Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Meridian Audio's new... sub-format called MQA. (Read 147005 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Meridian Audio's new... sub-format called MQA.

Reply #250 –
If you're not familiar, he's spent a lot of time worrying about phase shifts, but very little (if any!) time trying to determine whether they're audible.  Since our rules won't allow him to espouse the immense virtues of his work, he's left doing what every other person who resents TOS8 does: try to pit one expert or popular belief against another.

Re: Meridian Audio's new... sub-format called MQA.

Reply #251 –
With music you can usually get away with large phase shifts.... but we are not even talking about that, but relatively small and smooth phase shifts with music.
Since aj has to know better he decided to throw in the red herring of special synthetic test signals that when distorted can sound different.

But anyway, if someone is worried about pre-ringing at >21 kHz then simply resample (can be done during playback) by x2 with a min phase filter.
"I hear it when I see it."

Re: Meridian Audio's new... sub-format called MQA.

Reply #252 –
If anyone missed that one:
The Linear vs. Minimum Phase Upsampling Filters Test
Even when we have very steep settings in that test there seems to be something to it even when Archimagos test not meets anyones academic expectations.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Meridian Audio's new... sub-format called MQA.

Reply #253 –
Quote from: Wombat on
If anyone missed that one:
The Linear vs. Minimum Phase Upsampling Filters Test
Even when we have very steep settings in that test there seems to be something to it even when Archimagos test not meets anyones academic expectations.

As Archimago notes,  ' the calculated p-values are not impressive ".  Only one comparison, for one musical sample  showed a significant difference.

Whatever the something is, it's not much.


Re: Meridian Audio's new... sub-format called MQA.

Reply #254 –
Quote from: Green Marker on
I am intrigued that one set of industry gurus say that microsecond-level timing differences are highly audible (MQA), and others say that we can heavily distort the signal with phase shifts (in non-corrected speakers for example) without it being audible.
The problem is that most proponents of the audibility of such small timing differences are too vague in their prose. I have no doubts that some are deliberately vague, while others don't see through it. It is quite a subtle matter.

Hardly anybody denies the significance of relative timing between the channels in a stereo or surround signal. That's where the figure of around 6..10 µs apparently originates. However, there's no problem with meeting this requirement with CD-type signals, or with signals of even lower resolution (even 8kHz telephone-type encoding would suffice for that). It may not be obvious for people who have difficulties with understanding the finer points of sampling, but it is demonstrably so.

But that's not the same as temporal resolution in a single signal channel. You would have to define precisely what you actually mean with this term. You may well find that what you call "such small timing differences" mean something rather different between the different authors you refer to. In Toole's case it is almost certainly the single-channel case, where the phase shift is between different frequency components in a single signal, as can be caused by filtering, or by the construction of a multi-way loudspeaker, just to state a few examples.

In the case of MQA, I have tried to figure out exactly what they mean with temporal resolution, and they are being too vague to tell. I am convinced that this is deliberate. In their 2014 AES paper (Hierarchical Archiving and Distribution), they have a large number of literature references, but fail to clear up the fog. I quote:

"By exploiting population coding, temporal resolution can approach 8 μs and this precision reflects neural processing, rather than being strictly proportional to our 18 kHz bandwidth (an estimate of the upper ‘bin’ of the cochlea and upper limit of pitch perception) [56][58][4][5][6][7][70][71]."

Some of the references given there clearly relate to interaural time differences, which is the multichannel case that doesn't need higher sampling resolution. Some of the references are papers by Dr. Kunchur who doesn't seem to understand this point, and produces erroneous conclusions based on his misunderstanding, as has been discussed elsewhere in this forum. I don't see a paper there that specifically relates to the single channel case, but I haven't been able to get hold of all of them (has anyone here read those papers?).

Later, in the same paper, they say (after discussing inter-aural timing, i.e. the multi-channel case):

"Other mechanisms have been investigated that hint to similar discrimination limits within a channel, i.e. monaurally, including: temporal fine structure in pitch perception, the comprehension of speech against a fluctuating background [91] and other cues [34]."

I haven't got access to the two references given there, but I note that the second poses the thesis as a question, and in any case Stuart et.al. use prose that doesn't suggest that the investigations have produced an affirmative result. So we should be careful not to assume that the papers give sufficient evidence. You would certainly wish that something like MQA would be based on something more solid than some "hints".

So the quoted experts may not actually be in any disagreement, because they're not talking about the exact same thing. The apparent discrepancy is the result of a mixup of different concepts.

As far as my comprehension goes, the references given by Stuart et. al. in defense of MQA do not (yet) support the conclusion, that higher sampling rates than 48 kHz are needed. A 48 kHz sampling rate provides enough temporal resolution for the multi-channel case, where phase differences between channels can be represented that are much smaller than the required 6..10 µs. It also provides enough phase fidelity among the different frequencies in a single signal channel, even when taking reconstruction filters and antialiasing filters into account (temporal resolution is not a useful term here, since such phase shifts can be continuous even in a digital signal).

There's a real danger that by perpetually citing each other's flawed or misunderstood articles, the proponents of high-res create the appearance of something that's well researched and underpinned with evidence, when in reality it is not.

Re: Meridian Audio's new... sub-format called MQA.

Reply #255 –
Quote from: pelmazo on
Quote from: Green Marker on
I am intrigued that one set of industry gurus say that microsecond-level timing differences are highly audible (MQA), and others say that we can heavily distort the signal with phase shifts (in non-corrected speakers for example) without it being audible.
The problem is that most proponents of the audibility of such small timing differences are too vague in their prose. I have no doubts that some are deliberately vague, while others don't see through it. It is quite a subtle matter.
He's deeply (or willfully) confused about the issues of multi-driver loudspeaker phase audibility, which as Toole notes, is very difficult detect (though not impossible) in typical usage, despite often having XOs in the area where our hearing sensitivity is highest...vs filtering at or beyond hearing limits.
In desperation he sees "phase" and is getting his hopes up because he doesn't understand the issues.

I find the whole idea of MQA retroactively "fixing" filtering because it is possible, though remote, that "audible" filtering was used somewhere in the recording chain, to be absurd.
The BS test apparently showed it's possible to concoct such filtering, not that it actually exists to any extent in actual recordings.
It's clearly aimed at audiophiles who want the 5th version of the same 2ch construct. They're always starving for more.

cheers,

AJ
Loudspeaker manufacturer

Re: Meridian Audio's new... sub-format called MQA.

Reply #256 –
Quote from: ajinfla on
In desperation he sees "phase" and is getting his hopes up because he doesn't understand the issues.
He would certainly not be alone with that. I'd say that a large part of consumer land doesn't understand it, and it seems that there is a significant number of professionals who have their problems with it, too.

If you read the paper by Stuart et.al. which I was quoting from above, you can't help thinking that they actually do understand it, and that they are quite conscious of the different contexts and meanings of the term. I wouldn't have expected otherwise from them. But that means that they must be obfuscating the matter deliberately. They tiptoe back and forth between notions of bandwidth and resolution which have no simple relation with each other and can not be translated in the way they implicitly suggest.

Quote
I find the whole idea of MQA retroactively "fixing" filtering because it is possible, though remote, that "audible" filtering was used somewhere in the recording chain, to be absurd.
The BS test apparently showed it's possible to concoct such filtering, not that it actually exists to any extent in actual recordings.

I thought the idea was to have MQA everywhere in the chain, so that there's no need to remove such artefacts. That doesn't make it any more sensible, however. Its necessity for transparency hasn't been demonstrated, neither by their use of scientific references, nor by their own listening tests. We're still (and that's after decades of propaganda) at the stage of "hints".

Re: Meridian Audio's new... sub-format called MQA.

Reply #257 –
@Pemazo, et al.:
You've omitted what is possibly the most important part of AJ's post:
Quote from: ajinfla on
He's deeply (or willfully) confused about the issues of multi-driver loudspeaker phase audibility, which as Toole notes, is very difficult detect (though not impossible) in typical usage, despite often having XOs in the area where our hearing sensitivity is highest...vs filtering at or beyond hearing limits.

There's been a lot of wanking going on in this discussion; little of which is actually on-topic.

Maybe I should re-open the other BS topic and let you all have back at it.

...of course it still doesn't open the door for those who are butt-hurt because they can't talk about their still-off-topic pet interests.
 

Re: Meridian Audio's new... sub-format called MQA.

Reply #258 –
No confusion on my part, but thanks for your concern, AJ.

I am just trying to cut through the woolliness of this whole timing argument. We have detailed discussions on the minutiae of the timing implications of sample rates, bit depths etc. probably playing the MQA people's 'game' for them, but then listen to the results on speakers which distort the phase and/or don't have time-aligned drivers. Presumably many of the people being suckered by MQA are doing the same thing. It would be nice to point out the absurdity of it without getting drawn into a cycle of competitive 'fact' regurgitation.

(why not save your blood pressure, AJ, and simply ignore my comments?)

Re: Meridian Audio's new... sub-format called MQA.

Reply #259 –
Quote from: Green Marker on
I am just trying to cut through the woolliness..
..but then listen to the results on speakers which "distort" the phase and/or don't have "time-aligned" drivers.
No (self assessed) confusion on my part
Sigh
Loudspeaker manufacturer

Re: Meridian Audio's new... sub-format called MQA.

Reply #260 –
Quote from: Green Marker on
I am just trying to cut through the woolliness of this whole timing argument.

By complaining about nothing to people who do not care?   This is the least effective way you could possibly do that. 

Quote from: Green Marker on
It would be nice to point out the absurdity of it without getting drawn into a cycle of competitive 'fact' regurgitation.

If you don't like the existing evidence, I'd think it would be fairly straightforward to replicate any timing difference you like and produce your own data.  Not sure it would be interesting, but you could certainly do it. 

Re: Meridian Audio's new... sub-format called MQA.

Reply #262 –
^From that link:
Quote
Note that even though I was aware that the MQA decoded file had a higher noise floor, I did not purposely take advantage of this by listening to a quiet portion of the file with volume pumped up to listen to the hiss. That would be cheating.
-Archimago

So there is an audible difference, the background hiss, however he assures us he didn't consciously listen for it. And what assurance do we have that he didn't subconsciously hear it and/or key on that hiss difference without being consciously aware that was doing so?

Just like we can't accept data from people who say, "Well, there was an audible difference in level between A and B, but I assure you, I'm special and can hear passed that", we also can't accept data from people who say, "Well, there was an audible difference in background hiss, but I assure you I only listened to song sections where it shouldn't matter, and I can't possibly be mistaken on that point because I'm special."

Re: Meridian Audio's new... sub-format called MQA.

Reply #263 –
It is the first source i trust about any of these MQA results. When you read older posts of Archimago he often failed to abx things he could have easily cheated, especially with his nice equipment around.
Maybe ask him for short samples at his blog if you are curious?
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Meridian Audio's new... sub-format called MQA.

Reply #264 –
Just to be clear, I never accussed anyone of consciously cheating.

We know that often when there is a small fraction of a dB difference in level, say around .25 to .5 dB, many if not most people are unaware that the sound difference they can successfully differentiate (ABX) is nothing more than this difference in level. They however often attribute it to "clarity, detail,. de-blurring" etc..  It is an uncanny illusion and just like the McGurk Effect being well educated on the illusion doesn't make one immune to it. Even the world's top expert on the McGurk Effect, Dr. Laurence Rosenblum, says he himself is still fully susceptible to the illusion even though he knows better than anyone how it works. [reference available upon request]

Re: Meridian Audio's new... sub-format called MQA.

Reply #265 –
So it seems there is some ABX'able difference, maybe just the noise floor. And it's just different, not necessarily better. I'm sure that's pretty much what we all expected, should there be an actual audible difference.

His statement that he did not feel very confident in his choices and that he could only identify any difference by switching back and forth and presumably listening to the same short section over and over again, echoes my own experiences with ABXing MP3 close to my personal limit of transparency. Extremely subtle differences that are very hard to pin down or put into words, but still significant enough for a (near) 100% success rate. I would never be able to identify the tested tracks as FLAC or MP3 in ordinary listening, though. Not even very focused and concentrated listening.

Talk about diminishing returns. With the files nulling out to -74dB, the differences can only be very subtle. And the much-vaunted "de-blurring" can be applied to any PCM signal in the studio, so there's absolutely no reason why you couldn't do it to everything, without having to distribute it in MQA format.

Re: Meridian Audio's new... sub-format called MQA.

Reply #266 –
Quote from: mzil on
^From that link:
Quote
Note that even though I was aware that the MQA decoded file had a higher noise floor, I did not purposely take advantage of this by listening to a quiet portion of the file with volume pumped up to listen to the hiss. That would be cheating.
-Archimago

So there is an audible difference, the background hiss, however he assures us he didn't consciously listen for it. And what assurance do we have that he didn't subconsciously hear it and/or key on that hiss difference without being consciously aware that was doing so?

Just like we can't accept data from people who say, "Well, there was an audible difference in level between A and B, but I assure you, I'm special and can hear passed that", we also can't accept data from people who say, "Well, there was an audible difference in background hiss, but I assure you I only listened to song sections where it shouldn't matter, and I can't possibly be mistaken on that point because I'm special."

This seems like carping on your part.

His ABX results indicate he heard a difference, and he went looking for what it might be .   In the end his most reasonable candidate is the 'de blurring' step.  He's on solid ground in presuming that a small difference in noise floor  (not overall level) would be masked during music played at normal levels.  Even then he doesn't rule out that the changing noise floor with frequency , audible in a difference file played at high level, might account for his ABX

 He's not a fan of MQA.  

Quote
Realizing that Meridian/MQA has provided essentially no technical details or objective results, if I am correct about what is going on as described above, I am personally not interested in MQA as a format I feel I would want. There’s no “magic” here and there are evident compromises when trying to be everything to everyone as MQA seems to be aiming for.


Re: Meridian Audio's new... sub-format called MQA.

Reply #267 –
I apply the exact same standards to everyone, regardless of what camp they fall in, including myself. If there is an audible difference in level or background hiss between A and B, then I don't even bother to run a test on myself.

Re: Meridian Audio's new... sub-format called MQA.

Reply #268 –
But you understand that 'can be audible' doesn't mean 'is always audible', right?

Re: Meridian Audio's new... sub-format called MQA.

Reply #269 –
He measures a difference in noise at -74dB. This is not hiss as i understand it and worth an abx.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Meridian Audio's new... sub-format called MQA.

Reply #270 –
Quote from: mzil on
... So there is an audible difference, the background hiss, ...

He didn't say it was audible. He said that there was a measurable difference, but he had not tried to unrealistically boost the level to see if he could hear it.
Regards,
   Don Hills
"People hear what they see." - Doris Day

Re: Meridian Audio's new... sub-format called MQA.

Reply #271 –
I invited our forum member Archimago, I assume the same person, to join us and post his samples. I also attempted to do so at his blog but after registering through my Google account, it seemed to get stuck in an endless loop and didn't work. Could someone else please try?

I was curious what version of ABX he was using, meaning if it was current, so I opened a test of my own just to see what version I had. [I don't keep track of the version numbers; they are Greek to me.] One of the songs I happened to have ready to go on my playlist you'll see happens to be the same name as his song, since I wanted to see what kind of a song it was, so I downloaded a 1 minute sample. The other song, B,  is just dead air on Arny's setup. [Where is he by the way?]

Note my time entries. This is from just clicking "This is X", "Next trail", ten times, and I beat his score just by random chance!  What a weird fluke!

Code: [Select]
foo_abx 2.0.1 report
foobar2000 v1.3.9
2016-02-15 18:27:58

File A: Blagutten 1 minute.flac
SHA1: f60ec14800aaf64032491a6d13e3bf9df15a76c6
File B: digital_black_2496.wav
SHA1: 7965703e5d9eea3f3d60ab5399a1423727a4e992

Output:
DS : Primary Sound Driver
Crossfading: NO

18:27:58 : Test started.
18:28:17 : 01/01
18:28:18 : 01/02
18:28:19 : 02/03
18:28:21 : 03/04
18:28:22 : 04/05
18:28:24 : 05/06
18:28:25 : 06/07
18:28:27 : 07/08
18:28:29 : 08/09
18:28:30 : 09/10
18:28:30 : Test finished.

 ---------- 
Total: 9/10
Probability that you were guessing: 1.1%

 -- signature -- 
2fe5385a85cd5e851b057697d3e46ee534c31f32

Considering he only made one test and didn't even hit a p-value of .05 or better, it is odd how many seem convinced he heard a difference. Stranger things have happened. . .in fact one just did: my 9/10 score by random chance!

Re: Meridian Audio's new... sub-format called MQA.

Reply #272 –
Are other people seeing my ABX data in multi colors? I just cut and paste the data. I have no idea where the colors came from.

Re: Meridian Audio's new... sub-format called MQA.

Reply #273 –
Colored, yes. The way you argue foobar abx means nothing then, never when 9/10 are hit. So don't wonder if he won't come here to proof anything.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Meridian Audio's new... sub-format called MQA.

Reply #274 –
Archimago used foo_abx 2.0 and foobar 1.3, as can be seen from the screen capture in his post.

I've notified him in comments there that 8/10, at p=0.055, is still shy of 'significant' by standard criteria.