Regardless of why if you read the manufacturers response to the Stereophile article you will see that this is being eliminated with the latest implimentations of MQA even though it was not audible at high volumes and in very quiet conditions.
Oh and futhermore if you read the article you will find this was not a typical MQA file but an “outlier” (the term used in the article itself) so is not characteristic of all MQA files but just found on some unusual files.
But neither of these points fit with the story some want to propagate so they don’t mention them.
The A10 misapplication of MQA to a non-MQA files is done deliberately by the manufacturer for seamless switch (to avoid click and Pops) without having to switch from default DAC filter to a MQA filter. However, you can see due to the short and slow roll off filter, ultrasonic noise above 20kHz gets reflected back to the audio range. This is a fact. Nobody use this type of ‘leaky’ filter due to aliasing issue.
There’s a separate test results comparing MQA to a Hi-Res files:
Here you can see MQA is 20dB more noisy to a Hi-Res(DXD) from 10kHz to 30kHz. These are technical data that shows MQA is NOT a solution to Hi-Res.
Another interesting reading from Archimago’s Musing:
I’ve not heard from a ‘wrong decoder’ mentioned anywhere. The test files are from reputable music mastering, 2L. They produce one of high quality recordings and I brought several albums from them, mostly DXD masters
I’ve seen reference to a better algorithm and the 2L MQA files being recreated. Not a wrong encoder.
MQA is basically a DSP, it can be improved and changed. For streaming this if fine but I would hope people realize how it’s a big strike against purchasing MQA content.
It is unclear to me exactly what is going on, but MQA is an end-to-end system; this cannot be ignored. MQA’s inventors have forgotten more about digital filters than most of the forum “experts” will ever know. The filters may look leaky as isolated filters. However, this analysis completely disregards the DSP applied during the MQA creation process which, I believe, is complementary to the decoding DSP. In other words, aliasing is taken care of within the system as a whole.
The end-to-end system is the biggest problem. If it takes off and the record labels eventually only release MQA for digital delivery consumers only have one option, the MQA sound.
I don’t like the sound of minimum phase, zero pre and ver little post ringing. I prefer the sound of linear phase filters and hope to have the option down the road.
MQA creation process starts at the very front end ADC which uses so called ‘leaky filters’ to achieve reasonable good impulse response even when not fully decode at end users. At the DAC side the filter too is ‘leaky’ this whole process allow ultrasonic noise to get ‘reflected’ downstream causing aliasing. The whole purpose here is to get the best impulse response and sacrifice some noise and distortion.
My arguments is there’s no perfect solution here; either time or frequency domain or strike a balance in both of them and I believe MQA is trying to do just that.
From that article, I see nothing to imply that the end-to-end MQA process produces a minimum phase output or uses minimum phase filters. MQA DACs might have those filters for non-MQA material (or even for if MQA decoding can be turned off in them), but that’s completely different.
Here is Bob Stuarts response to the added noise in this file - even though it is audible, as Jim Austin demonstrated.
Here is the bottom line if you don’t want to read it all and since it is obvious many people are not reading it.
“2L-078 is one of a dozen albums that used the early Sphynx2 converters; … Fig.2 shows that, using our latest tools, there is no longer added noise in the audioband (even though it was already inaudible and sounded great)”
So next generation of MQA will not have this in any files.
Note also that Jim Austin points out the following:-
“At this point I would love to show you what an MQA-decoded file looks like, point out all its nuances, and tell you what they mean. But I’ve recorded and analyzed—or tried to—dozens of MQA files, and have noticed few patterns. I’ve likely encountered most of MQA’s 2000+ encapsulation algorithms: Every MQA file seems to do things in a different way, diverging from the apparent non-MQA source in many different ways, and sometimes not at all. Every time I thought was getting a handle on some aspect of MQA, I’d look at another file and find that it didn’t conform to my theory.”
Austin also makes that the point that this file that it seems everyone is considering a characteristic of how MQA works is on “outlier”. And I quote again.
“So instead of showing you a typical MQA file—there’s no such thing—I’ll show you an outlier.”
So taking this file as typical of MQA and drawing conclusions about the way MQA always works is plain wrong. Seems that could be more than 2000 ways MQA works. Furthermore since future MQA files will not contain this noise there is little point in continuing to debate it.
But what choice do you have once encoded to MQA? Linear or Minimum or whatever. It’s an end to end solution to a problem that doesn’t exist. Allot of people seem to like the sound of the solution and that’s fair, but I don’t want that solution eventually forced on me.
@DrTone You don’t have any control of any encoding today at all. So I that is a non argument. You don’t buy from HDT or Qoboz with any knowledge of any of that.
As for decoding, if you get first unfold of lossless data then you can then do with that as you wish. Just don’t unfold any of the higher data rate should that be present.
The majority of MQA is 44 and 96k anyway.
The text above says the issue was with the first dozen albums as the earliest/first adopter. You can worry about “what if’s” if you want. But it is a glass half empty p.o.v.
Studios screw up mastering and releases all the time. Always have. I don’t recall ever a studio or label actually letting you know that the first twelve albums had an issue and it was resolved already.
Provenance is provided by some retailers. CD provenance to master can be researched on places like the Hoffman forums.
Take MQA out of the picture and now I don’t have to worry about their algorithm messing up with a certain content. Or if they develop MQA 2.0 encoding.
PS) The first unfold doesn’t produce lossless data, it’s not a binary compression.
“The Ferrari F12 is so pathetic. It doesn’t have keyless go, or adaptive cruise control, or blind spot warning, it can’t even park itself!” “What is it like to drive? How does it perform on the road?” “Oh, I’ve never driven a Ferrari.”
“The Hasselblad X1D is so pathetic. It doesn’t have a live histogram or a tilting screen or a focus point joystick, and the fastest shutter speed is 2000!” “What is it like to use? What do your pictures look like?” “Oh, I’ve never used a Hasselblad.”
“MQA is so pathetic. Leaky filters, and minimum phase, and noise floor bla bla bla…” “What does it sound like on the music you listen to?” “Oh, I’ve never listened to MQA.”
I didn’t say decoder I said encoder… In the article you posted MQA notes at the end:
“Finally, a comment about the analysis of 2L’s Veslemøy Synsk (2L 2L-078). This album (released in May 2011) was one of the first to be encoded to MQA (in December 2015). DXD is the most revealing and challenging source; successive generations of converters vary a lot in terms of noise floor (see fig.1). Over the last two years we have steadily improved the encapsulation and deblurring stages to best match the various converter designs, and in the case of 2L, several albums have been re-encoded where [2L recording engineer] Morten [Lindberg] and we felt it was advantageous (footnote 3).
2L-078 is one of a dozen albums that used the early Sphynx2 converters; many of these were scheduled to be remastered this winter, and should be available from 2L by the time this issue is available. Fig.2 shows that, using our latest tools, there is no longer added noise in the audioband (even though it was already inaudible and sounded great).—Bob Stuart, MQA”
The noise resulted from a 2015 encoder hat was not optimized for the 2L converters. The 2017 encoder fixed the issue—it was an anomaly.
