MQA is Vaporware

[Jud]

Just now, The Computer Audiophile said:

 

That's where I was headed. 

 

Glad I could help. 

[audiventory]

Just now, The Computer Audiophile said:

 

Is there a scientific way to do figure out what someone likes best?

 

1. Careful double blind test only. For such test need pure software encoder and decoder of full MQA cycle.

 

When I listen, that MQA sound better than original I have in mind next things:

 

A) WAV1 > encode - lossy - decode > WAV2 = sound better

B) WAV3 > encode - lossless - decode > WAV4 = sound better

 

Better sound may be achieved as some lossy tricks well known to sound engineers (boost high frequencies, reverberation, 1-2 dB boost gain, easiest compression, etc.).

But, I suppose, it is not about case A.

 

However, If playback tool implementation for WAV2 better than for WAV2, WAV2 may sound better, despite loses. Also may be difference in volume.

 

In case B WAV3 is binary identical to WAV4. Why WAV4 sound better than WAV3?

 

 

 

 

2. Before we discussed that MQA is slightly lossy, but better that mp3, as example.

 

Today morning, I look to its site again. I read MQA benefits more carefully.

 

Quote

MP3 brings you just 10% of what was recorded in the studio. Everything else is lost to fit the music into a conveniently small file. MQA brings you the missing 90% – the full, rich experience – without any loss of convenience.

Source: http://www.mqa.co.uk/customer/how-it-works

 

I even don't know how understand it: 10% + 90% = 100%

It is lossless or not?

 

[audiventory]

13 minutes ago, Jud said:

 

Even if you do, there remains the question of what "likes best" actually means.  The story of the Pepsi Challenge and New Coke is instructive.

 

For me it more like to MP3 vs. AAC.

[jhwalker]

2 hours ago, Fyper said:

I had read about the Stereophile article: John Atkinson is not a producer and 4 positive identifications out of 7 isn't statistically enough not to be pure chance.

About the Audiostream article, it is clearly stated  : "The following document was provided by MQA"

I'd be happy to read about more independent and professional sources.

 

Obviously, you didn't read the links.

 

John Atkinson made the comments about *his own recordings*.

Morten Linberg made the comments about *his own recordings*.

 

 

[jhwalker]

2 hours ago, audiventory said:

 

1. Careful double blind test only. For such test need pure software encoder and decoder of full MQA cycle.

 

When I listen, that MQA sound better than original I have in mind next things:

 

A) WAV1 > encode - lossy - decode > WAV2 = sound better

B) WAV3 > encode - lossless - decode > WAV4 = sound better

 

Better sound may be achieved as some lossy tricks well known to sound engineers (boost high frequencies, reverberation, 1-2 dB boost gain, easiest compression, etc.).

But, I suppose, it is not about case A.

 

However, If playback tool implementation for WAV2 better than for WAV2, WAV2 may sound better, despite loses. Also may be difference in volume.

 

In case B WAV3 is binary identical to WAV4. Why WAV4 sound better than WAV3?

 

 

 

 

2. Before we discussed that MQA is slightly lossy, but better that mp3, as example.

 

Today morning, I look to its site again. I read MQA benefits more carefully.

 

 

I even don't know how understand it: 10% + 90% = 100%

It is lossless or not?

 

 

They are *very careful* not to say it's "lossless".  And as we know, it is not.

[Jud]

4 minutes ago, jhwalker said:

 

They are *very careful* not to say it's "lossless".  And as we know, it is not.

 

I think the reason to make the compression lossy was to make whatever encryption is used harder to bypass/crack, because you can't know what went in from what comes out.

[mansr]

1 minute ago, Jud said:

I think the reason to make the compression lossy was to make whatever encryption is used harder to bypass/crack, because you can't know what went in from what comes out.

That's not the reason. The difficulty of "cracking" the encryption doesn't depend on the lossy or lossless nature of the compression.

 

The main reason for making it lossy is that achieving the desired compression ratio is impossible otherwise.

[audiventory]

21 minutes ago, jhwalker said:

 

They are *very careful* not to say it's "lossless".  And as we know, it is not.

 

1. In the patent used "lossless" term https://www.google.com/patents/US9548055

 

2. 10%+90% also good linked with point #1

[audiventory]

Just now, mansr said:

The main reason for making it lossy is that achieving the desired compression ratio is impossible otherwise.

 

It may be reason.

[Jud]

2 hours ago, audiventory said:

 

For me it more like to MP3 vs. AAC.

 

Hi Yuri.  The point Chris (who will correct me if I misrepresent him) and I were making is this:

 

- There were people who preferred Pepsi to Coke under certain testing conditions.

 

- As a result, New Coke was formulated, which tasted more like Pepsi than the traditional Coke formula.

 

- Under non-testing conditions, the market resoundingly rejected New Coke, including many of the same people who had preferred Pepsi in the testing.

 

So this is about how test conditions affect preferences.

[Jud]

5 minutes ago, mansr said:

That's not the reason. The difficulty of "cracking" the encryption doesn't depend on the lossy or lossless nature of the compression.

 

The main reason for making it lossy is that achieving the desired compression ratio is impossible otherwise.

 

Thanks for the correction.

[Jud]

6 minutes ago, audiventory said:

 

1. In the patent used "lossless" term https://www.google.com/patents/US9548055

 

2. 10%+90% also good linked with point #1

 

Yeah, "audibly" lossless, which doesn't have a strictly mathematically defined meaning.

[audiventory]

3 minutes ago, Jud said:

So this is about how test conditions affect preferences.

 

Hi Jud,

Test conditions more important than test even

 

Market response more complex than better product. First is bigger number of people and their issues, that solved via product. I not once heard recommendations don't use focus-groups.

 

Some people think that double blind test between 2 option is enough careful.

 

But I heard opinion, that need add 3rd option (mp3 320k, as example), but nobody must know about it, only push "like"/"dislike"

 

May be test with 2 options assigned to MQA and 1 option for WAV or contrary. And listener should assign MQA or WAV for each option.

[Jud]

5 minutes ago, audiventory said:

 

Hi Jud,

Test conditions more important than test even

 

Market response more complex than better product. First is bigger number of people and their issues, that solved via product. I not once heard recommendations don't use focus-groups.

 

Some people think that double blind test between 2 option is enough careful.

 

But I heard opinion, that need add 3rd option (mp3 320k, as example), but nobody must know about it, only push "like"/"dislike"

 

May be test with 2 options assigned to MQA and 1 option for WAV or contrary. And listener should assign MQA or WAV for each option.

 

Hi Yuri.  I don't think that stuff works a lot of the time for audio, due to the short time span of echoic memory and other confounders.

[MarkS]

17 hours ago, Sal1950 said:

On what basis or evidence do you make the claim of the MQA files sounding "better"?

Any simple tweak of level or freq response could make the file sound "better",  but more accurate is an entirely different kettle of fish.

 

6 hours ago, Hifi Bob said:

The master is the reference. When comparing the master against MQA or any other lossy format, the master always wins—even if the listener doesn't personally like the sound, it's what the producer intended.

 

 

Response to quote 1: my ears

 

Response to quote 2: This is the nub of the issue. You're saying the master is the reference, but MQA is saying the sound of the instruments in the space in which they were recorded is the reference. MQA says that digitization of the analogue signal introduces a form of distortion (timing error) that is worse to the ultimate SQ than any issues associated with the lossiness.  MQA is basically saying that the losslessly encoded "master" sounds less like what the instruments/music sounded like than MQA because of MQA's correction of the digitization error (i.e., the "master" sounds more worse than the music that was recorded than their correction ?).

 

I think I'm done on this. I liked the demo I heard and hope others get a chance to hear MQA one day to form their own judgments. I also assume, as I initially stated, that MQA cannot turn a poor sounding recording into a good sounding recording. 

[Ralf11]

I just hope I'll be able to use MQA on my cassette deck....

[Fitzcaraldo215]

2 minutes ago, Ralf11 said:

I just hope I'll be able to use MQA on my cassette deck....

Get out your soldering iron!

 

[botrytis]

9 hours ago, Jud said:

 

Well, um, that kind of undercuts your calls for third party blind tests (you said double blind, but I don't think you actually meant double blind as correctly defined) for anyone wanting to say they like MQA, doesn't it? 

I do want to do a double blind test - because then you won't know which file is which. This is a common scientific testing method to rule out as many possible causes of bias as possible. If I don't know which file is MQA, one can't prejudge now, can they.

[Fitzcaraldo215]

2 hours ago, Jud said:

 

Yeah, "audibly" lossless, which doesn't have a strictly mathematically defined meaning.

Right, and this brings up one of the fundamental issues.  Stuart and his cohorts base much of this technology on psycho- and perceptual acoustics.  Some think that is tantamount to BS, primarily out of ignorance.  Others think, if it is based on careful experimentation, there may be some validity to perceptual and psycho acoustics.  I am in the latter camp.

 

I would have to see the context in which they used "audibly lossles".  Personally, I think this whole "lossless" idea is a red herring merely intended to discredit MQA.  MQA is a process claimed to be predicated on "improving" the sound.  One cannot  "improve" sound without changing the bits making up the audio signal in some way.  But, if the sound is changed for the purpose of improvement, how can something be judged to be "audibly lossless", which implies there is no change to the sound, with and without the process.  That is a standard useful only for compression/decompression schemes, not MQA, which is only partly such a scheme.

 

Put the other way, the "lossless" idea in the strict sense would preclude any audible improvement to the sound. I quite agree that if MQA were truly "lossless", it would be worthless, just another arbitrarily different compression scheme, ultimately just producing the same sound we already have.  I agree with those who say "we don't need that".  But, that is not what MQA purports to be.

 

I also agree with @MarkS. The idea, if at all valid, is that MQA attempts to reach beyond the existing infrastructure of AD and DA converters and closer to the true source, the sound of real instruments in real space.  Does it work as advertised?  "Lossless" is in no way helpful in assessing that.

[miguelito]

1 minute ago, botrytis said:

I do want double blind - because then you won't know which file is which. This is a common scientific testing method to rule out as many possible causes of bias. If I don't know which file is MQA, one can't prejudge now, can they.

I listened to MQA in March 2015 at Meridian in NYC. Long story short, MQA files sounded a lot better than the purported non-MQA files. The issue with all of these demos is it is unclear whether MQA is sounding better because of a remastering from the original source, or whether there's really a technological advantage in MQA. So I don't really know what A/B means in this case. 

[botrytis]

2 minutes ago, miguelito said:

I listened to MQA in March 2015 at Meridian in NYC. Long story short, MQA files sounded a lot better than the purported non-MQA files. The issue with all of these demos is it is unclear whether MQA is sounding better because of a remastering from the original source, or whether there's really a technological advantage in MQA. So I don't really know what A/B means in this case. 

Well, actually the problem is when Meridian does these demos, they talk about the file (I am assuming like going to hear, for example a demo of a new speaker being introduced by the manufacturer). This adds in a basic bias and your brain EXPECTS the files to sound better. If it doesn't then you brain fills in the information - it is all psycho-acoustics.

 

psychoacoustics

[miguelito]

35 minutes ago, botrytis said:

Well, actually the problem is when Meridian does these demos, they talk about the file (I am assuming like going to hear, for example a demo of a new speaker being introduced by the manufacturer). This adds in a basic bias and your brain EXPECTS the files to sound better. If it doesn't then you brain fills in the information - it is all psycho-acoustics.

The differences between the files was massive - no psychoacoustics required to explain them. Most brutal case was a 24/192 file followed by what was purportedly just an MQA encoding decoded to 24/192. But the difference was so stark that there's no way it was just the original file run through MQA encode/decode, it was surely a file that was poorly mastered in it's original 24/192 version and was remastered to MQA. If one would have remastered to a standard 24/192 and then run MQA encode/decode that would be a valid comparison. The staff was particularly dodgy about answering questions on all of this.

 

Joni Mitchel's "Blue", which I have in many forms, sounds to me ever so slightly better in MQA decoded by TIDAL to 24/96 than the HDTracks 24/192 file. But this is quite marginal, nothing like the difference I heard in MQA demos.

[Jud]

2 hours ago, Fitzcaraldo215 said:

That is a standard useful only for compression/decompression schemes, not MQA, which is only partly such a scheme.

 

Hi Fitz -

 

As to the filtering part of MQA, I agree.  Very few filters used in AD or DA are lossless.

 

However, the compression used in MQA is itself lossy, when as you've noted, lossless audio compression is ubiquitous.  That's where I fault MQA for using the term "lossless," which has a precise mathematical meaning when talking about compression, to mean an entirely different thing: that its lossy compression supposedly doesn't throw out any information that is perceptually significant.

 

I suppose it is somewhat pedantic to want certain words to have their traditional definitions rather than becoming malleable marketing concepts.  But once we start playing with those meanings, confusion can result, as I think is occurring with regard to MQA.

[botrytis]

1 hour ago, miguelito said:

The differences between the files was massive - no psychoacoustics required to explain them. Most brutal case was a 24/192 file followed by what was purportedly just an MQA encoding decoded to 24/192. But the difference was so stark that there's no way it was just the original file run through MQA encode/decode, it was surely a file that was poorly mastered in it's original 24/192 version and was remastered to MQA. If one would have remastered to a standard 24/192 and then run MQA encode/decode that would be a valid comparison. The staff was particularly dodgy about answering questions on all of this.

 

Joni Mitchel's "Blue", which I have in many forms, sounds to me ever so slightly better in MQA decoded by TIDAL to 24/96 than the HDTracks 24/192 file. But this is quite marginal, nothing like the difference I heard in MQA demos.

 

It is still psychoacoustics, no matter what. You cannot divorce bias from that type of listening.

 

It is not what I consider critical listening.

[jhwalker]

41 minutes ago, Jud said:

 

Hi Fitz -

 

As to the filtering part of MQA, I agree.  Very few filters used in AD or DA are lossless.

 

However, the compression used in MQA is itself lossy, when as you've noted, lossless audio compression is ubiquitous.  That's where I fault MQA for using the term "lossless," which has a precise mathematical meaning when talking about compression, to mean an entirely different thing: that its lossy compression supposedly doesn't throw out any information that is perceptually significant.

 

I suppose it is somewhat pedantic to want certain words to have their traditional definitions rather than becoming malleable marketing concepts.  But once we start playing with those meanings, confusion can result, as I think is occurring with regard to MQA.

 

MQA do not market or describe their process as lossless.  I've seen others use the term lossless rather loosely, but I've not seen that from MQA themselves - they are quite careful *not* to say it.