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

[gatucho]

As a bystander…
All I’m reading is theoretical jargon and flowery marketing terms from de-facto mqa supporters in this thread. 74 pages in and I still haven’t found a described benefit that’s not either vague or has been totally debunked.

You are talking with people that would buy a no-feedback amp with lower than 80dB SNR (just about 13bit BTW)and defend it like it is made of angel tears. However, in this forum there is a ranking with DACs/AMPs close to 120dB and it is supposed to be AScienceR.
It is not a surprise that a bunch of "audiophiles" will defend any nonsense they are indoctrinated to believe.

While it may technically feasible to encode some additional data using some tricks as those used by MQA, it will be only under particular scenarios (music-like signals). Therefore it will never be lossless. Say it like it is, "MQA is lossy but clever" and let people decide.

 

[voodooless]

Without mentioning lossing EQ capability. Since "bit perfect" connection to the DAC is expected! I really do not understand how @amirm is ok with this, him being a proponent of EQ

His beloved Roon can do EQ just fine on MQA tracks, so there is really no argument there. Of course that is not a full unfolding.

 

[gatucho]

His beloved Roon can do EQ just fine on MQA tracks, so there is really no argument there. Of course that is not a full unfolding.

IDK about roon, but the software I'm using cannot. So MQA is costing me more both in HW and SW and just for something I cannot be convinced I need!

 

[TurbulentCalm]

Without mentioning lossing EQ capability. Since "bit perfect" connection to the DAC is expected! I really do not understand how @amirm is ok with this, him being a proponent of EQ

I agree. The lack of EQ annoyed me and now I’ve dumped Tidal it is so nice to have it back. Though I am not sure how Tidal handles Replay Gain or Cross Fade in the Tidal App while still maintaining a pristine MQA stream that can still be ’Authenticated’ by the DAC.

 

[maverickronin]

Though I am not sure how Tidal handles Replay Gain or Cross Fade in the Tidal App while still maintaining a pristine MQA stream that can still be ’Authenticated’ by the DAC.

It's already a long thread, so all quotes!

Roon just got permission to fake it by adding back in the "authentication" code after the software DSP.

Lol is this true? Pretty sad..

IIRC, @mansr said it was all in the modulation of "noise" in the LSB. A DSP can just just keep track of the original LSB of each sample, make it's manipulations, and then replace the new LSB with the original before sending it to the DAC.

Roon just got MQA/Bob Stuart to agree to let them ship it in a licensed product even though it makes a mockery of their "authentication".

That is exactly how Bluesound/NAD do it. They take the output of the "core" decoder, save the LSB, do some processing, then put the LSB back.

 

[TurbulentCalm]

It's already a long thread, so all quotes!

Thanks.

And now we find that this stupid audio codec can’t even live up to its own name. What a joke!

 

[awdeeoh]

MQA is a gift to audiophiles who are experienced enough to know the magic of a good high-res file.

 

[PaulD]

Because it is a very interesting thing what they are doing, and it is a shame that people trash it just because they are told this could be worse (without even hearing it!),

Wrong. I have heard it and compared it to lossless CD quality (single blind, in a studio), MQA provides no audible advantage. ZERO. Given how it mangles the signal, I am surprised it did not sound worse.

Also because I think this encapsulation is a way to make a healthy music industry, allowing them to deliver the best possible quality and still have their assets protected, and I really, really love music

No, that will not happen. It is only making purveyors of "alternative facts" rich. This will not stop artists being ripped off - remember artists have mostly done poorly when there was buckets of money sloshing through the commercial music world, mostly the management got rich, the musicians far too often lose out. We all love music, it IS my life and it has been my livelihood in one way or another for most of my life... If MQA could improve this aspect of the industry I might even support it even as bad and deceptive as it is.

I am fully with John the Restorer, if it cannot encode any reasonable (band limited) signal properly it is broken, by definition, because then it cannot encode a music signal properly. It is clearly a step BACKWARD in quality. It also seems to be at least borderline illegal in some jurisdictions with the claims made for it. MQA is poison and some people cannot tell how bad it is, or they see a way to make an easy buck.

 

[guenthi_r]

Maybe the MQA-makers loves LSB-Steganography (or watched THIS video)...

 

[sandymc]

Forget the “3 bits” that people are mentioning. It’s a red herring (or maybe the sound of axes grinding).

We should be a little clear as regards the (now famous) 3 bits. So let me lay out my understanding of the JA/MQA argument as to why 13 bits is a "red herring":

1. the lower 3 bits are used by MQA for encrypted MQA data, and so are no longer audio in the sense that a non-MQA DAC can decode, other than as (shaped) noise,
2. the encrypted data in the these lower three bits are encoded as pseudo random noise, so as not to create audible artifacts on non-MQA decoders,
3. that encrypted data includes (probably) some of the information that was in the original lower three bits, encoded perceptually.

It's point (3) that creates confusion. If you replay the MQA track on a non-MQA DAC, the original lower three bits are gone. If you play the audio on an MQA decoder, then some of the three bits can be recovered. However, basic signal theory tells us that not all of the three bits can be recovered because of (a) the pseudo random noise encoding, and (b) there's other stuff in there.

So a 16 bit MQA file on an MQA decoder is, at a guess, somewhere between 13.5 and maybe 15 bit of effective resolution, depending on how good the MQA perceptual encoder really is. However, without actual technical details of the MQA encoder, that is an (educated) guess.

 

[voodooless]

encoded perceptually

I’ve now read this several times in several pro and contra posts. Somebody might shine a light on this burning question I have: how do you perceptually encode sound above 20 kHz?

 

[sandymc]

I’ve now read this several times in several pro and contra posts. Somebody might shine a light on this burning question I have: how do you perceptually encode sound above 20 kHz?

You have precisely identified the point at which I become very, very confused about how MQA works.

 

[mansr]

Thank you for providing such a detailed reply.

Detailed, perhaps, but not representative of what MQA actually does.

 

[TurbulentCalm]

I’ve now read this several times in several pro and contra posts. Somebody might shine a light on this burning question I have: how do you perceptually encode sound above 20 kHz?

Maybe by removing anything and everything above 22.05kHz before it is encoded into MQA.

Now admittedly the following is crude and may in the end prove to be in error, but I’ve been going through the various versions of the downloadable 2L audio files using NeutronMP and its inbuilt spectrum analyser and I’m yet to find a single track that shows any frequencies above 22.05kHz.

Not in any of the FLAC, the DSD or MQA encoded files.

If this is true, then the 2L tracks in effect contain only 16bit/44.1kHz, though some of the FLAC versions are multiples 48kHz (this is strange that they haven’t encoded these consistently).

So it would follow that most, if not all of these files are not truly Hi-Res and can’t not be used to assess MQA’s handling of ultrasonic data contained in the original source data.

Has anyone seen any Hi-Res tracks that actually contain ultrasonic frequencies that have also been encoded as MQA?

Even though there is no guarantee that original data of these two tracks would match perfectly, it would indicate which MQA tracks to study and compare what ultrasonic frequencies have been reconstructed via MQA.

Surely a visual analysis would indicate if there was major issues and inconsistencies in the official MQA files and help backup up @GoldenOne.

 

[mansr]

I’ve now read this several times in several pro and contra posts. Somebody might shine a light on this burning question I have: how do you perceptually encode sound above 20 kHz?

Oh, that's easy. Since it can't be heard, you don't encode it at all.

 

[mansr]

Has anyone seen any Hi-Res tracks that actually contain ultrasonic frequencies that have also been encoded as MQA?

There are plenty in the 2L set. I suspect you are mistaken about what NeutronMP does.

 

[TurbulentCalm]

There are plenty in the 2L set. I suspect you are mistaken about what NeutronMP does.

I think it’s worth others investigating because NeutronMP does identify and play MQA encoded FLAC files and outputs the correct bit depth and sample rate to the DAC which plays the decided music the same as playing these 2L tracks via Tidal in Master quality.

Not sure how I’m mistaken about this.

Further it would provide an accessible PCM data stream that could be compared directly with a non-MQA FLAC of the same bit-depth and sample rate.

Oops… edited. Sorry @mansr I mistook you for someone else.

 

[voodooless]

Oh, that's easy. Since it can't be heard, you don't encode it at all.

I knew you would say that . Would love to hear from the proponents how they think this works? BTW, do we actually know what kind of compression is used? All I read so far is speculation an conjecture. Also note that you can store a whole high bitrate Vorbis file in those lower bits, so there should be plenty of room to store some HF stuff.

 

[mansr]

I think it’s worth others investigating because NeutronMP does identify and play MQA encoded FLAC files and outputs the correct bit depth and sample rate to the DAC which plays the decided music the same as playing these 2L tracks via Tidal in Master quality.

Not sure how I’m mistaken about this.

If you're not seeing any spectral content above 22 kHz, then the MQA hasn't been decoded. Could you post a screenshot of the spectrum you're seeing?

 

[mansr]

I knew you would say that . Would love to hear from the proponents how they think this works? BTW, do we actually know what kind of compression is used? All I read so far is speculation an conjecture. Also note that you can store a whole high bitrate Vorbis file in those lower bits, so there should be plenty of room to store some HF stuff.

Looking at the decompiled decoder, it contains elements suggesting the use of linear predictive coding (very common in audio applications), but I never figured out how all the parts fit together (it would take weeks of full-time work). There are probably other techniques used as well.