MQA is Vaporware

[adamdea]

I don't think Apple has nearly as much influence and power in this area as it once had. If labels want to deliver MQA to Apple rather than Mastered for iTunes, Apple doesn't have a choice. Accept it or stop selling / renting the music.

This raises an interesting point: I have been a little confused by the marketing blurb around MQA- it seems to imply that MQA is the original master, blessed by the artist, rather than being an indistinguishable compressed version of that original.

 

Is there are evidence that the final master is, or is ever likely to be produced in MQA as opposed to 24/96 (or more rarely I imagine 24/192 or DXD). How would this work? Is there editing software working in MQA? PLugins?

 

If not then why would the label ever want to deliver MQA to Apple? To stop them gettign the Crown jewels? (which is of course odd if MQA is the Crown Jewels.)

 

For me this is the most interesting part of this curious portmanteau product: if it is just a way of packaging reduced bitrate 24/96, then no one should really be perturbed by a claim that it is transparent to the original (it's possible although it might manage to fail. Straight 16/96 would do the job.) But it claims to be better (technically) AND to be the original master. I can;t at the moment see how it could be both of those things.

[adamdea]

Read this Bob Stuart interview. You have to parse the words carefully, because after all, he is marketing his product. You can't blame him for that as long as he's not being deceitful. You can then form your own opinion.

 

And yes, you must have enough trust in Stereophile to believe they are not in on the conspiracy.

MQA and Warner: the Real Scoop | Stereophile.com

Clearly there are some cases where work is done to restore a master with some effort to reverse engineer some ADC failings. In the case of the Universal catalogue might be a batch convert from whatver 24 bit masters they have; certainly he has not specifically indicated that details of the originally ADCs (mastering software? plugins?) were provided or compensated for.

 

In neither case is it specifically stated that the original artist has approved the final MQA version.

[adamdea]

@Miska has done the bandwidth analysis. If I recall correctly, the bandwidth required to stream an MQA file is ~ 2x that of a redbook FLAC, and about the same as a 18bit/96KHz FLAC which incorporates all of the data in an MQA file.

 

But regardless... Anyone able to stream MQA could stream a 24/96 file. Bandwidth of this sort is generally not a problem today. The set of people that could stream MQA reliably but not a 24/96 file is of measure zero. I say this bc as someone pointed out in the past, not all player will be able to play a 18/96 FLAC.

Possibly but you can get the same result AFAIK by "bit freezing" the bottom digits of 24/96. If you stop the bottom digits presenting as noise then FLAC will pack them very tightly. Or you can noise shape 16/96 which should work pretty well.

[adamdea]

We already have a very good, open source, widely distributed and well supported  system for producing perceptually indistinguishable, reduced size version of studio masters. It's called 16/44. It is marginally preferable to 256 Khz AAC as it can be further transcoded transparently. 

 

If people want something better than 2 channel 16/44 they would be better off working on some version of surround/HRTF or whatever.

[adamdea]

50 minutes ago, Shadders said:

Hi Fokus,

You have claimed i do not know what i am talking about, and this was not correct.

You have claimed that temporal blur is caused by ringing, and this is not correct.

You have claimed that temporal blur is not dispersion, and this is not correct.

You are trying to find fault where none exists.

Regards,

Shadders.

OK. Deep breath called for. Let's step back a moment.

"Temporal blur" may well be used in different (or even vague and undefinable) senses in some parts of the MQA literature, particularly relating to a general vague claim to undo something or other done wrong in the ADC process BUT

There can be no doubt that the main thrust of the MQA papers and the famous MQA Q and A was to do with targeting an impulse response equal to 10m of air. The critcism of conventional non MQA filters was also aimed at their impulse response and time blur seems to have been equated mainly with the width of the impulse response.

 

Now there may be all kinds of reasons why the MQA position is horseshit but it really does seem as though  it was the pre-and post ringing which amounted to the alleged "time domain" (actually meaning impulse response) characteristics of conventional filters which MQA  was attacking.

 

Of course it is difficult to speak definitively about what temporal blur is or isn''t for the same reason that it is is difficult to speak definitely about the breeding characteristics of unicorns.

 

[adamdea]

empty

[adamdea]

8 minutes ago, Shadders said:

Hi,

The AES paper from 2014 on page 2 defines blur as dispersion.

I will defend inaccurate claims made against myself.

Regards,

Shadders.

I dare say. But "The AES paper" is not the one Werner was referring to- which is by Craven (note not Stuart), and kicked off the minimum phase apodising filter thing (and I think it's referred to in the MQA documentation)

 http://www.aes.org/tmpFiles/JAES/20180125/JAES_V52_3_PG216.pdf
it's all about the impulse responses. But does not use the expression blur.

 

Anyway, whichever way you look at it  impulse response is supposed to be the MQA target (10m of air). It can be very confusing, because we are to a large extent, tilting at windmills. 

[adamdea]

Even if it is dispersion, I fear that it isn't going to help with the the point that the PCM output of a delta sigma adc with a conventional filter will end up  with whatever the (real or imagined) time domain thing is that MQA is supposed to cure.

[adamdea]

1 hour ago, mansr said:

This "10 m of air" talk makes no sense. The speed of sound in air is independent of frequency, so there is no dispersion. High frequencies are attenuated more than low ones. For 10 m of air at 50% relative humidity, the attenuation is 0.047 dB at 1 kHz, 1.6 dB at 10 kHz, 5.2 dB at 20 kHz, and 33 dB at 100 kHz (http://www.kayelaby.npl.co.uk/general_physics/2_4/2_4_1.html). I'd like to see explained how the MQA filters are derived from this.

Obviously it's pretty much horseshit, but the claim never was about group delay or dispersion in any specific technical sense, (even if, like many sciencey-sounding things, the term dispersion is borrowed here or there) it was about the impulse response of 10m of air. It's enough of a straw man as it is not to need to be dressed up further as a straw man. 

[adamdea]

7 minutes ago, Shadders said:

H

Again, the impulse response does not prove the dispersion effect.

That is, a wide or slim impulse response does not mean a dispersive, or non-dispersive effect of a filter.

The phase of the filter determines whether the signal undergoes a dispersive effect.

As long as the filter phase is linear across the required signal bandwidth then there is no dispersion to those frequencies of interest.

Regards,

Shadders.

Shadders please,

you are arguing yourself round in a circle.The property which MQA is targeting is not dispersion in the sense you are using it. QED.

The MQA argument is about some alleged time property of filters shown here in the slide attached by manisandher. Whatever it is, a linear phase filter has it. It is not dispersion in the sense you use it. That may mean that somewhere along the line the MQA literature has used the word in one sense and then again in another. No surprises there  

[adamdea]

By the way, this 1997 paper by dcs also uses the heading "dispersion" to refer to the effects of ringing in a linear phase filter. (very similar to the  MQA graphic in my previous post). 

http://www.cirlinca.com/include/aes97ny.pdf

 

Obviously it is irrelevant whether this is a correct usage of the engineering term dispersion. It is the phenomenon from which we are supposed to need saving.

[adamdea]

17 minutes ago, Shadders said:

Hi,

If you examine page 4, it states "Energy Dispersion at Different Sample Rates  - Figure 6 shows the energy associated with the transient response".

This plot is a filter response to an impulse - transient. Again - it does not show that the audio signals in the relevant bandwidth undergo dispersion.

As long as the audio signal is in the linear phase region, then there is NO dispersion. An impulse contains ALL frequencies up to infinity (there are practical limitations). The figures in the paper you have referenced do NOT show dispersion occurring at audio frequencies, just the difference in impulse response.

Regards,

Shadders.

P.S. There is a cable company that in their demonstration (youtube) stated that they have "discovered" a new form of distortion/noise. This was 2009 - still no proof, or rewriting of the text books.

As I understand it the energy which is or might be redistributed in time (I am not going down the road of using the word dispersion) is in the transition band of the filter. Of course, for a sensible anti alias or anti imaging filter  it is therefore energy at inaudible frequencies and that energy can only arise in the first place to the extent that it is in the signal. It is therefore grossly exaggerated by a notional dirac input 

[adamdea]

I should point out that this is very much the subject of my post over on Pinkfish (itself referring to an article by Archimago) which has been made the subject of another thread on this forum. I have also challenged Jim Austin on the Stereophile forum for his use of impulse responses to (purport to) analyse  the "time domain" behaviour of MQA and non MQA filters. 

O tempora o mores.

[adamdea]

14 minutes ago, esldude said:

That is my understanding of what MQA is claiming as much as one can understand their tales mixing real with fantasy.  That long steep filters disperse energy in the transition band over time more than shallow short filters.  Then you make up the tale about air.  Air is absorbing frequencies, but if you graph the amplitude it can look like a filter dispersing energy.  

As I understand it air is somewhat elastic and behaves like a filter (not linear phase or even minimum phase). Interestingly this has the effect that for all practical purposes there is no such thing as sound in air over a few hundred kHz. (can't remember the exact figure but somewhere before 1 MHz you get to the point where 1cm of air attenuates at 90dB or more). Anything you find above a certain level in a recording is an electronic artefact.

You can repeat something like that argument above an order of magnitude lower for most mics (though they might be minimum phase)

[adamdea]

9 minutes ago, Shadders said:

Hi,

Yes - this was the statement behind my original post many pages back - the filter on the front end of the ADC has a cut off near the 1MHz frequency - and that ensures that any statement of "blur" is inconsequential - it does not happen with ADC's from the mid 1990's onwards that are Sigma-Delta.

In fact, a filter with a cut off at or near 96kHz will not cause "blur" since there are no audible signals above 50kHz. The filter will be linear phase at this region and above (to a point - filter spec dependent).

Regards,

Shadders.

Yes. The issue is the effect or otherwise of the anti alias filters and anti imaging filters required to get to and from the distribution sample rate. These are unproven and dubious at 16/44 , entirely implausible at x/96 and comically absurd at x/192.   But the issue is the effect on the filter of the small to zero  amount of program material in the transition band of a sharpish LF filter ie around 20-24Khz for 16/44 40-55 khz for x/96 and maybe 80-120 khz for x/192 

[adamdea]

15 minutes ago, Lee Scoggins said:

 

That is not what I said. I said, "In other articles I have read like in Stereophile, Bob claims it is around 23 bits so I think I need to see if there is a way to reconcile and that we are not taking him out of context."

 

So I am saying we need to investigate why the discrepancy exists and make sure it is not taken out of context.  I am clearly open to the possibility that Stuart meant it on the 17 bits comment.

It is confusing of course- but that is because it is meant to be. However if you use some sort of sense you must be able to see that a 24/48 container can only contain so much information, however cleverly it is stacked. The general gist is to have undecoded MQA read using the top 15-16 bits. That leaves the bottom 7-8 bits to represent everything over 22/24 Khz. Now there remains some issue as to how efficiently this can all work, but that's the size of it. And it supposedly works because the target area is not a "rectangle" 24 bits deep and 48 khz wide but a triangle (or something like it) say 16 bits deep at one end and zero somewhere (maybe at 48 khz or maybe a bit higher). That's how it works -it simply can't  give 23 bit resolution everywhere and probably not anywhere.

[adamdea]

22 minutes ago, Shadders said:

Hi,

For the 44.1kHz/48kHz sample rate, then possibly the energy near 20kHz is an issue. For 96kHz, the energy at 40kHz+ is approx -80dB to -100dB down from full scale. Two issues - it is too low to hear, and no one can hear in this range. (i am examining Hifi News Hi-Res download plots).

If the ADC is a Sigma-Delta - then none of this is an issue. Only an issue for recordings at the lower sample rates.

Brian Lucey said he uses the Pacific Microsonics ADC at 44.1kHz, i have Depeche Mode's - Spirit album - everything is clear and precise - a good recording like most of Depeche Mode's.

Regards,

Shadders.

I think we are probably in agreement provided it is understood that all pcm recordings (and probably almost all dsd recordings) are at "a lower sample rate". The immediate  output of the adc's modulator is really neither here nor there (where are the multibit 14Mhz sample rate recordings?).  

The amusing thing about the HFNRR plots is that  PM actually notes in places that most piano recordings have no information high enough to warrant hi rez. Does that stop people buying them?

 

The amusing point is that I can discern no difference in the enthusiam expressed for  hi rez recordings with payload in the range that might  make a difference from recordings with nothing there (most old recordings); same with sighted filter preferences. What does that tell us?

[adamdea]

1 minute ago, esldude said:

That any quality difference is in all the other aspects of good recording/mastering practice which doesn't require hirez.  And/or the idea of there being more in hirez makes people hear more regardless of whether there even is more there.  Like old recordings released in hirez that are sourced from 40-50 year old reel tape.  

mais oui. It has always tickled me that the terrible flaw with Meyer and Moran was that unbeknownst to the participants (who IIRC were asked to bring their favourite sacds) some of them were upsampled 16/44. 

[adamdea]

1 hour ago, FredericV said:

The best part is that those researching MQA and it's filters, came to an even better filter config than what MQA is offering:

http://archimago.blogspot.be/2018/01/musings-more-fun-with-digital-filters.html

The intermediate phase filter as proposed by Archimago does not suffer from the artefacts which MQA suffers from, including severe aliasing and phase errors.

Whilst I remain unpersuaded that there is any sound case for anything other than linear phase., I welcome Archi's work as ever. It's interesting isn't it that the cheerleaders for minimum phase (ie the same publications as now cheer-lead for MQA) did not in all their researches point out the time domain problems of minimum phase. 

[adamdea]

5 minutes ago, manisandher said:

 

 

Maybe I'm really dense, but I'm find this confusing. Aliasing occurs during analogue-to-digital conversion, and not during digital-to-analogue conversion, and yet he talks about using an anti-alias filter in a DAC. Are the terms 'aliasing' and 'imaging' really so interchangeable?

 

Happy to be put right.

 

Mani.

Spot on. The "aliasing region" graph suggests that he must be talking about A/D or downsampling.

[adamdea]

Another day another rabbit hole.

[adamdea]

35 minutes ago, Lee Scoggins said:

 

I recently visited Dave Wilson in Provo and he also showed me research on the ear’s ability to hear timing differences which he used to design the Chronosonic flagship. He was excited about it because this ability of the ear to hear tiny timing differences holds true into one’s 70s.

Well who knows what he meant? It is my understanding that age related hearing loss is not confined to frequency range but also affects certain temporal effects including gap recognition for example. These temporal phenomena are AIUT highly relevant to speech comprehension.  It's difficult to engage with the entire spiel about timing differences/time resolution/time blur because none of it appears directly to map to any genuine measured or recognised effect.

 

Our old friend the 7us resolution is of course about inter aural time differences and has nothing to do with anti alias filters. I have tried looking into what time resolution might mean (within one audio channel) and cannot see that there is any single fixed measure. But so far as I can recall they all seem to be in the millisecond range.

 

Which is just as well or vinyl would sound unbearable 

[adamdea]

[  ]

[adamdea]

14 hours ago, Fokus said:

 

Interesting.

 

Let’s call an attorney with a bit of footing in DSP and audio.

 

I know such a person ...

 

 

Interesting. No specialist knowledge here but....

You may recall that In the UK back in 2007 the Advertising Standards Authority rejected a complaint against Nokia's implicit claim that 128 kbps AAC was CD quality, interpreting the meaning as being that most people could not tell the difference. They concluded that most people could indeed not tell the difference, based on the ISO survey, Report on the MPEG-2 AAC Stereo Verification Tests. That ruling strikes me as about as favourable to a sound quality claim as one could imagine. My guess is that the most hardened proponents of perceptual codecs amongst us would probably set the bar a bit higher.

 

But of course "lossless" doesn't strictly mean "indistinguishable" but "identical". Would the ASA interpret lossless as meaning "mathematically identical"? It would seem odd to interpret lossless as meaning "perceptually indistinguishable", since that would accord no meaning to the distinction between "lossy" and "lossless". The problem is IMHO that the number of people likely to understand the issues is pretty small. That causes a problem when it comes to identifying the meaning of the expression to the average consumer. I would like to think that the answer lay in identifying what the expression would mean  to someone a) interested in hi resolution audio and b) who could be bothered to think hard about it. But that's you, me and uncle fred, right?

 

So it's difficult to see what the outcome would be, but it would be amusing to see how MQA justified the claim if someone made the complaint.

 

Neverthless, it's not 100% clear whether they really have made the claim that MQA is lossless in their advertising. If they have then let's see. 

 

[adamdea]

10 minutes ago, adamdea said:

deleted duplicate

 

By the way, while I'm at it "Tabula rosa" has no meaning I'm aware of. If it meant anything it would be "pink slate".  Tabula rasa on the other hand...