• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

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

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
12,418
Likes
14,919
Location
Riverview FL
The pre-amp switches between DAC A and B so there is only L and R on the ADC.

My mistake...

Didn't squint at the drawing hard or long enough.

Oh well...
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,921
Likes
8,432
Location
Prague
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.
Hi, I measured DacMagicPlus jitter on Toslink input years ago and it was excellent

dacmagic_toslink_jitter.png


Yes, Topping optical output is bit-perfect and we are just at the precision of 24-bit digital files in case of this test, talking about data. Technical reserves of DACs and SOTA DACs are still in analog section, however deep deep below any sign of audibility.

Topping D10s USB to Toslink digital bridge 24bit.png

Topping D10s USB to Toslink digital bridge 24bit.png
 

Mnyb

Major Contributor
Forum Donor
Joined
Aug 14, 2019
Messages
1,904
Likes
2,536
Location
Sweden, Västerås
PMA did a good enough test IMHO it's not a six sigma result from Cern :) but actually it is good science to repeat test now and then , the result is expected , but one should test basic premises now and then .

A more stringent test meant for a scientific paper would have specified 0.1dB or less and ABX. but for a home test 0.4dB and AB is better than most will achieve I cant do this . I'm glad someone else does it .

The sheer weight of the multitude of similar tests done over the years is the real deal. PMA's test reinforces the current consensus .
That's how we build "belief's" in a scientific results if test after test gives similar results it reinforces the results .

In some fields you make meta studies . you compare all tests in a specific field statistically . In some sciences results are not as clear cut as this and this approach is needed.
For example medicine , to avoid cherry picking results .If most test give a certain result we go with that conclusion.
This can be ripe with biases etc which studies to include ?

In this kind of test the protocoll of what to include is relatively easy. You compare only level matched blinded tests with each other.
And sighted test is not considered information , not even massive amounts of anecdotes contains any actual information in this sense (there are many good examples of this ).
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,921
Likes
8,432
Location
Prague
A more stringent test meant for a scientific paper would have specified 0.1dB or less and ABX.
You can do it, the post #1 has a link to level matched test files from both DACs. This is the level matching (enough, I hope):

Trimmed 0 samples ( 0.00ms) front, 0 samples ( 0.00ms end)
Initial peak values Reference: -8.407dB Comparison: -8.416dB
Initial RMS values Reference: -25.295dB Comparison: -25.299dB

Final peak values Reference: -8.407dB Comparison: -8.414dB
Final RMS values Reference: -25.295dB Comparison: -25.297dB

The fact that I could not tell the difference from speakers with levels matched to <0.5dB only indicates that ABX with the perfectly matched files will give zero success.
 
Last edited:

Mnyb

Major Contributor
Forum Donor
Joined
Aug 14, 2019
Messages
1,904
Likes
2,536
Location
Sweden, Västerås
You can do it, the post #1 has a link to level matched test files from both DACs. This is the level matching (enough, I hope):



The fact that I could not tell the difference from speakers with levels matched to <0.5dB only indicates that ABX with the perfectly matched files will give zero success.
Yes that's what i think too , hence this is good enough, your not going to space and no one dies if your wrong
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,921
Likes
8,432
Location
Prague
More on Topping digital bridge (USB to Toslink) transparency and if it may "affect" the sound (No!):

1) first file is from Arta Generator, 1kHz/24bit undithered
2) second file is (1) passing through Topping D10s to optical Toslink output, then fibre, then Toslink opto receiver and converter to USB. See the results below.

sig1k_Arta_24bit.png


sig1k_optobridge_24bit.png


The grass below -150dB is in the original file and it is nice to see how it is replicated.

In fact, the resulting files using the optical fibre are much cleaner than if there was a direct USB to DM+ or even the 2nd opto isolator (still capacitive currents contaminating single-ended links). Tested.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,921
Likes
8,432
Location
Prague
Why is the Topping in their twice why not connect both dacs directly to the PC ( two USB isolators)
Keith
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

In addition to spectra shown I made the test when original 999.76Hz 24-bit sine was sent through Topping D10s via USB to its optical SPDIF Toslink output (as in post #1), the optical fibre connected to SPDIF receiver and converter to USB. The original file and the file created by the optical path described here above were both exported into ASCII text files. 24-bit signal covers 16 717 216 discrete levels. The ASCII files (original - opto) were subtracted and the difference in individual samples is either 0 or 1 or -1 (from 16 777 216 levels) as a result of mathematical rounding of the 24-bit data, i.e. +/- 1LSB. 65536 data samples are shown in the attached document with the result of subtraction. To speak absolute true and precisely, there are 4 results of subtraction vith value "2" from the 65536 data samples, as the biggest error between original and USB-Toslink /SPDIF-USB digital data.

The method with optical fibre is in fact much better than 2 direct USB paths, as it completely eliminates any signal ground loop. Signal ground loop error in fact pollutes the resulting analog signal with disproportionately higher error. How do you check the data error and analog cleanliness in your connection via two USB ports?
 

Attachments

  • sig1k_orig-digibridgeopto.zip
    1.5 MB · Views: 9
Last edited:

tuga

Major Contributor
Forum Donor
Joined
Feb 5, 2020
Messages
3,730
Likes
3,804
Location
Oxford, England
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.

Did you compare the two DACs or the two ADC captures?
 
  • Like
Reactions: pma
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,921
Likes
8,432
Location
Prague
Did you compare the two DACs or the two ADC captures?
Both. At first, by listening on my main system, with speakers, A/B method. Then, when I found no difference by listening, I made 2 ADC captures. The procedure and test schematics is described in post #1.
 

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,118
Likes
551
...it can't affect SQ any more than your ISP providers switches
But those make a HUGE difference! I changed from AT&T to Comcast and the music sound 16.88% better, it is like "night and day" it is so "obvious" the "tremendous" difference!
 
Top Bottom