MQA Deep Dive - I published music on tidal to test MQA

[blueone]

Another observation and conclusion:
MQA is THE solution against loudness war . Because heavily clipped tracks contain square waves which cannot be MQA encoded. So the mastering is forced to avoid clipping, otherwise the track may not find its way to Tidal.

I don't think so. The loudness war is played out in pop music, and compression is often applied too. Digital clipping is a mastering problem, perhaps more common than we think, but if MQA was more broadly adopted we might just get a little more care in mastering, or just more compression.

 

[Sir Sanders Zingmore]

So I guess in summary. I'd consider myself an objectivist, but I think many other objectivists expectation-bias themselves into being unable to hear a difference between two given devices because they don't expect that X level of the metric they're testing for would be audible.

This is an interesting point.
How do we guard against false negatives that might come about because the person being tested has an expectation bias against hearing a difference?

 

[amirm]

Here's an arbitrary 50 seconds, to match your plot length, I just stretched a selection to get the time, some short term sections are worse:

You have massive vertical scale and you have stretched the horizontal scale so much that it makes anything look flat. Follow the aspect ratio like mine and show the graph again.

 

[Jim in Florida]

I just played this Classical tract on Amazon Hi Res, Tidal MQA through Roon, and DSD through JRiver on a Gustard X16 MQA DAC.
I gave up trying to play MQA without Roon.

The Amazon and DSD sounded similar. Roon / MQA actually sounded better, The orchestra and violins especially weren't as hash, yet there seemed more detail.

Next I'll try the same thing with headphones....although I have three different equalizers for each source (Roon, JRiver, and Marantz) so probably not a good test.

 

[amirm]

A version of iOS?

The Android kernel was built on Linux, not iOS. MacOSX was built on BSD UNIX. Neither of those share a code base. iOS is Unix-like, but was not based on either Linux or Unix code.

Or are you saying that because iOS came first? Even though there are some GUI and functionality differences which now iOS has incorporated from Android, which make them similar?

Surely you were not so pedantic to run off with this, were you? My reference there was to the phone platforms, not lineage. "Somewhat" referred to the fact that Android too locks you into their ecosystem although not nearly as bad as Appl.

As long as we are being pedantic , MacOS was built on Mach OS variant of Unix, not BSD.

This is my book by the way:

So be careful in thinking you can pull a stunt with your knowledge of Unix. The first decade and half of my professional experience was working on Unix kernel and its variants.

 

[Inner Space]

This is an interesting point. How do we guard against false negatives that might come about because the person being tested has an expectation bias against hearing a difference?

Honestly, I think that's one of those "Atheism is a religion too!" arguments. My observation is that if the test is well set up, so-called objectivists will try very hard to hear a difference - partly to honor the process, and partly because it would be very cool to hear a difference, hunt down the hidden measurement, and establish a whole new thing. I think engineers secretly long for such a result, and I know for sure scientists do.

 

[amirm]

Honestly, I think that's one of those "Atheism is a religion too!" arguments. My observation is that if the test is well set up, so-called objectivists will try very hard to hear a difference - partly to honor the process, and partly because it would be very cool to hear a difference, hunt down the hidden measurement, and establish a whole new thing. I think engineers secretly long for such a result, and I know for sure scientists do.

I have had one famous objectivist just vote randomly to get through the test to prove there was no difference. I caught it because the trial time lag was a second or two! No way he was listening.

But yes, it is tricky business to fake it because I took the above test and passed it, super embarrassing the other fellow.

 

[Sir Sanders Zingmore]

Honestly, I think that's one of those "Atheism is a religion too!" arguments. My observation is that if the test is well set up, so-called objectivists will try very hard to hear a difference - partly to honor the process, and partly because it would be very cool to hear a difference, hunt down the hidden measurement, and establish a whole new thing. I think engineers secretly long for such a result, and I know for sure scientists do.

If “trying really hard” were sufficient to overcome bias then we could all do sighted tests and just try really hard.
Bias can be unconscious and surprising. Good experimental design should mitigate the risk of this bias causing false negatives. I’m asking how it might be done?

 

[dmac6419]

Surely you were not so pedantic to run off with this, were you? My reference there was to the phone platforms, not lineage. "Somewhat" referred to the fact that Android too locks you into their ecosystem although not nearly as bad as Appl.

As long as we are being pedantic , MacOS was built on Mach OS variant of Unix, not BSD.

This is my book by the way:

So be careful in thinking you can pull a stunt with your knowledge of Unix. The first decade and half of my professional experience was working on Unix kernel and its variants.

Ka-Boom

 

[earlevel]

You have massive vertical scale and you have stretched the horizontal scale so much that it makes anything look flat. Follow the aspect ratio like mine and show the graph again.

OK, but I only included 50 seconds because you did, not because I thought it was an important duration. This one is only a couple of seconds, and not the whole window because RX 8 doesn't let me set the vertical axis magnification, only resize to stretch it:

Again, I said something about if you aren't dealing with the extreme lows (the first two verticals are 20 and 40Hz, next is 100). But we're focusing too much on this. Again, it was an arbitrary example, picked at random, I didn't hunt for the best example. it was only about expectations for music.

But to UliBru's point, even this seems at odds with Bob Stuart's paper, no? That is, it's a question we have on the meaning in the paper, and on MQA limits. As I've pointed out, my curiosity in this is all about wanting to know MQA expectations, limitations. Again, failure of MQA to encode any particular thing is not reason to say it's useless, but I think it's a reasonable thing to want to have some feel for what these limitations might be.

 

[Inner Space]

If “trying really hard” were sufficient to overcome bias then we could all do sighted tests and just try really hard.
Bias can be unconscious and surprising. Good experimental design should mitigate the risk of this bias causing false negatives. I’m asking how it might be done?

Yeah, I immediately realized I was talking about desire, where you were talking about bias. My bad. Apologies. But a good question.

 

[amirm]

But to UliBru's point, even this seems at odds with Bob Stuart's paper, no?

I don't know, now you don't have frequency scale.

Regardless, if I were building MQA, I would simply analyze the music, distinguish between noise and real signal, and burry the extra info below the noise floor of the main audible envelop. Having a triangle doesn't do anything. So as noted, that may just be a marketing thing as there is no such thing as triangular encoding and even if there were one, it would wasteful by itself. Indeed, the main audible envelop in 44.1 and 48 khz is fully preserved anyway as far as spectrum so the triangle can't apply to that as they are showing.

 

[PierreV]

If “trying really hard” were sufficient to overcome bias then we could all do sighted tests and just try really hard.
Bias can be unconscious and surprising. Good experimental design should mitigate the risk of this bias causing false negatives. I’m asking how it might be done?

It's a real issue. In other settings, misdirection would be used. The subjects of a psychological test would have to be informed they are part of a psychological experiment (because informed content is required). They would be told the test would be of their arithmetic skills under different levels of noise and be split into two classrooms, with the same noise level, but painted in different colors because the impact of the color of the environment is what the psychologists would actually want to study. A bit hard to implement in this case I guess. (bogus example, just to illustrate the idea).

 

[pkane]

This is an interesting point.
How do we guard against false negatives that might come about because the person being tested has an expectation bias against hearing a difference?

The normal way is to include additional tracks as hidden references in the test to determine if the subject is capable of hearing differences that are well-established to be audible. The subject doesn’t know which track is the hidden reference and which is the test track. If the subject fails to detect a hidden bad track, the results of all their other tests can be thrown out. Either the subject was guessing, not sensitive enough or biased.

 

[raistlin65]

Surely you were not so pedantic to run off with this, were you? My reference there was to the phone platforms, not lineage. "Somewhat" referred to the fact that Android too locks you into their ecosystem although not nearly as bad as Appl.

As long as we are being pedantic , MacOS was built on Mach OS variant of Unix, not BSD.

This is my book by the way:

So be careful in thinking you can pull a stunt with your knowledge of Unix. The first decade and half of my professional experience was working on Unix kernel and its variants.

You are correct about Mach OS UNIX.

However, Android people won't agree with you that their phone platform is an "iOS version." That's pretty weak. Guess Windows 98 was an Apple OS version to you guys at Microsoft. lmao

But good deflection, proxy! This exactly how a proxy would act.

 

[DimitryZ]

I don't know, now you don't have frequency scale.

Regardless, if I were building MQA, I would simply analyze the music, distinguish between noise and real signal, and burry the extra info below the noise floor of the main audible envelop. Having a triangle doesn't do anything. So as noted, that may just be a marketing thing as there is no such thing as triangular encoding and even if there were one, it would wasteful by itself. Indeed, the main audible envelop in 44.1 and 48 khz is fully preserved anyway as far as spectrum so the triangle can't apply to that as they are showing.

I think it was more conceptual?

The baseband must be LPCM, so rectangular? And the ultrasonics can be something fancy, because they are not limited there.

 

[Sir Sanders Zingmore]

The normal way is to include additional tracks as hidden references in the test to determine if the subject is capable of hearing differences that are well-established to be audible. The subject doesn’t know which track is the hidden reference and which is the test track. If the subject fails to detect a hidden bad track, the results of all their other tests can be thrown out. Either the subject was guessing, not sensitive enough or biased.

I like this one

 

[dc655321]

As long as we are being pedantic , MacOS was built on Mach OS variant of Unix, not BSD.

macOS certainly has BSD components - the XNU kernel is an odd composition of the Mach/NeXTSTEP micro-kernel and FreeBSD/NetBSD network stack, filesystems, user-space utilities, etc. Can't do very much with just a (micro-)kernel...

To also be pendantic.

 

[Raindog123]

I guess no one here had to carry a BlackBerry… You know, with its ‘most trusted OS’.

 

[earlevel]

I don't know, now you don't have frequency scale.

Haha, sorry about that. Same as the other one, but anyway just for fun, a couple with the frequency scale showing, and from Apple Music for a little higher top end. The opening ~10 seconds of the song (top frequency line is 20k):

At about 1:30 into the song, that basically-noise-plus-squeal sound:

Like I said, just a comment on musical expectations, I think we've killed it. (PS—I have to listen to this at so much of a lower dB setting on my DAC than for typical tunes that it's ludicrous.)

Regardless, if I were building MQA, I would simply analyze the music, distinguish between noise and real signal, and burry the extra info below the noise floor of the main audible envelop. Having a triangle doesn't do anything. So as noted, that may just be a marketing thing as there is no such thing as triangular encoding and even if there were one, it would wasteful by itself. Indeed, the main audible envelop in 44.1 and 48 khz is fully preserved anyway as far as spectrum so the triangle can't apply to that as they are showing.

Fair enough, thanks.