• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

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

Status
Not open for further replies.
I can see where you're coming from, but it would make a reviewer's job a lot more difficult if they hear things that aren't represented in the measurements, since they'd have to set up a blind test every time.
Aww, poor "reviewers"... we certainly wouldn't want to make their difficult and stressful job harder now would we? It must take a lot of effort to come up with all those fancy words to write a "review"... one can only assume they all have a good thesaurus handy. ;)

Correct, blind test every time... or it's simply someone's ears which is pretty damn random.



JSmith
 
@UliBru , you are exactly naming the confusion I'm also seeing ;-)
The Triangle of Confusion.

Also, as mentioned, MQA is able to handle high-energy high-frequency bursts or even sustained signal within the range it folds the higher frequency data into.
 
I hear you... Yet, while not available in commercial digital-audio implementations, this - the huge amount of bandwidth (processing and communication) - can be arranged, see my post above... Let's all think about if/how we can orchestrate this.
I think we've been there, already. Use full software unfold like in ROON.
 
Speaking of triangles in published music...
FLAC snippet here:
https://hydrogenaud.io/index.php?topic=120491.msg993500#msg993500

triangle.png
 
There's embellishing and there's straight out and out lying. MQA lies. About lossless, about "authenticated", etc. It's as simple as that. Therefore, I wouldn't take anything they say about their product on faith. The opposite: a good chance they are lying there, too.

don't get me wrong, i don't care for MQA either, not when i can just subscribe to Apple Music and call it a day, especially since i'm a student so prices for what is soon-to-be hi-res music assuming i even care about that (i generally don't) is effectively next to nothing...
 
I dont care about MQA at all but the side discussion about what can be considered music is interesting to me. I would like your opinion on the possibility of encoding some more experimental music.

A poster above already mentioned Ryoji Ikeda. A japanese electronic music artist that is known for using sine waves at the end of the audible spectrum and random noise. I actually enjoy listening to this on occasion.

I am including an audissey spectrum from his album dataplex, 02. Data Simplex. Flack redbook. https://en.wikipedia.org/wiki/Dataplex

Please note the warning: "Caution! This CD contains specific waveform data that performs a data-read test for optical drives. The last track will cause some CD players to experience playback errors, with no damage to equipment."

image.png
 
"better than lossless" that's exclusive to MQA.
It's marketing... but that said it's an interesting idea from a technical perspective as it's in effect mathematically lossy and potentially perceptually lossless at the same time. It's not lossy in the same way as MP3 for example.

If MQA was a little less focussed on marketing spin, I think there would be more respect for the engineering behind it. Many people may not like the marketing style of MQA (understandable and myself included), but from a purely technical perspective it is a very interesting codec.

It's reasonably clear what MQA does, how it does it and why. I'll cite this article published in 2016;

https://www.soundonsound.com/techniques/mqa-time-domain-accuracy-digital-audio-quality

"High-res" audio is often also a scam and can even be worse in theory than a well performing perceptually lossless codec due to ultrasonics being a liability for potential intermodulation distortion artefacts, post filter, so it is surprising that there is not more focus on that. Many unsuspecting customers are getting scammed already with "high-res" audio, however this appears to not be discussed that often. In reality there really is no need for anything higher than 24/48. If that had been the case then there would never have been any desire to develop MQA as one perspective.

I don't support embellished marketing, but I do support interesting ideas to potentially further the enjoyment of audio and the exploration of same.

Anyway, it looks like MQA is in a precarious position if Tidal decide for some reason to pull the pin and I see no longevity in that from a commercial perspective.



JSmith
 
Do we care about how MQA actually performs or do we care about bullshit marketing talks?

I'd say both. Especially by the questions some of the newcomers seem to ask, I bet there are thousands of consumers out there that don't ask these questions and based on the messaging assume things that are not correct.

MQA team produced an award winning peer reviewed paper that showed filtering high res audio can create audible differences.

And highly criticized as I pointed out earlier. It's highly doubtful the results can be extrapolated to current state of art DAC's. That doesn't mean the results of the paper are not valid or peer review failed. But as always, the devil is in the details.

Oh man, I had to dig long to find this next one ;)

You partially answer your own question there. There are two misunderstandings here:
We see that the spectrum nicely drops off as frequencies increase -- precisely what MQA algorithm is counting on.

I've read this now several times, but do we have any source other than MQA itself to confirm that the compression is actually contingent on this? Because a whole lot of the discussion hinges on this element? Because as far as I know we actually don't know a lot about how the lossy compression works. Because so far I found some information that might indicate the contrary?

Up to now we haven’t found information whether the compression algorithm can be put in a state of overload when the upper frequency energy and entropy is that high that the linear prediction algorithm generates residual errors too numerous to be encoded in the lower 7 bits of the baseband
 
Last edited:
@amirm I have to confess to being somewhat lost on the current state of the signal in the discussion as opposed to the noise. My understanding is that you feel that MQA is to be commended as a schme for capturign all the musical information in a 24 bit file in a manner which is essentially lossless (although not strictly interchangeable with the original). There are several particular points which I cannot seem to find the answer to
1) is there any evidence that it actually does do so reliably? After all we all know that there are any amount of nonsense files sold as high rez and surely this forum's raison d'etre is testing claims. Sorry if I missed this.
2) is there any evidence as to how it does so?
3) I still can't understand why the (or at least a) correct comparator is not noise shaped 16/96 or 16/88. What's wrong with that? AFAIK MQA offers little or no file size advantage over such a file.
4) where the musical triangle extends above 48KHz (which IIRC it is alleged to albeit not that often) how is this captured by MQA without significant aliasing in the <48Khz range?
 
Last edited:
@amirm I have to confess to being somewhat lost on the current state of the signal in the discussion as opposed to the noise. My understanding is that you feel that MQA is to be commended as a schme for capturign all the musical information in a 24 bit file in a manner which is essentially lossless (although not strictly interchangeable with the original). There are several particular points which I cannot seem to find the answer to
1) is there any evidence that it actually does do so reliably? After all we all know that there is any amount of nonsense files sold as high rez and surely this forum's raison d'etre is testing claims. Sorry if I missed this.
2) is there any evidence as to how it does so?
3) I still can't understand why the (or at least a) correct comparator is not noise shaped 16/96 or 16/88. What's wrong with that. AFAIK MQA offers little or no file size advantage over such a file.
4) where the musical triangle extends about 48KHz (which IIRC it is alleged to albeit not that often) how is this captured by MQA without significant aliasing in the <48Khz range?
You need to read all the threads about MQA,your answers are in them.
 
I was thinking of digitally capturing the second unfold, not only first.
Though I agree, there is nothing there that matlab would not tell.
I was under the impression that ROON does the full unfold, creating the final signal with all the bells and whistles, at a higher sample rate (2x, or even more -- 4x or 8x really would be preferable) which is send to a DAC which has support for those high sample rates. The DAC's filter then would not matter anymore.
 
You need to read all the threads about MQA,your answers are in them.
DMac I have read a load of them. I'm interested in @amirm 's position. It goes without saying that I have the highest regard for him.
I'm aware of findings of mansr achimago and others. If you could point me to this forums official position on those points without troubling @amirm then that would be helpful.
 
If you could point me to this forums official position on those points without troubling @amirm then that would be helpful.
Search this thread using the forum search function for all Amir's recent posts in date order. Also, "official position" is an odd term to use in this case.



JSmith
 
Search this thread using the forum search function for all Amir's recent posts in date order. Also, "official position" is an odd term to use in this case.



JSmith
Sorry, but this sort of thing is just not helping me. The premise of my post is that I cannot find the answer.
edit
I don't know whether that was a reference to this post-
https://www.audiosciencereview.com/...-tidal-to-test-mqa.22549/page-127#post-798595
I can see that that might in a sense answer my first question. But definitely not the others. Even then I'm afraid I'm lost as to what if anything is thought to be known. I'm confused by the file size comparisons and references to "16 bit" and all sort of things.
 
Last edited:
I was under the impression that ROON does the full unfold, creating the final signal with all the bells and whistles, at a higher sample rate (2x, or even more -- 4x or 8x really would be preferable) which is send to a DAC which has support for those high sample rates. The DAC's filter then would not matter anymore.

No, just the core (first) decode- the one up to 88/96. Applies the DSP etc then adds back the MQA flag to be rendered on the DAC.

EDIT- https://help.roonlabs.com/portal/en/kb/articles/roon-x-mqa#Can_I_use_DSP_with_MQA_content_in_Roon
 
I was under the impression that ROON does the full unfold, creating the final signal with all the bells and whistles, at a higher sample rate (2x, or even more -- 4x or 8x really would be preferable) which is send to a DAC which has support for those high sample rates. The DAC's filter then would not matter anymore.

No, as @Jimbob54 said, Roon only does the core decoding (to 88 or 96ksps). That’s why ‘canonically’ it needs a ‘MQA-rendering’ DAC - that will upsample to 192ksps… (Though it is pretty obvious, this rendering DAC can’t do anything a traditional up-sampling DAC with the same filter would not do.)
 
Why exactly are we using a codec that can't handle arbitrary signals?

How different from an impulse response does something need to be to not be an impulse response? Like would the crackle from the Hyper Light Drifter sound track make it choke? https://music.disasterpeace.com/track/vignette-visions
Ha, I can't make this up, Disasterpiece has music on Tidal but the Hyper Light Drifter soundtrack isn't among it.

If a codec can't handle white noise and square waves then it doesn't even cover the capability of the original gameboy, which has a "wave" channel (this one can be modified with more complex wave forms), two square wave channels, and a noise channel. https://bit-shifter.bandcamp.com/track/hexadecimal-genome . Piece of hardware that was obsolete when it was released in 1989, too much for this modern codec.

And how noise like can something be? Does it choke on 80's synth like ELO?

And WHY is it using the audible band to contain this extra data? Are they planning to press it to vinyl and give the DBX concept another shot?

Does this mean the number of (PC-)software full unfolders is zero? That'd really be dissapointing. So tapping the output of a firmware MQA final renderer chip like some XMOS is the only way?

Yes, full unfolds are only allowed to be performed by analog out only devices.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom