Tidal ditch MQA

[Jim Audiomisc]

Excellent. Really pleased to hear that as I’m now a Tidal subscriber and don’t have (or want) the decoding technology. There is still a lot of stuff that identifies as MQA on there and of differing resolutions, though I guess I only get to hear them as lossy watermarked files. Good riddance to bad technology.

The snag being that - like HDCD - some future re-issues may still have been messed about with MQA along the way to the customer. But now come with the flags lost and the added garbage re-arranged! I have more than one CD which has no sign of HDCD in the blurb - but signs of it being affected/afflicted in the audio data!

 

[Jim Audiomisc]

MQA certainly isn't foo and those who suggest otherwise need to get better equipment.

It certainly does have issues though.

Depends on your defn of "Foo". If your kit or your ears likes the changed filter or the added HF 'noise' fine. But it is something you can do without MQA.

 

[Jim Audiomisc]

PS The problem with MQA was always the locked proprietary system, the lies and grift. A more honest open source solution could have some use.

The "secret sauce" was the killer from my POV. But in practice the reality is that if people feel they prefer 'high rez' then they can simply get hi rez LPCM. 20 years ago when we were limited in sample rate and/or sample depth magic origami might have had some use. But HDCD didn't thrive. Same reason as MQA. Either you keep the full info or you don't. Either you need it, or you don't. You can't have the bun *and* the sixpence.

 

[Jim Audiomisc]

The codecs like flac are supposed to be lossless so are still high res. I don't know enough about MQA to say whether it's lossless at 24/192 but I thought that was the point. What the compression rate is is another matter, there must be a mathematical limit to what compression rate can still be lossless (although every file will be different as I think it relies on not repeating the same chain of 1s and 0s but I could be completely wrong about that).

The trick here is the idea of "audibly* 'lossless' in the sense that you can't tell it has been messed about with. The process in information theory terms is not lossless. You can't get a bit perfect recovery of what went in to the encoder. All you can do is hope/claim it sound the same.

 

[Jim Audiomisc]

Just had this email from Tidal....

"On July 24, 2024, we’re replacing the music in TIDAL’s MQA catalog with FLAC versions. I

My "Apples != Oranges" detector just buzzed at that.

You can certainly convert MQA LPCM to flac and then expand that back into a perfect copy of the original MQA LPCM series of samples. No info is lost or altered by FLAC. So are they simply encapuslating the MQA-affected items into FLAC?

 

[John Phillips]

Depends on your defn of "Foo". If your kit or your ears likes the changed filter or the added HF 'noise' fine. But it is something you can do without MQA.

Jim, I wonder if the selectable reconstruction filters in the MQA DAC are an interesting, separable part of the MQA "package". To be used without the "folding". Paired with the particular sampling filter used by the studio with some advantages over the usual sampling/reconstruction (but maybe some disadvantages too).

I think Shannon-Nyquist sampling and Whittaker reconstruction are not the only ways to (in theory if not in engineering practice) get mathematically perfect analogue audio from studio (sampling) to home (reconstruction).

AIUI (from a dinner a few years ago with Jamie Angus - Prof. of audio technology at U. of Salford) there are alternative sampling filters and reconstruction filters based on beta-spline impulse responses. These have impulse responses that drop towards zero much faster than sinc. They can be much shorter for the same accuracy, but AIUI the DAC filter has to be specifically paired with the specific sampling filter used.

It was just a dinner conversation. There were no details (and in discussing audio technology with Jamie I felt that I had brought a knife to a gunfight - so beware of me misunderstanding this). However Jamie has published this JAES paper: https://www.academia.edu/73065157/Modern_Sampling_A_Tutorial.

So might the multiple MQA filters, usually described in conventional terms as "leaky", be an interesting separable element of the MQA package? I do wonder if what Lenbook is proposing to do at MQA labs might separate out the elements of MQA. Maybe FOQUS = sampling; QRONO = reconstruction? Just speculation, of course.

 

[Jim Audiomisc]

Jim, I wonder if the selectable reconstruction filters in the MQA DAC are an interesting, separable part of the MQA "package". To be used without the "folding". Paired with the particular sampling filter used by the studio with some advantages over the usual sampling/reconstruction (but maybe some disadvantages too).

I think Shannon-Nyquist sampling and Whittaker reconstruction are not the only ways to (in theory if not in engineering practice) get mathematically perfect analogue audio from studio (sampling) to home (reconstruction).

AIUI (from a dinner a few years ago with Jamie Angus - Prof. of audio technology at U. of Salford) there are alternative sampling filters and reconstruction filters based on beta-spline impulse responses. These have impulse responses that drop towards zero much faster than sinc. They can be much shorter for the same accuracy, but AIUI the DAC filter has to be specifically paired with the specific sampling filter used.

It was just a dinner conversation. There were no details (and in discussing audio technology with Jamie I felt that I had brought a knife to a gunfight - so beware of me misunderstanding this). However Jamie has published this JAES paper: https://www.academia.edu/73065157/Modern_Sampling_A_Tutorial.

So might the multiple MQA filters, usually described in conventional terms as "leaky", be an interesting separable element of the MQA package? I do wonder if what Lenbook is proposing to do at MQA labs might separate out the elements of MQA. Maybe FOQUS = sampling; QRONO = reconstruction? Just speculation, of course.

Two basic points:

1) How you *calculate* (i.e. convert) a series of sample values into an analogue output shape is up to you. But from Shannon and Nyquist the *defined* shape you should expect is as the standard mathematical approach is defined. i.e. if you use a 'non standard' approach to the calculation that 'fills in between the dots' then you should get the same analogue output pattern.

2) the reconstruction filter form MQA is nominally known. It is what I got out when I put a sigle non-zero sample in the middle of a sequence of zeros and played that out at the 'tail' of an MQA recording using Meridian's DAC.

The way MQA *encoding* may futz about with the input is inbuilt. Playing 'pretty shapes' at the output generates something else on the basis of the MQA argument that essentially says it will "sound better than vanilla*.

So, its just like some DAC/player makers decided to do a 'non Nyquist' output filter/process because they think it "sounds better". But with a coating of secret sauce to make it harder to find out what they did. And of course some people may like some added HF noise. Matter of how much sauce you like on your dinner.

Pesonally, I find the approach of using noise shaping makes far more sense. It gets more resolution into the more audible regions at the expense of a tad higher HF noise at low levels. Of course, that isn't patented or covered in sauce.

 

[davidsrsb]

You can certainly convert MQA LPCM to flac and then expand that back into a perfect copy of the original MQA LPCM series of samples. No info is lost or altered by FLAC. So are they simply encapuslating the MQA-affected items into FLAC?

I doubt FLAC could compress a MQA file much, the data encoded into the LSBs must look like noise. For the same reason FLAC does not work well for 24 bits depth. You cannot compress random data.

 

[colasblue]

Depends on your defn of "Foo". If your kit or your ears likes the changed filter or the added HF 'noise' fine. But it is something you can do without MQA.

You say that as though that's all it does. It isn't. It does genuinely encode more information into a smaller amount of data, albeit in a slightly lossy way. I'll take MQA over standard 44.1/16 95% of the time. There are clearly a few works it seems unsuited to for some reason.

I agree that from a technical point of view it's pretty much a concept without purpose if you have unlimited storage and bandwidth and was a dead horse from the outset if the objective was to create a higher quality physical media CD like platform.

The major point that doesn't seem to have been grasped by its detractors is that at least some of the music industry was buying into it as a mechanism to release a higher quality digital product without going quite all the way to the full studio master. A thing that they have mostly resisted to the nth degree in the CD years. The consequence is that there is now quite a lot of Hi-Res stuff out there on Tidal and other platforms on MQA and other Hi-Res formats at very reasonable cost which simply wasn't available without either paying a fortune to shysters such as HDtracks or in most cases at all just a few years ago.

The antagonists should consider very carefully what they might have shot down here. If all the MQA tracks revert to 44.1/16 and the push for higher resolution formats loses momentum the audiophile community will have lost a very great deal in my view.

It's never been about what can be done with the entertainment industry giants (at least not for the last 30 or 40 years). It's about what they will allow you to have. Never forget that. Just look at the home theatre world to see the full raft of licensing bullshit and restricted functionality. Faff free SACD ripper anybody? Full Matrix audio without a dedicated AV processor?

 

[Jim Audiomisc]

I doubt FLAC could compress a MQA file much, the data encoded into the LSBs must look like noise. For the same reason FLAC does not work well for 24 bits depth. You cannot compress random data.

It is simple enough for people to check that by converting an MQA encoded wave file to flac. Although the result will be limited by the added 'noise' due to errors in the MQA being treated as real info.

 

[Meursault]

It highlights the need for continuing support from the streamer manufacturers.
There’s some concern amongst Arcam users, particularly of SA30 units that Arcam haven’t confirmed that they intend to switch on streaming above 44.1 FLAC. It supports full res files in MQA but only seems to get to 44.1 on a straight FLAC file and obviously those nice MQA files disappear in July. The problem seems to be at the Arcam end and not the Tidal end. I think it might affect the ST60 as well. These units are still being sold by dealers and are only three years old at a maximum.
Arcam might yet update the service but it’s a clear warning that any expensive streamer is a gamble, particularly where there are a number of companies involved as in this case with Tidal and Arcam.

 

[Jim Audiomisc]

You say that as though that's all it does. It isn't. It does genuinely encode more information into a smaller amount of data, albeit in a slightly lossy way.

Not sure what meaning is being given to "more". How do you seperate the wheat from the chaff in the MQA? Real info is lost on the argument that people won't miss it. The problem here is the way two things are being conflated. One is a change in sample rate or size. The other is the added 'approximation' (polite term for it) and the loss.

Apples meet onions.

 

[colasblue]

That's a comment that could clearly only come from somebody who's never auditioned a range of MQA material against the 44.1/16 or Vinyl equivalents!

 

[adamdea]

That's a comment that could clearly only come from somebody who's never auditioned a range of MQA material against the 44.1/16 or Vinyl equivalents!

Why on earth would you say that? There is a range of opinion on the subject of MQA vs non-MQA from always sounds better to always sounds worse. Quite frankly if it always sounded different that would be a remarkable thing.
As for comparison with vinyl, you might as well compare it with peanut butter.

 

[colasblue]

I make my comment based on auditioning with a very good system.

I find it very difficult to believe that anybody who has actually taken the trouble to audition MQA seriously before engaging in the debate could be entirely negative about it. Certainly a minority of MQA material isn't at all to my taste compared with other formats, but it is a minority.

Don't forget that the people behind it aren't exactly idiots. They're amongst the best digital audio engineers on the planet!

 

[adamdea]

I would expect MQA if applied properly and not messed up in application to sound pretty much identical once unfolded to 24/96. But not because I think it's any good just because I doubt that 24/96 adds anything practcially useful to 16/44, so whilst I consider MQA to be inferior, I doubt it really matters. That said I have no confidence that something close-source and uninspectable will in practice be aplied properly and not messed up in application. There are anough examples of ordinary hi rez that turns out not to be. Either way I did listen to it and heard nothing that interested me.

AFAIK MQA does not fit in more information (audible or otherwise) than noise shaped 16/96 nor does it pack it into a smaller space and has several disadvantages. On the other hand if not unfolded then its basically noise shaped 13/44 in a much bigger package than 16/44.
As for the idiocy or otherwise of the people behind MQA, I shrug my shoulders.

 

[colasblue]

Well if you honestly think 44.1/16 sounds as good as 96/24 then I think you'd have the whole pro industry calling you out as deaf! The lie that CD gives "perfect" sound must surely be well and truly dead by now mustn't it?

Tidal does at least have a few 96/24 and 192/24 FLAC offerings and I'm 100% sure they better my redbook CD's.

Mind you I don't think that applies to all recordings. I doubt whether a lot of modern compressed CD's really ever use more than about 8 bits!

Didn't Archimago do a 96/16 vs 96/24 comparison a few years back which pretty much proved the extra bits weren't really audible to anybody provided the recording level was set correctly? That suggests the difference is in the 96 rather than the 24.

And of course if you listen to MQA without it being properly unfolded then you get a compromised 13 bit output, but why would anybody running it on Tidal get that? A PC can unfold it if your DAC isn't capable and transmit on the processed version and I assume that also works for other devices.

Tidal wold have been a self defeating audiophile platform if they hadn't thought about that surely since most DACs in use don't have native MQA decoding.

 

[davidsrsb]

And of course if you listen to MQA without it being properly unfolded then you get a compromised 13 bit output, but why would anybody running it on Tidal get that? A PC can unfold it if your DAC isn't capable and transmit on the processed version and I assume that also works for other devices.

Properly decoded, the up to 22 kHz band is still 13 bits, but the encoded HF noise below that is now removed and there is also the >22 kHz ultrasonics restored.
Why MQA was a bad idea for 44.1/16

 

[Jim Audiomisc]

I make my comment based on auditioning with a very good system.

I find it very difficult to believe that anybody who has actually taken the trouble to audition MQA seriously before engaging in the debate could be entirely negative about it. Certainly a minority of MQA material isn't at all to my taste compared with other formats, but it is a minority.

Don't forget that the people behind it aren't exactly idiots. They're amongst the best digital audio engineers on the planet!

Well, to my ears as well as via my analysis, it does nothing useful. However it measurably adds distortion and noise. Given that it also requires extra payment for its 'benefits' it makes no sense to me. if someone wants genuine 'high rez' then choose that. These days it is easy to get and use. However if someone can't hear - or care about - the added noise and anharmonic HF distortions - or like them - Fair enough. But again it would be easy enough to add them as 'effects' without paying someone a fee.

 

[Jim Audiomisc]

Why MQA was a bad idea for 44.1/16

In essence, the same loop we went round with HDCD. Curiously, IIRC in that case some people claimed to enjoy the 'filter switching' factor as it was said to add back some > 22kHz info. Yet later on it became clear that HDCD simply didn't do that aspect of what was claimed in the patents. And I've never found *any* HDCD track that has the flags for it.