CD player vs. DAC sound - can you hear a difference? (listening test)

[pma]

To contribute a bit to discussions on digital sources sound, I have prepared a test. The same song is recorded with ADC once from a CD player analog output and also from a DAC analog output. Two files, level matched before recording by a good quality analog pot and time aligned are uploaded here:

https://drive.google.com/file/d/137lGBoMJF2eZjmJhIeu2jP4rQeBBealv/view?usp=sharing

So if you are interested, please download the file, unzip, listen to the music files and let me know your thoughts. As always, foobar ABX test record would be appreciated.

 

[Blumlein 88]

Some interesting things in the results using Deltawave. I'll give it a listen tomorrow, and see how audible it is.

 

[restorer-john]

Just when I thought I'd had enough of ABX tests, along comes @pma and I can't resist.

 

[pma]

Some interesting things in the results using Deltawave. I'll give it a listen tomorrow, and see how audible it is.

Thanks. Filters, mostly.

 

[abdo123]

my best chance was with the CD player having more crosstalk but that wasn't the case.

it's pretty transparent.

 

[charleski]

As usual, on first listen I thought the second sample (beta) had a slightly better frequency balance. But no, going back and forth I couldn't replicate that. Do a frequency scan and beta (in green) has very slightly more top-end from about 5kHz on.

Trying to null out the samples leads to the usual problems with sub-sample phase alignment. I found that adding a 1-sample delay to beta resulted in a slightly better null.
Nulled as-is:

Nulled with a 1-sample delay to beta:

 

[Blumlein 88]

As usual, on first listen I thought the second sample (beta) had a slightly better frequency balance. But no, going back and forth I couldn't replicate that. Do a frequency scan and beta (in green) has very slightly more top-end from about 5kHz on.
View attachment 140661
Trying to null out the samples leads to the usual problems with sub-sample phase alignment. I found that adding a 1-sample delay to beta resulted in a slightly better null.
Nulled as-is:
View attachment 140663
Nulled with a 1-sample delay to beta:
View attachment 140662

Get Deltawave it's free and does sub sample alignment for you. Do a search for deltawave and pkane who is the author.

 

[artbur]

foo_abx 2.0.6d report
foobar2000 v1.6.4
2021-07-12 14:47:15

File A: alpha.wav
SHA1: 2edddc4333a3ee2f43f3d2c02602433f5b09f783
Gain adjustment: +2.25 dB
File B: beta.wav
SHA1: d6e8965b8daaeb938d38f415a952adbd5bd8df00
Gain adjustment: +2.25 dB

Used DSPs:
FabFilter Pro-Q 3

Output:
ASIO : MOTU M Series
Crossfading: NO

14:47:15 : Test started.
14:48:08 : 00/01
14:48:38 : 01/02
14:49:04 : 02/03
14:49:11 : 03/04
14:49:17 : 03/05
14:49:24 : 03/06
14:49:33 : 04/07
14:49:42 : 04/08
14:49:51 : 04/09
14:50:37 : 05/10
14:50:37 : Test finished.

----------
Total: 5/10
p-value: 0.623 (62.3%)

-- signature --
bcaa1ad04f8c481ea27fb820fde19a1bf210ba65

 

[pma]

As usual, on first listen I thought the second sample (beta) had a slightly better frequency balance. But no, going back and forth I couldn't replicate that. Do a frequency scan and beta (in green) has very slightly more top-end from about 5kHz on.
View attachment 140661
Trying to null out the samples leads to the usual problems with sub-sample phase alignment. I found that adding a 1-sample delay to beta resulted in a slightly better null.
Nulled as-is:
View attachment 140663
Nulled with a 1-sample delay to beta:
View attachment 140662

Thanks for trying the test. I think you got it quite right, the biggest difference is in filters. Though the "alpha" has minimum phase filter, "beta" has steeper linear phase filter. There is a difference at the highest frequencies, audible in a difference file, but most probably inaudible in the ABX with complete files.

The linear difference in amplitude/phase response is a part of the transfer function and possible audibility and should not be removed with diffmaker or deltawave - this is a message to @Blumlein 88 .

 

[pma]

foo_abx 2.0.6d report
foobar2000 v1.6.4
2021-07-12 14:47:15

File A: alpha.wav
SHA1: 2edddc4333a3ee2f43f3d2c02602433f5b09f783
Gain adjustment: +2.25 dB
File B: beta.wav
SHA1: d6e8965b8daaeb938d38f415a952adbd5bd8df00
Gain adjustment: +2.25 dB

Used DSPs:
FabFilter Pro-Q 3

Output:
ASIO : MOTU M Series
Crossfading: NO

14:47:15 : Test started.
14:48:08 : 00/01
14:48:38 : 01/02
14:49:04 : 02/03
14:49:11 : 03/04
14:49:17 : 03/05
14:49:24 : 03/06
14:49:33 : 04/07
14:49:42 : 04/08
14:49:51 : 04/09
14:50:37 : 05/10
14:50:37 : Test finished.

----------
Total: 5/10
p-value: 0.623 (62.3%)

-- signature --
bcaa1ad04f8c481ea27fb820fde19a1bf210ba65

Thanks for trying the test!

 

[artbur]

What CD-Player, DAC and ADC have you been using btw? Or will you reveal the used devices as well as alpha and beta only after the poll has closed?

 

[pma]

What CD-Player, DAC and ADC have you been using btw? Or will you reveal the used devices as well as alpha and beta only after the poll has closed?

Yes, later. It is not very important.
I will reveal at least something now. The sound source is CD. Once it was taken directly from the balanced output of the CD player, to the balanced preamp which served to equalize volume. From the preamp balanced output the signal went to balanced input of the sound card.
In the second recording, the coaxial SPDIF output from CD went to DAC input. From DAC balanced output to the same preamp and ADC as described above.

 

[pkane]

The linear difference in amplitude/phase response is a part of the transfer function and possible audibility and should not be removed with diffmaker or deltawave - this is a message to @Blumlein 88 .

View attachment 140677 View attachment 140678

Why would phase difference be removed? DeltaWave does not do any filtering on the files unless you ask it to, and then you have a choice of linear-phase or minimum-phase filters.

Here's the phase difference between the two files:

 

[3125b]

The differences aren't really audible (certainly not just played back to back with foobar).
Not the most comprehensive tool, yet Dynamic Range Meter does show them to be slightly different.

 

[charleski]

Get Deltawave it's free and does sub sample alignment for you. Do a search for deltawave and pkane who is the author.

Yeah, if you lin-phase upsample the files you have more resolution for aligning them, which is what I assume Deltawave does. But I can't get Deltawave to produce a difference spectrogram with a log frequency axis, which makes it harder to see what's going on.

 

[pkane]

Yeah, if you lin-phase upsample the files you have more resolution for aligning them, which is what I assume Deltawave does. But I can't get Deltawave to produce a difference spectrogram with a log frequency axis, which makes it harder to see what's going on.

No, the files are not upsampled unless you specifically ask for this. All the processing occurs at the original sampling rate.

 

[Tatr76]

I use foobar2000 so will have a go once figure out abx test. Just to add an only anecdotal observation is when Playing CDs in the past and had someone switch inputs on my amp between dac input fed by optical in from CD or straight from CD player via rca could not really tell when it was done let alone discern any difference in musicality.

 

[pma]

Not the most comprehensive tool, yet Dynamic Range Meter does show them to be slightly different.

Here is the more thorough result shown with Adobe Audition. IMO not bad taking into account the fact that the level matching was made purely in the analog path by means of blue Alps volume pot.

 

[pkane]

But I can't get Deltawave to produce a difference spectrogram with a log frequency axis, which makes it harder to see what's going on.

By the way, thanks for reminding me! This is something I planned on adding more than a year ago, but couldn't find the time. Should have this added later today

 

[charleski]

No, the files are not upsampled unless you specifically ask for this. All the processing occurs at the original sampling rate.

So how are you doing sub-sample alignment? Does it interpolate between samples?

By the way, thanks for reminding me! This is something I planned on adding more than a year ago, but couldn't find the time. Should have this added later today

That will be great.