Article: MQA: A Review of controversies, concerns, and cautions

[mitchco]

1 hour ago, John_Atkinson said:

 

In my interviews with Bob Stuart, he has told me that the intention is that the analog signal output by the consumer's D/A converter is identical to that output by the mike preamps (in a purist recording) or the mixing console (in a conventional recording).  That the A/D conversion, transmission, storage and D/A conversion be transparent, other than there being an ultrasonic rolloff equivalent to a signal path of a few feet in air. This has been written about in the magazine.

 

 

It is MQA's time-domain behavior, the claimed 'temporal deblurring," that is fundamental to Stuart's explanation. And as I have said before,  that is what I will be investigating in a future article. Up to now, what we have investigated and written about is MQA's  frequency-domain behavior and MQA's societal and commercial aspects. I don't comprehend why you and other posters to CA don't regard that as journalism, investigative or otherwise. Unless you are confusing Stereophile with another magazine?

 

John Atkinson

Editor, Stereophile

 

 

 

Hi John, as an ex 10 year recording/mixing engineer, DSP aficionado, and software engineer by day, I wonder what is the definition of deblurring or temporal deblurring?

 

In the world of DSP,  deblurring is usually associated with digital image restoration that can sharpen an out of focus image for the eyes. The only time I have seen it applied in the audio world is when using digital loudspeaker and room correction DSP. This is where the digital signal of the music file is convolved with a custom digital correction filter, so the music arriving at ones ears matches as closely as possible to the music encoded in the digital media file. There are several articles written about this here on CA.

 

Technically, we are talking 65,536 filters taps which is 2000 more powerful than MQA's 32 taps. Also, digital loudspeaker and room correction DSP corrects the entire signal path (i.e. playback chain) to ones ears, both in the frequency and time domain. As far as I know, currently today, this is the only way to achieve fully accurate sound reproduction where the frequency response is flat along with flat phase response and group delay to ones ears. Note this also corrects a loudspeakers timing response, as some 95% of loudspeakers are not time aligned. What would be the point of MQA deblurring if it can't deblur the loudspeaker? Makes no sense.

 

John, I am a real fan of your measurements and wanted to personally thank you for this, especially your talks on loudspeaker measurements. This is why I am puzzled, and obviously others, as to why you would even entertain such a thing?

 

Kind regards, Mitch

[james45974]

1 hour ago, John_Atkinson said:

In my interviews with Bob Stuart, he has told me that the intention is that the analog signal output by the consumer's D/A converter is identical to that output by the mike preamps (in a purist recording) or the mixing console (in a conventional recording).  That the A/D conversion, transmission, storage and D/A conversion be transparent, other than there being an ultrasonic rolloff equivalent to a signal path of a few feet in air. This has been written about in the magazine.

Why when I read things like this does the word smug always come to mind?

 

It seems that those investigating this claim have shown that what reaches the consumer is an adulterated facsimile of the original!

[sullis02]

1 hour ago, wdw said:

Not attempting to defend anyone but I do not need to degrade someone to make my points.  There are a number of N. Young fans on this site although I am not one of them.  I enjoyed Tellig although never took it as anything other than entertainment.

 

I'm sure you, like any adult, can distinguish the merit of an artist's work  (I like Neil's)  from his opinions on, say, technology. Right?

 

As for Tellig, how do you know he wouldn't find your opinion of his work 'degrading' too?

[wdw]

3 minutes ago, sullis02 said:

 

I'm sure you, like any adult, can distinguish the merit of an artist's work  (I like Neil's)  from his opinions on, say, technology. Right?

 

As for Tellig, how do you know he wouldn't find your opinion of his work 'degrading' too?

OK, fair comment,  Truce?

[sullis02]

1 hour ago, John_Atkinson said:

 

In my interviews with Bob Stuart, he has told me that the intention is that the analog signal output by the consumer's D/A converter is identical to that output by the mike preamps (in a purist recording) or the mixing console (in a conventional recording).  That the A/D conversion, transmission, storage and D/A conversion be transparent, other than there being an ultrasonic rolloff equivalent to a signal path of a few feet in air. This has been written about in the magazine.

 

And it always makes me ask: does Bob mean that he's going to somehow enforce the use of original mixdown master sources, rather than, say, versions that have been subsequently 'remastered' to reduce their dynamic range?   

 

IOW will it be 'audiophile quality' all down the line, or will it be like HDtracks, who must accept whatever mastering the record companies supply?

 

That was the *only* interesting aspect of MQA, to me.

 

[sullis02]

19 minutes ago, mitchco said:

 

Hi John, as an ex 10 year recording/mixing engineer, DSP aficionado, and software engineer by day, I wonder what is the definition of deblurring or temporal deblurring?

 

I have a premonition that the name 'Milind Kunchur' will be invoked at some point.

 

I hope JA consults other experts like JJ Johnston, if so.

[Doug Schneider]

1 hour ago, John_Atkinson said:

It is MQA's time-domain behavior, the claimed 'temporal deblurring," that is fundamental to Stuart's explanation.

 

Indeed, this is fundamental because the compression over typical networks is useless these days. But with that in mind, MQA is 3+ years old and this should have been questioned from the outset -- and it should be them who is made to prove their claim, not the other way around. Paul Miller was, I believe, the first print journalist through his "Hi-Fi News and Record Review" to take a critical look at MQA and all mentions by him so far have questioned the validity of the claim. On the other hand, all the early raves by other journalists attributed what they heard to improved time-domain performance. Don't you think that's putting the cart before the horse? If there is no time-domain improvement, what then? What accounted for the revolutionary improvements they supposedly heard?

 

Doug

SoundStage!

[miguelito]

19 minutes ago, mitchco said:

As far as I know, currently today, this is the only way to achieve fully accurate sound reproduction where the frequency response is flat along with flat phase response and group delay to ones ears. Note this also corrects a loudspeakers timing response, as some 95% of loudspeakers are not time aligned. What would be the point of MQA deblurring if it can't deblur the loudspeaker? Makes no sense.

As you said... Consider what MQA does not account for:

1- DAC's analog stage

2- Pre/Amplifier

3- Speakers

4- Room

 

Moreover, given the design itself, and the restrictions to hardware rendering (ie upsampling with MQA-specified filters), MQA's design explicitly prevents you from applying DSP that would correct for the above.  All in the name of licensing revenues for MQA.

 

I do think people are entitled to get paid for their inventions. This however is a hostage situation sold on false pretenses in my opinion.

[miguelito]

I am still confused by the "temporal deblurring" feature touted around... My interpretation is that the use of the much softer rolloff upsampling filters results in much milder phase shifts at high frequencies, which means that a harmonic (that would have a bunch of frequencies) has the signal arriving in a narrower time interval than if some of the frequencies were phase shifted (phase shifts in the frequency domain are time offsets in the time domain). Is this what this is all about? And if so would this not be incredibly imperceptible?

 

 

@Archimago or @mansr - Can you comment on this? 

[miguelito]

One more thing: I would love to see a transfer function for MQA vs the original file, if at all possible. I suspect some of the sound differences are, basically, eq. Listening to Keith Jarrett's Koln Concert, I get the distinct impression that the MQA version has been eq'd compared to the original high res I have (24/96 from HDTracks).

[Dr Tone]

5 minutes ago, miguelito said:

One more thing: I would love to see a transfer function for MQA vs the original file, if at all possible. I suspect some of the sound differences are, basically, eq. Listening to Keith Jarrett's Koln Concert, I get the distinct impression that the MQA version has been eq'd compared to the original high res I have (24/96 from HDTracks).

 

I thought some analysis showed around a .5 db increase in volume with MQA, same master.  I think it was in the big MQA thread on the Roon forums.  That would be enough for me to think it was better, my silly brain.

[Doug Schneider]

This is actually a very good video, even if you start at 8:00. Right off the bat, he talks about having his eyebrows raised and being skeptical about the "claims." This idea of being skeptical/critical or whatever you want to call it seems to have been completely lost on some members of the press. Also, the topic of DRM is front-row and center here. What the heck happened three years ago? Whatever it was that had people drinking the Kool-Aid is a mystery to me. It's guys like these guys who will probably save the music world. It's a boring video, but give it a try... What's more, CA's "MQA is Vaporware" is in there!...Doug

 

 

 

 

[WiWavelength]

15 minutes ago, Dr Tone said:

 

I thought some analysis showed around a .5 db increase in volume with MQA, same master.  I think it was in the big MQA thread on the Roon forums.  That would be enough for me to think it was better, my silly brain.

 

Yes, about a year ago, I made some bit perfect 30 second captures of a few Tidal streams, both HiFi and MQA, then ran those captures through the foobar2000 DR Meter plug in to compare peak and RMS level metrics.  You can find those measurements in the Roon Community forums in and around this post:

 

https://community.roonlabs.com/t/mqa-general-discussion/8204/1235?u=wiwavelength

 

In summary, some masterings are the same, others are level shifted, others are different and dynamically compromised or dynamically improved.  All in all, a mixed bag, as expected.

 

AJ

[WiWavelength]

36 minutes ago, miguelito said:

I am still confused by the "temporal deblurring" feature touted around... My interpretation is that the use of the much softer rolloff upsampling filters...

 

Per my understanding, the "temporal deblurring" additionally or exclusively refers to pre processing as part of MQA encoding.

 

MQA has claimed to research the provenance of recordings -- namely, to determine the ADCs used in digital recording or mastering.  The pre processing then could entail reversing phase shifts from analog brickwall filters in early ADCs and/or apodizing out Nyquist frequency ringing from linear phase digital filters in oversampling ADCs.

 

Whether those temporal characteristics at high frequencies are audible, whether reversing/removing them is a desirable practice, and whether ADC provenance of most commercial recordings can be adequately defined are debatable, of course.

 

AJ

[mansr]

57 minutes ago, miguelito said:

I am still confused by the "temporal deblurring" feature touted around... My interpretation is that the use of the much softer rolloff upsampling filters results in much milder phase shifts at high frequencies, which means that a harmonic (that would have a bunch of frequencies) has the signal arriving in a narrower time interval than if some of the frequencies were phase shifted (phase shifts in the frequency domain are time offsets in the time domain). Is this what this is all about? And if so would this not be incredibly imperceptible?

 

 

@Archimago or @mansr - Can you comment on this? 

Linear phase filters have no phase distortion. They do have the (for no good reason) dreaded pre-ringing, which is a) not an issue, and b) nonexistent at 192 kHz sample rate. MQA is solution to a non-problem.

[rickca]

8 hours ago, John_Atkinson said:

I was intending to examine the deblurring.

Were your readers not interested in this two years ago?

[Archimago]

2 hours ago, miguelito said:

I am still confused by the "temporal deblurring" feature touted around... My interpretation is that the use of the much softer rolloff upsampling filters results in much milder phase shifts at high frequencies, which means that a harmonic (that would have a bunch of frequencies) has the signal arriving in a narrower time interval than if some of the frequencies were phase shifted (phase shifts in the frequency domain are time offsets in the time domain). Is this what this is all about? And if so would this not be incredibly imperceptible?

 

 

@Archimago or @mansr - Can you comment on this? 

 

Until MQA/Bob Stuart clearly tells us what they/he's talking about, we're left speculating.

 

- Is it truly a DSP process separate from the "encapsulation"/"origami" that can be applied to standard resolution or hi-res PCM?

 

- Does it require impulse response data from all the ADCs, DSP processing used in the whole production chain, for each track that was affected, to work optimally? (ie. maybe it analyzes all the potential group delays and applies a "correction" filter to the track? One obvious issue with this is that in a multitrack recording, this also implies a need to remix.)

 

- If there is no data about the equipment/process in the production chain, what does it do in a "batch" processing scenario? Important since it looks like the majority of the thousands of albums were not remixed and were not specially "authenticated"/supervised by the original artist/audio engineer.

 

- Are they just talking about their minimum phase, slow-roll playback filters applied based on the whim of the "authentication" entity... eg. "I think for this 192kHz file, filter #4 sounded good, so let's go with that", and there was really nothing special?

 

We await answers from MQA and/or Stereophile if they can truly report on the heart of the matter.
 

Remember folks, don't hold one's breath on this since it has been 3 years! We might never get an answer because it's quite possible that it was never anything of substance to begin with .

 

Even if there seems to be some interesting algorithm at play, it's important to keep in mind @mitchco's reminder that this cannot be applied to your particular setup because whatever it's doing lacks the potency of true speaker/room digital correction. 

[Archimago]

9 hours ago, Priaptor said:

Well hopefully the following is NOT the future. 

 

One of my favorite recordings of all time is the redbook version of Radka Toneff & Dobroogsz "Fairytales" which just got Stereophile's Recording of the Month "award" (for whatever that is worth) as a new Original Master Edition (MQA).  Whether one likes the redbook recording is up to them; I love it; my concern is the method of remastering chosen for this amazing album.

 

I guess when it becomes available we can all compare it to the original

 

Yes Priaptor,

 

A beautiful album and recording. Just checked and my CD has a DR13 value. Will be very interesting to see if this gets altered by the "Original Master Edition".

 

[miguelito]

3 hours ago, Archimago said:

- Does it require impulse response data from all the ADCs, DSP processing used in the whole production chain, for each track that was affected, to work optimally? (ie. maybe it analyzes all the potential group delays and applies a "correction" filter to the track? One obvious issue with this is that in a multitrack recording, this also implies a need to remix.)

This MQA argument only makes sense in most cases where there is a mix of ADCs used or some tracks in the mix are purely digital like synthesizers, etc. 

 

Quote

- If there is no data about the equipment/process in the production chain, what does it do in a "batch" processing scenario? Important since it looks like the majority of the thousands of albums were not remixed and were not specially "authenticated"/supervised by the original artist/audio engineer.

Consider that about 8,000 albums have been encoded in MQA. Imagine there are 10 teams doing this encoding, and they manage to sort out all of the information of each album (ADCs etc) such that they can encode 1 album a day per team. And these teams work 252 days per year - ie every business day each team cranks out an album. This means that it will take 3.1 years for these teams to do these 8,000 albums. Ponder...

[mansr]

9 minutes ago, miguelito said:

This MQA argument only makes sense in most cases where there is a mix of ADCs used or some tracks in the mix are purely digital like synthesizers, etc. 

That's 95% of all music, minimum.

[crenca]

2 hours ago, Archimago said:

 

- Does it require impulse response data from all the ADCs, DSP processing used in the whole production chain, for each track that was affected, to work optimally? (ie. maybe it analyzes all the potential group delays and applies a "correction" filter to the track? One obvious issue with this is that in a multitrack recording, this also implies a need to remix.)

 

 

If MQA is not talking about phase delay introduced in what I take to be their idealized situational application:

 

Analogue input > ADC {here phase delay/distortion added} > MQA encoding {here phase delay/distortion removed} > MQA dac {here they are all over the place as to rather MQA dac makes a "deblurring" difference}

 

What else could they be talking about?  Has Bob Stuart discovered something "new" in signal processing?

 

Not that JA will add anything fundamental - how can he as he admits he has neither special access or the time/ability to reverse engineer.  The most he can do is to come up with an argument that puts a bit of coherency to MQA marketing materials, but so what?  He won't know if it is true any more than we will.

 

What am I missing?

[miguelito]

8 minutes ago, mansr said:

That's 95% of all music, minimum.

Exactly. And I meant to say the MQA argument makes no sense.

[Archimago]

Hey guys, I was looking over some of the oldest comments I made about MQA back in early 2015 when I wrote my first speculations on January 15, 2015. I said this in the Squeezebox forum on January 16, 2015:

 

--------------------------------------

Desperation?

As others have noted about a "solution in search of a problem", I think there's another angle here. I suspect the recording houses in their heart do not like the availability of the true "studio masters" going out the door. And I think they're realizing that the "HRA" angle isn't selling.

Other than yet "moar loud" or EQ'ed remaster, a good quality "studio master" is the end of the revenue stream; they must keep the door open for potential re-buying of the album and MQA allows another step in that. It's "high resolution" in that it's potentially better than CD (Wombat, I'd love to see if there is indeed evidence that it could be <16-bits) - at least 48kHz is better than 44kHz, right? But it's not true 24-bits nor true 96/192kHz. Call me conspiratorial if you like but I think there's a big DRM/copy protection issue here as well. Of course, we can bet this will be hacked pretty quickly just like all that have come before. If I am correct, the big labels will get on board with the idea of DRM capabilities and provide backing in a way that something like Pono would never have. They of course need another way to sell this to a public wary of copy-protection already and that's why we keep seeing these fluffy comments about "connecting" the studio/artist to the consumer, or the "analogue front end to the back end" or even the use of the term "Authenticated" (through some indicator light). I think we're in for an interesting ride in terms of how they will continue to market and push this "format".

BTW: I find Bob Stuart's "sales job" interesting... Here's a smart man with technical experience talking ad-speak putting himself out there like this as a mouthpiece. Unlike Neil Young who is clearly out of his depth with Pono, we've got a smart fellow trying to get the public to buy into his scheme. The responses appear scripted and vague by design. I smell desperation.

Another implication of this suspicion is that over time, the big names will start deprecating HDTracks and Pono as it plays favourites with MQA. Death to the "studio master", long live the almighty MQA? We shall see :-).

--------------------------------------

 

Reading that comment again in 2018 reminds me of how this has played out as suspected back then. Remember that MQA had their release party at The Shard in London on December 4, 2014 so the above comment was written just over a month later. Furthermore, Spencer Chrislu made the "crown jewels" comment in the Stereophile article published in August 16, 2016. And of course years before Mans' work opened up the technical details and explored the cryptographic signature embedded in the control stream.

 

Back then, there were many concerns about Neil Young, and indeed I still feel the same way about Bob Stuart as the mouthpiece and his stature as a technically informed, yet intentionally vague spokesperson.

 

If I and others on the forum at that time were already suspicious about all of this, it really is a marvel that all those media folks invited to The Shard including especially the audiophile press did not demonstrate any critical thought about the intent of this "format"! (LOL - perhaps it's precisely because they attended The Shard that they didn't question some of the obvious!)

 

 

[crenca]

3 minutes ago, miguelito said:

Exactly. And I meant to say the MQA argument makes no sense.

 

On the other hand, they are doing something - we all have heard it (i.e. the volume increase, the digitus or "instrument separation" depending on your take, etc.).  So perhaps it "works" to some extent even in a non-idealized situation of 95% of modern recordings:

 

https://www.computeraudiophile.com/forums/topic/38700-article-mqa-a-review-of-controversies-concerns-and-cautions/?page=24&tab=comments#comment-795567

 

at least, to some extent.  What that "something" is on the encoding side is the question.

 

[miguelito]

5 minutes ago, crenca said:

 

On the other hand, they are doing something - we all have heard it (i.e. the volume increase, the digitus or "instrument separation" depending on your take, etc.).  So perhaps it "works" to some extent even in a non-idealized situation of 95% of modern recordings:

 

https://www.computeraudiophile.com/forums/topic/38700-article-mqa-a-review-of-controversies-concerns-and-cautions/?page=24&tab=comments#comment-795567

 

at least, to some extent.  What that "something" is on the encoding side is the question.

Perhaps it works is not really an endorsement is it?