• 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!

Archimago's MQA listening test results

Ken Newton

Active Member
Joined
Mar 29, 2016
Messages
190
Likes
47
If this time domain is a concern to you then the emphasis is in the wrong place. Have you ever taken a look at what a typical hifi speaker does to the time domain?

I'm neither concerned nor unconcerned. I'm simply addressing the technology behind digital audio time-domain prioritization, which Bob Stuart says MQA is concerned with.
 

Don Hills

Addicted to Fun and Learning
Joined
Mar 1, 2016
Messages
708
Likes
464
Location
Wellington, New Zealand
I'd like to see a test (properly designed and proctored) that shows there is an audible difference between a "standard" brick-wall filtered 192 KHz DAC and an MQA filtered one.

I'm happy enough with his encoding the 24-48 KHz range and storing it in the low 8 bits of a 24/48 channel. That's logical and elegant. But I see the messy imaging/alias of the 48-96 KHz range as unnecessary and snake oil.

... Hmmmn. What are the coefficients of the conjugate filters used to split (and later rejoin) the input into the 0-24 and 24-48 ranges for the "origami" fold?
 

Ken Newton

Active Member
Joined
Mar 29, 2016
Messages
190
Likes
47
I'd like to see a test (properly designed and proctored) that shows there is an audible difference between a "standard" brick-wall filtered 192 KHz DAC and an MQA filtered one.

Yes, I'd like to see that as well. That is the primary question, as far as subjective sound performance.

The shame could be, that should time-domain optimized MQA prove popular with consumers, the audio industry can only blame itself for that success. There is every opportunity to produce a time-domain optimized mastering standard based on 24b/192kHz PCM. That's all that's required, an industry open standard, primarily affecting the anti-alias filter response specification. That, plus some marketing logo or graphic symbol identifying albums which have been so mastered, is all it would take. No magic DSP algorithms, no change in manufacturing process or content distribution would be required.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
Fyi I have the mqa cheeky cd. I plan to buy the high res version so can create comparison.

Problem is finding a Windows audio capture that doesn't use kernel mixer. Anyone know any? Need to capture output of decoded ma from roon.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
I'd like to see a test (properly designed and proctored) that shows there is an audible difference between a "standard" brick-wall filtered 192 KHz DAC and an MQA filtered one.

I'm happy enough with his encoding the 24-48 KHz range and storing it in the low 8 bits of a 24/48 channel. That's logical and elegant. But I see the messy imaging/alias of the 48-96 KHz range as unnecessary and snake oil.

... Hmmmn. What are the coefficients of the conjugate filters used to split (and later rejoin) the input into the 0-24 and 24-48 ranges for the "origami" fold?

That was my thinking too. A nice way to add response if you think it important. The other is a lot of shite thinking. Mostly meant to show the sound of impulses that don't occur in the real world as un-smeared. I think the whole blurring thing is just marketing brouhaha. The "blur" is just spreading of energy in the transition zone which we can't hear anyway.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
I thought archimago emulated ma filter in his PCM samples. On the phone and can't check.

He did, and mansr, helped him do that. mansr has done alot along with a user named soxr and Miska (HQ Player author) to figure out what is really happening.
This mansr.... https://github.com/mansr Mans Rullgard.

The best that can be determined MQA DACs are implementing a loose filter that lets aliasing occur along with a more conventional filter. It appears there is none of the fixing the time domain at the other end, nor anything more advanced than this going on. It gives a visually pretty impulse response, but that doesn't really fix time domain issues. So if fits with other activities that suggest it is more a marketing scam to enrich the perpetrators and potentially function as DRM. If my posting has errors they are mine and not Mans Rullgards. The rest is just unfolding, sometimes upsampling using those hidden lower 8 bits for higher frequency info.

So yes, Archimago's files should be like MQA as it so far is implemented. Is that full MQA or is more of the promised things possible though not yet being done? All we have is data from what is out there. If there is nothing more then lots of promises of MQA either cannot or are not being fulfilled.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
This discussion is going in an absurd direction. Discussing the merits of MQA in a world where MQA controls the process from capture to playback - which is not going to happen - is meaningless.

Discussing an unlikely scenario (total MQA control of the AD-DA chain) is the definition of waste of time.

For audio science to be relevant it has to deal with real-world problems.
 

Ken Newton

Active Member
Joined
Mar 29, 2016
Messages
190
Likes
47
The "blur" is just spreading of energy in the transition zone...
Exactly, and all music being composed entirely of transient signals means that music features many many such transition zones.

...which we can't hear anyway.
That seems the key audiophile question. What's your evidence for asserting that we cannot hear it? Meridian has pointed to academic studies indicating that we can hear it. I've come to no conclusion myself.
 

Werner

Active Member
Joined
Mar 2, 2016
Messages
109
Likes
135
Location
Europe
Exactly, and all music being composed entirely of transient signals means that music features many many such transition zones.

Ken, you know better than that. A filter's transition zone is defined by the filter, not by the signal. If the signal does not contain energy at the transition point, then the filter does not ring. This is easy to verify.

As for the research supporting MQA: if you delve a bit deeper in to these tens of papers they refer, you will find that most of them do not even remotely seem to apply. It is almost comical.
 

Ken Newton

Active Member
Joined
Mar 29, 2016
Messages
190
Likes
47
Ken, you know better than that. A filter's transition zone is defined by the filter, not by the signal. If the signal does not contain energy at the transition point, then the filter does not ring. This is easy to verify.

As for the research supporting MQA: if you delve a bit deeper in to these tens of papers they refer, you will find that most of them do not even remotely seem to apply. It is almost comical.

[I've substantially edited my below response to Werner, as upon re-reading of his comment I realized that I hadn't originally interpreted it correctly.]

Werner, except that transient signals can contain such energy as they transition in and out, so the filter still can ring. This is easiest to illistrate via a digital FIR brickwall filter.

Here's an illustration: Take 100 cycles of an Fs/4 digitized tone burst (one-half the system Nyquist frequency) to represent a transient and pass it through an 255 stage linear phase FIR filter. The 100 cycle burst will, on average, have a duration of 4 samples/cycle = 400 total sample periods. After the burst is passed through the 255 stage FIR filter the tone burst will be stretched from 400 sample periods to 400 + 255 = 655 sample periods. This stretching mechanism is absolutely no issue with infinitely continuous signals. However, it does objectively distort transient signal information, which, of course, is what music is complexly composed. The issue isn't the frquency of the transient, it's the fact that by definition a tramsient signal is not continuous.
 
Last edited:

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
All the above said, it doesn't necessarily mean transient blurring in digital audio is audible. Of course, Meridian believes it is audible.
Correction: Meridian says it believes it is audible.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
There is still the issue that supposedly 'de-blurred' audio is being played back over the most blurry devices imaginable: passive speakers. You know, those boxes where there are phase rotations through the crossovers, and if the drivers are not at least on a sloping baffle then the smearing is measured in milliseconds not microseconds.
 

Fitzcaraldo215

Major Contributor
Joined
Mar 4, 2016
Messages
1,440
Likes
632
I think I can see most all sides of the technical arguments. I think part of the underlying problem is that MQA goes so far out on a limb by using claimed psychoacoustic advantages, including many that are not widely accepted, if they indeed ever will be.

As much as Stuart, et al, have published justifying their technical approach and alluded to psychoacoustics, they have not concretely proven the advantage to the listener. And, skeptics who do not accept those psychoacoustic assumptions and lean heavily on traditional theory, give MQA a hard time and throw rocks, to put it mildly. Of course, other counterproductive and often nasty speculations and red herrings - DRM, Meridian/MQA ownership/profitability/financing, that Stuart is a liar and a fraud, etc., etc. - have also been spawned by those who do not wish to see the technical status quo upset in any way.

If it is fundamentally about psychoacoustics, then only carefully controlled listening tests can truly demonstrate any superiority of MQA. So, I agree with others in this thread on that. I think Archimago (I am usually a fan of his) has offered a nice try, but one that might be open to serious questions as to its rigor. I say that in the middle of reading Toole's excellent Sound Reproduction -3rd Edition, in which he clearly defines how to go about conducting such rigorous psychoacoustic testing based on listener preferences and correlating those to measured, technical observations, brilliantly and convincingly I might add. But, speakers in rooms are his sole interest.

We do not have such tests. MQA could not themselves conduct them. They would be dismissed as tainted. So, who might conduct them and when? I note that DAC-maker Berkeley claims to have done this as part of their due diligence before accepting MQA, but it is unpublished. Maybe no one ever will publish it, and it will be left for audiophiles in the marketplace, subject to the usual chaos and lack of scientific discipline.

I find one thought by Ken most interesting: the notion of a true end to end, analog to analog recording/playback chain using MQA directly in the a-d and throughout the chain, as opposed to MQA being used to post process the result of traditional a-d later in the chain. MQA has not openly described this possibility. But, I would love to hear the results alongside a conventional hi rez chain, difficult though this might be with today's studio equipment infrastructure.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
I think I can see most all sides of the technical arguments. I think part of the underlying problem is that MQA goes so far out on a limb by using claimed psychoacoustic advantages, including many that are not widely accepted, if they indeed ever will be.

As much as Stuart, et al, have published justifying their technical approach and alluded to psychoacoustics, they have not concretely proven the advantage to the listener. And, skeptics who do not accept those psychoacoustic assumptions and lean heavily on traditional theory, give MQA a hard time and throw rocks, to put it mildly. Of course, other counterproductive and often nasty speculations and red herrings - DRM, Meridian/MQA ownership/profitability/financing, that Stuart is a liar and a fraud, etc., etc. - have also been spawned by those who do not wish to see the technical status quo upset in any way.

If it is fundamentally about psychoacoustics, then only carefully controlled listening tests can truly demonstrate any superiority of MQA. So, I agree with others in this thread on that. I think Archimago (I am usually a fan of his) has offered a nice try, but one that might be open to serious questions as to its rigor. I say that in the middle of reading Toole's excellent Sound Reproduction -3rd Edition, in which he clearly defines how to go about conducting such rigorous psychoacoustic testing based on listener preferences and correlating those to measured, technical observations, brilliantly and convincingly I might add. But, speakers in rooms are his sole interest.

We do not have such tests. MQA could not themselves conduct them. They would be dismissed as tainted. So, who might conduct them and when? I note that DAC-maker Berkeley claims to have done this as part of their due diligence before accepting MQA, but it is unpublished. Maybe no one ever will publish it, and it will be left for audiophiles in the marketplace, subject to the usual chaos and lack of scientific discipline.

I find one thought by Ken most interesting: the notion of a true end to end, analog to analog recording/playback chain using MQA directly in the a-d and throughout the chain, as opposed to MQA being used to post process the result of traditional a-d later in the chain. MQA has not openly described this possibility. But, I would love to hear the results alongside a conventional hi rez chain, difficult though this might be with today's studio equipment infrastructure.

MQA have easy access to funds if they wanted to arrange state of the art quality listening tests by objective third parties. They even have financial backing to make whatever end-to-end equipment, say in a series of 500 copies (to get some scale and distribution) if they wanted to.

So why don’t MQA make this investment if they believe in the technology’s superiority. As we know, Johann Rupert is trigger happy if he sees financial opportunities that water out his tobacco money.

This is kindergarten level logic as opposed to sophistry, i.e. discussing what if scenarios that are not going to happen.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
...skeptics who do not accept those psychoacoustic assumptions and lean heavily on traditional theory, give MQA a hard time and throw rocks, to put it mildly. Of course, other counterproductive and often nasty speculations and red herrings - DRM, Meridian/MQA ownership/profitability/financing, that Stuart is a liar and a fraud, etc., etc. - have also been spawned by those who do not wish to see the technical status quo upset in any way.
I think it is more than that. The most mild-mannered sceptic can sense which way the wind is blowing: that aggressive marketing and somewhat sharp practice may not only result in mugs paying for the privilege of imagining sonic improvements, but that he may be forced to do the same. The non-decoded MQA version is inferior to straight CD. And record companies will start releasing genuinely inferior masters for the non-MQA formats - or will at least drop hints that they are - knowing that anyone who cares about quality will be forced to stump up the cash, while everyone else won't notice a thing. We already see such scandalous shenanigans beginning to creep in because of the cult of vinyl - e.g. the Sgt. Peppers remaster that applies "gentle limiting" (actually hard clipping) to the CD version, but not the LP.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
[I've substantially edited my below response to Werner, as upon re-reading of his comment I realized that I hadn't originally interpreted it correctly.]

Werner, except that transient signals can contain such energy as they transition in and out, so the filter still can ring. This is easiest to illistrate via a digital FIR brickwall filter.

Here's an illustration: Take 100 cycles of an Fs/4 digitized tone burst (one-half the system Nyquist frequency) to represent a transient and pass it through an 255 stage linear phase FIR filter. The 100 cycle burst will, on average, have a duration of 4 samples/cycle = 400 total sample periods. After the burst is passed through the 255 stage FIR filter the tone burst will be stretched from 400 sample periods to 400 + 255 = 655 sample periods. This stretching mechanism is absolutely no issue with infinitely continuous signals. However, it does objectively distort transient signal information, which, of course, is what music is complexly composed. The issue isn't the frquency of the transient, it's the fact that by definition a tramsient signal is not continuous.

The difference in the signal before and after the filter is the stretching will contain energy between 20 khz and the cutoff (22.05 khz?). You can see it on the waveform in software. You aren't going to hear that. If you use 88 or 96 khz sampling then everything is over 40 khz in the stretching. Not to mention your tone burst is a sharper transient than is usually contain in music recordings. Even crashing cymbals take a few dozen samples to reach peak intensity.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Exactly, and all music being composed entirely of transient signals means that music features many many such transition zones.

That seems the key audiophile question. What's your evidence for asserting that we cannot hear it? Meridian has pointed to academic studies indicating that we can hear it. I've come to no conclusion myself.

As already pointed out some of the studies don't seem to support MQA even though quoted by them. Others like the ones they did are very curious. They used extraordinarily sharp filters to compare to 192 khz recordings. The transition band of those downsampled filters were 500 hz and I think 423 hz. They used even smaller transition zones in training people to hear this. Used no dither or a known to be audible dither to downsample. Then used an extraordinarily wideband playback system. I don't fault them for that. But after these unusual conditions with trained listeners in an excellent listening environment instead of 50:50 results the listeners chose the files above 5% chance level (just barely) at 56:44 ratios. At a minimum that means the audibility is fleeting and minor.

But okay, grant them their results, the next step is to use proper dither, use the 2khz or wider transition band and see where audibility disappears. Do we need more to make the effect inaudible do we need much more? Is 96 khz not enough? How big is the problem MQA solves if it really amounts to something specially trained listeners hear 56% of the time? So their testing seems geared merely for PR for their system instead of an investigation of what is needed. Not much money to be made if the answer is use 48 khz FLAC or 88 khz/16 bit FLAC or if there was no effect if you used proper downsampling.

Now I have read JJ Johnston describe how sharp transition bands could be marginally audible to small numbers of young people with excellent hearing. This is like 1-2% of adults between 18-30 who have some hearing response to 22-24 khz. JJ thought 48 khz very nearly solved any issue even for those people in all likelihood. Said if he wanted a fully blameless format he would have chosen 60-65 khz sampling with the transition band starting at 25 khz. Which isn't to say anyone had shown they could hear the effect in existing sample rates. That is one of the reasons MQA in some places says it keeps response to 30 khz and rolls off the rest (though they said differently in other places).
 

Ken Newton

Active Member
Joined
Mar 29, 2016
Messages
190
Likes
47
The difference in the signal before and after the filter is the stretching will contain energy between 20 khz and the cutoff (22.05 khz?). You can see it on the waveform in software. You aren't going to hear that. If you use 88 or 96 khz sampling then everything is over 40 khz in the stretching. Not to mention your tone burst is a sharper transient than is usually contain in music recordings. Even crashing cymbals take a few dozen samples to reach peak intensity.

I specifically did not address issues of audibility, because I've not personally experimented with this question. Only that transient spreading due to brickwall filtering objectively exists.
 
Top Bottom