go in my signature.

 

[pma]

@Warlequin , @Holdt - thank you, guys!

 

[Sokel]

Thank you. I can't hear any difference. -Bookmarked for reference. Might even go in my signature.

Yep,same here.
I even tried to use @pma 's tell about that slight hiss but nope.
If I want to be honest there's no difference.

Edit:Listening was through my main speakers,2.8m at MLP.

 

[pma]

I even tried to use @pma 's tell about that slight hiss but nope.

This might be audible only with no signal and ear close to the tweeter. It cannot be audible from the music recording, at the level used. The samples start with 2s of digital silence. After those first 2s any possible system noise is masked with the noise or just signal in the music sample.

I was thinking about recording at even lower level, like 100mV max. or less. However, the test would be questionable then, because everyone would play it at "normal" listening level, and not at the original one. Manipulating with process gain would kill the normal listening conditions.

 

[pma]

BTW this vote "I can hear a difference and I do not have an ABX result" posted in the poll may be considered interesting, but, to me, statistically irrelevant. ABX report, if possible, please, to be taken serious.

 

[Thomas_A]

Nice to see these ABXs trials. However, I already know this is below my threshold of audibility and there will be a null result.

 

[voodooless]

I was a naughty boy and voted "I cannot hear a difference" already yesterday without even downloading the file

So I just tried for real... I don't even need an ABX, the difference between the files is like night and day for a blind person! My vote stays where it was.

 

[pma]

the difference between the files is like night and day for a blind person!

. That's a good one.

 

[voodooless]

I forgot to show the difference in frequency responses between Chain1 and Chain2, so here it goes:

View attachment 328846

That’s not perfectly level matched, is it? The green one is the iConnect? It has quite a bit more passband ripple. The digital filter is way less sofisticated. It seem to use AK4627. Ripple is 0.02 dB. Looks about right.

 

[pma]

That’s not perfectly level matched, is it? The green one is the iConnect? It has quite a bit more passband ripple. The digital filter is way less sofisticated. It seem to use AK4627. Ripple is 0.02 dB. Looks about right.

It is perfectly matched, talking about music samples, within 0.01dB order (check Deltawave result between the files). The REW result you see is just the difference between the systems (there is no way to match the level absolutely perfect by volume knobs, iCA has digital volume knob and OPFET has stepped attenuator, both with finite steps), before the final level matching of music samples in Adobe Audition. Ripple is coming from Topping D10s DAC, as a rule. But we are here not to evaluate visual differences, but audible differences.

This is the result of Deltawave amplitude analysis (before Deltawave starts matching)

Report

Initial peak values Reference: -0.755dB Comparison: -0.773dB

Initial RMS values Reference: -15.341dB Comparison: -15.39dB

 

[dweeeeb2]

Thanks for taking the time to put this together

 

[pma]

Original file

I am posting the original file as well:

https://pmacura.cz/dacamp_orig.zip

Please feel free to use it in your ABX comparisons as well, but do not mix it into the poll.

 

[pma]

It is very, very interesting that the chain with worse SINAD and worse noise, Chain2, is closer to original data than the Chain1 which measures better in a traditional view. I guess, just guess, that it would be because of filters used. Below please see analysis of test files against the original in Deltawave. And, when listening to the Deltawave matched difference, again, with Chain2 the audible difference is much lower. This clearly shows how difficult it is to make an interpretation of measurement results and their correlation with audibility.

 

[dweeeeb2]

It is very, very interesting that the chain with worse SINAD and worse noise, Chain2, is closer to original data than the Chain1 which measures better in a traditional view. I guess, just guess, that it would be because of filters used. Below please see analysis of test files against the original in Deltawave. And, when listening to the Deltawave matched difference, again, with Chain2 the audible difference is much lower. This clearly shows how difficult it is to make an interpretation of measurement results and their correlation with audibility.

View attachment 329066

View attachment 329067

I am happy to accept that I won’t understand the explanation but fyi this has confused me.

 

[miero]

It is very, very interesting that the chain with worse SINAD and worse noise, Chain2, is closer to original data than the Chain1 which measures better in a traditional view. I guess, just guess, that it would be because of filters used. Below please see analysis of test files against the original in Deltawave. And, when listening to the Deltawave matched difference, again, with Chain2 the audible difference is much lower. This clearly shows how difficult it is to make an interpretation of measurement results and their correlation with audibility.

View attachment 329066

View attachment 329067

Are these differences adjusted for a delay and a clock speeds? If not for both, then it is possible that the worse DAC has its clock more in sync with ADC than the other one.

 

[danadam]

Chain2 with iConnectAudio4 shows THD = 0.00098%, THD+N = 0.013% and SINAD = 77.72dB at 4Vrms
[...]
THD vs. frequency comparison

How come there's such a big difference in chain2 THD in those graphs (0.00098% vs almost 0.005%)?