GoldenSounds passes apparently ABX test for DACs (NOT Really)

[amirm]

Here are the DACs I think:

View attachment 366453

and here are the same and the ABX device he has to to compare amps,DACs,etc.

View attachment 366454

The test as he says (and as I understand) is about the audibility and the quality of the filter in each DAC .

None of that was used in final listening tests.

 

[amirm]

The test as he says (and as I understand) is about the audibility and the quality of the filter in each DAC .

But that is not what he tests. No filter in any dac was tested. He is listening to two software algorithms.

 

[Mad_Economist]

It's possible he might be hearing IMD on the headphones way below 21 kHz, isn't it? "True" hearing around 21 kHz at 26 would be quite unusual, but not impossible.

Also, it seems somewhat pointless to claim you can hear differences between DACs with different reconstruction filters. Sure, I can use the super slow filter with severe roll-off past 10 kHz and ABX it with the standard filter (doesn't have to be another DAC) and I might be able to pass the test. Doesn't mean I could ever differentiate between different DACs with identical filters. The main point of the video "DACs matter" is thereby wrong, as it's more something like "reconstruction filters matter and specifically these two I tested only really matter for me as I can hear higher tones than 99.6% of all adults".

Meh.

This is one of the few remaining possible spuria here, but I'd propose that it's more fantastical to imagine that Cameron can hear the distortion products (themselves =/<-40dB from signal level with a Susvara in that frequency range) of a very small band of high frequencies which are =/<-40dB versus his signal level. It's not impossible, and it's something we'll be looking at in future ABXs, but such low magnitude nonlinearities would almost certainly be masked by even quiet audio band signals.

Like, comparatively, "an individual can perceive slightly above 20khz" seems a lot less magical hearing to me Of course, nothing in the video says that this is a major or even significant difference, and it should be completely obviated simply by not using 44.1khz sampling with the rare tracks that actually have 20khz+ content.

Also I was waiting he would mention intersample overs and if any attempt was made to control them but no, so unless I missed it there’s reasonable doubt that the differences he heard are intersample overs which are likely with the steep filter he used.

A good question! This is, indeed, why the output waveforms from the DAC were recorded - if you see any apparent intersample peaks there, please point them out, I'd love to have another explanation for the outcome of this test.

 

[Mart68]

at 11.50 on he explains what he does, the whole test is created in software, he's not switching between two DACs.

 

[amirm]

A good question! This is, indeed, why the output waveforms from the DAC were recorded - if you see any apparent intersample peaks there, please point them out, I'd love to have another explanation for the outcome of this test.

No output was recorded. As I have said repeatedly, he is testing between two different upsampled files using software that created them. No hardware, DAC or otherwise, is being tested here.

Further, if you play these files, you are subjecting them to the filter in your own DAC! So there is that interaction as well.

 

[Blumlein 88]

^ Very good response to this type of question IMO.

The only possibly viable argument I've heard against this is based on interesting data about the capabilities of the human eardrum, inner ear and auditory nerve - apparently there's some data suggesting that higher frequencies do reach the nervous system, but the human brain essentially employs a low-pass filter on it (as do most animals that evolved *alongside* other animals that use high-frequency echolocation such as bats and insects use.) Because of this there's a slight possibility that we 'detect' the high frequency and the act of filtering it out could cause fatigue sooner in some people. It's only speculation at this point based on inner-ear and auditory nerve measurements only. The part that's interesting is that this would be something that's by definition not detectable via ABX tests, because the conscious mind is not 'hearing' the frequencies after the low-pass.

I'd like to know your sources on this. What I thought was the case is our cochlea is what feeds our auditory nerve. The cochlea is built to respond to around 15 khz. The fact we hear more is simply that the cochlea acts as a filter and leaks a little beyond 15 khz. Maybe this is the filtering you are referring to above. People can sense high levels of ultrasonics when young by the ear becoming non-linear and mixing multiple tones to create something like IMD distortion. This is not so much hearing as having an odd perception of something. People who hear above 20 khz usually only do so at very high levels of sound pressure over 100 db SPL. People also can hear higher via bone conduction which would indicate the auditory nerve has additional response.

I don't know why this would not be detectable via ABX testing. People have at times detected differences in ABX testing while consciously believing they heard no difference. Fatigue occurring over time would be missed in quick switching ABX, but is something of a different issue.

 

[amirm]

This type of testing was done and published as peer reviewed paper. See my write up here: https://www.audiosciencereview.com/forum/index.php?threads/high-resolution-audio-does-it-matter.11/

Yes, that Bob Stuart, the creator of MQA! In his larger scale testing, they did find statistically significant difference between two DAC filters: “The audibility of typical digital audio Filters in a high-fidelity playback system,”Convention Paper, Presented at the 137th AES Convention 2014

 

[gvl]

Further, if you play these files, you are subjecting them to the filter in your own DAC! So there is that interaction as well.

I thought a high enough rate would effectively disable the filter. He could also disable the filter on the May, but I didn’t hear he said he did.

 

[Sokel]

But that is not what he tests. No filter in any dac was tested. He is listening to two software algorithms.

You're right,now that I downloaded them I got it.

Here's the two files (the high performance filter and the normal one),original,aligned and their delta:

Edit:Check down to my post #49 for the mono comparison which is the proper one.

 

[gvl]

Oh wow, you don’t really need to have a bat gene to be able to hear those frequencies.

What’s happening here?

 

[Sokel]

Oh wow, you don’t really need to have a bat gene to be able to hear those frequencies.

What’s happening here?

Yes,differences are about 2db from 200-20Khz and 25db at the filter.

 

[gvl]

Yes,differences are about 2db from 200-20Khz and 25db at the filter.

Are we getting a glimpse into the actual “quality” of HQ Player processing?

 

[Sokel]

Are we getting a glimpse into the “quality” of HQ Player processing?

HQ player is about DSD as far as I know.
(I may be wrong though,it's been a long time I tried it)

 

[juliangst]

I did some intensive audiometry testing last week for a university project and it turned out that my experiment partner (23 years old) has really impressive hearing skills (-5dB to even -10dB below the usual hearing threshold).

But even he couldn't hear past 17-18kHz.
I would be really suprised if Goldensound at 26 years can really hear 21kHz

 

[Mad_Economist]

No output was recorded. As I have said repeatedly, he is testing between two different upsampled files using software that created them. No hardware, DAC or otherwise, is being tested here

We did record the output of Cameron's DAC specifically to check for audio bands misbehavior due to being fed higher frequency content from one of the two files. I'm out presently, but I'll make sure that the analog recording files are in the public folder by tonight (and I think they already are?)

Further, if you play these files, you are subjecting them to the filter in your own DAC! So there is that interaction as well.

The files were reproduced at 176.4khz on the DAC for this reason!

You're right,now that I downloaded them I got it.

Here's the two files (the high performance filter and the normal one),original,aligned and their delta:

View attachment 366464


View attachment 366462

View attachment 366463

This delta looks very similar to what I got manually FFTing the two files - do you have clock drift correction on in Deltawave?

 

[Sokel]

do you have clock drift correction on in Deltawave?

Yes,it's the default.

 

[AnalogSteph]

I'm impressed by a 21k upper limit. I don't think I made it much beyond 16.5k even at his age. Mind you, I think my genetics are a bit potato in this respect.

One of the filters he used is very steep with cutoff extremely close to fs/2, much like the old SSRC filter. I've done some tests using more approachable values of fs in the past (think 11.025k), and these will ring like a bell when excited around fs/2 much like Monsieur Fourier would suggest. If this falls well within your hearing range, that is definitely quite audible. This is why SoX defaults to 95% of fs/2 for cutoff nowadays instead of trying to get super close, here the Foobar2000 plugin as an example (Audacity would do just as well):

 

[gvl]

HQ player is about DSD as far as I know.
(I may be wrong though,it's been a long time I tried it)

It’s a Swiss Army knife for anything audio filtering, it can covert to DSD yes.

 

[Sokel]

I did some intensive audiometry testing last week for a university project and it turned out that my experiment partner (23 years old) has really impressive hearing skills (-5dB to even -10dB below the usual hearing threshold).

But even he couldn't hear past 17-18kHz.
I would be really suprised if Goldensound at 26 years can really hear 21kHz

He has a proof in the download files ABX'ing 20Khz to 21Khz.

 

[juliangst]

He has a proof in the download files ABX'ing 20Khz to 21Khz.

He did hear something. Others speculated that he might have heard IMD