MQA Listening Impressions

[manisandher]

Edit 30/11/17: Please use the following files (captured directly from the analogue outputs of my DAC using a Tascam DA-3000) to compare the original 24/88.2 against the 16/44.1 and 24/44.1-MQA... all from the same master.

 

Track: Christina Pluhar - Music for a While - Improvisations on Purcell - 09_Wondrous machine_, Z. 328_8 (arr. Pluhar)

 

24/192 PCM

 

24/88.2 original _ HQPlayer _ poly-sinc-short-mp to 705.6 _ TPDF:

https://drive.google.com/open?id=1ouSdts0JtlCyjlJCN40iIYvI0jtSETRk

 

16/44.1 _ HQPlayer _ poly-sinc-short-mp to 705.6 _ TPDF:

https://drive.google.com/open?id=1oTW7ra5cWln9R_fYQYotSm7q8nYVwA3K

 

24/44.1-MQA _ XXHighEnd _ first unfold to 88.2 _ Arc Prediction to 705.6:

https://drive.google.com/open?id=1m2npYzIraEKweh-I8Eu-u7VMXTCNzzx-

 

DSD128

 

24/88.2 original _ HQPlayer _ poly-sinc-short-mp to 705.6 _ TPDF:

https://drive.google.com/open?id=1EpKgFkcpkaTsm54Ot21ZM_9YW2CgnGR5

 

16/44.1 _ HQPlayer _ poly-sinc-short-mp to 705.6 _ TPDF:

https://drive.google.com/open?id=1DHXqn7IDeU0__WEb2IPvfQHlEZAMDfwh

 

24/44.1-MQA _ XXHighEnd _ first unfold to 88.2 _ Arc Prediction to 705.6:

https://drive.google.com/open?id=1i3PE4GB2VbLBSjGSnPQMTsiPM5_atAqC

 

So, which sounds best to you, and why?

 

********************************************************************************************

 

Just wanted to share my listening impressions of MQA. Happy for anyone else to share their listening impressions too, but ask that people refrain from turning this into yet another ' MQA is evil' or 'Bob Stuart is the devil' type of thread. Much appreciated.

 

I started by comparing the various 'Mozart: Violin concerto in D major - Allegro' free download files from 2L. I compared the original DXD (24BIT/352.8kHz ) to the 24BIT/96kHz, 16BIT/44kHz and MQA files. My preference was as follows (in descending order):

 

1. original DXD

2. MQA (very close to 1.)

3. 24BIT/96kHz (big gap between this and 2.)

4. 16BIT/44kHz (OK, if you have no other choice, I suppose)

 

I've since been playing a bunch of MQA albums from Tidal. Invariably, I prefer the sound of these MQA albums over the CD rips and/or downloads of the same albums I already have. They tend to have a clarity that my rips/downloads don't have. Overall just a nicer listening experience. But I have to say that apart from the 2L files, I've not managed to compare MQA to any original hirez files.

 

From a purely sonic perspective, MQA seems to do 'what it says on the tin'.

 

Mani.

[mansr]

What DAC did you use?

[kumakuma]

1 minute ago, mansr said:

What DAC did you use?

 

Probably a Yggdrasil...

[watercourse]

Interesting feedback. The Phasure is a non-MQA DAC, correct?

 

I have been trying out MQA partial processing via Audirvana+ to my non-MQA DAC, and have found MQA to be pleasant listening as well. However, MQA files often (over half the time) sound like the files have received more processing once comparing to the same non-MQA tracks. This is most evident IME in slightly attenuated treble extension and energy. At the same time, bass seems slightly tighter, and soundstaging moves forward a tad bit. Some albums have significantly better imaging than non-MQA tracks, which I assume to be a hallmark of additional processing, or using different source files. Very smooth and inoffensive, and easy to listen to.

 

Comparing local lossless flac and hi res files to streaming non-MQA Tidal Hi Fi files, I tend to prefer local files. Tidal MQA in general is more polite and processed-sounding in comparison, but so are regular Tidal flac files (I noticed this pre-MQA as well, sonically akin to MP3 processing, only better).

 

Since MQA files also generally sound like different masters, it’s hard to know if MQA processing benefits or harms the music.

[Miska]

3 hours ago, manisandher said:

1. original DXD

2. MQA (very close to 1.)

3. 24BIT/96kHz (big gap between this and 2.)

 

First question of course is what was used to create 96k from the original DXD? Did you create it yourself, or did you let the player do the conversion for you? What did you use as MQA decoder?

 

2L is not specifically the type of music I would listen. I have other types of music as both original 96/24 files and MQA versions of the same, and to me the original sounds better than the MQA'd version. And original anyway costs 2€ less.

 

In any case, why would anybody want to have anything else except the original!?

[PeterSt]

7 hours ago, Miska said:

In any case, why would anybody want to have anything else except the original!?

 

Let me try the commercial blurp : because it has been "deblurred". This shouldn't count really for a native (and well done) 24/96 though, although it can still be audible (why do we upsample to preferably 768 ?).

 

7 hours ago, watercourse said:

 

Comparing local lossless flac and hi res files to streaming non-MQA Tidal Hi Fi files, I tend to prefer local files.

 

Mani plays it locally. He didn't tell much about it for his own reasons (or is challenging perhaps), but indeed this makes the whole world of a difference. "We" use our own filters (so no MQA filters or nothing a tidal desktop app etc. would do or omit). And please, this is all just "test phase" and it certainly isn't so that a whole community now thinks that MQA is for the better, but this is merely because relatively hardly anyone tries ...

 

8 hours ago, watercourse said:

Interesting feedback. The Phasure is a non-MQA DAC, correct?

 

Allow me ... Correct. So once and again the hardware (filtering etc.) part is done in software. Nothing new for "us" at Phasure. Also announced in advance (to MQA). It takes quite some reasoning to get it right though. And not finished, I'd say.

 

8 hours ago, watercourse said:

MQA files often (over half the time) sound like the files have received more processing once comparing to the same non-MQA tracks.

 

They have. Call it the process of deblurring (and no, please, I am not the commercial wise guy - only telling what I have been telling a 100 times by now, in here).

 

8 hours ago, watercourse said:

This is most evident IME in slightly attenuated treble extension and energy.

 

This is exactly the other way around but the most hard to explain, reason out or even prove. Later today I will try to reason out a few things in the Lush thread (in answer to questions of Teresa in there). It can only be an attempt to explain how things work (out) because it is not supported by measurement.

 

10 hours ago, manisandher said:

I started by comparing the various 'Mozart: Violin concerto in D major - Allegro' free download files from 2L.

 

Mani, allow me ... this is exactly what one should not do. This is too much under control of the MQA and 2L combo and therefore should not be a showcase. It is too easy (for MQA ltd) and also IMHO not the example of music which contains everything and all (although one may think that a violin is a difficult beast).

By pure coincidence last night I played Yes - Fragile. Yes, the one and only you and me had a debate over, whether the 16/44.1 was indistinguishable over the 24/192 (or is it 24/96 ?) and where you won with the Hires version (we listened to it both in the same room). So try that one. Don't compare - just listen to it. Already for the so many things you never heard before (e.g. in Roundabout) you like the MQA over the other version(s). But, this is in absolute sense. What happens when comparing I don't know. Point could be that you just like it all over and that nothing disturbs and that there's nothing much reason to shut it off.

[PeterSt]

For a longer time by now I wanted to start a thread like this myself, but didn't because it would be the opinion of one person only (me) without backup of anyone just because there wasn't any possibility to listen to the same (plus I would be regarded biased for obvious reasons). Theoretically this possibility exists now but it still is quite "theoretically" indeed; It requires too many distinct prerequisites which not many people will like. But say that it is a start.

Possibly it is so that Mani himself does not realize that the prerequisites exist - or at least I don't see him talk about anything. Or maybe that is his style. But you can't "just" switch on the Tidal Desktop App, select some MQA album and have it for the better. No way (and not even smart of MQA to allow for this). You also can't say "hey, let's have some random MQA capable DAC and go for it". The reason is obvious : you must like the DAC's nature first and for the same reason we all mostly use different DACs, a random MQA DAC also not is the one you will like over your otherwise preferred DAC. So here too, not so wise decisions (of MQA) if you ask me. Unless of course their target is that in a few years everything and all is MQA capable. But we do not like that idea much, right ? (and here I certainly am biased and I hope you can see in what direction ).

 

So the key thing is to create distance to all the things we don't like to be forced into, but still perceive a better SQ from albums we know for a long time. And that without trespassing rules. Heck, it takes me 10 minutes already to write down this single sentence.

Speaking about forcefields ...

[testikoff]

That Mozart track has some in-band filtering applied to MQA-ed audio (B; decoded + rendered version from Archimago's test) compared to its DXD source (C):

 

- Spectra + delta:

 

- Spectrum delta (close-up):

[PeterSt]

32 minutes ago, testikoff said:

That Mozart track has some in-band filtering applied to MQA-ed audio

 

Not disagreeing. But envision that MQA's processing is processing for a reason - it changes something. Right ?

This happens prior to us receiving what ever version in which ever playback software, or more hardware streaming means if you prefer to look at it like that. Prior to first unfold.

 

So you can encounter anything in there.

[watercourse]

9 hours ago, PeterSt said:

 

Not disagreeing. But envision that MQA's processing is processing for a reason - it changes something. Right ?

This happens prior to us receiving what ever version in which ever playback software, or more hardware streaming means if you prefer to look at it like that. Prior to first unfold.

 

So you can encounter anything in there.

 

@testikoff's post seems consistent with my listening experiences, e.g. treble characteristics of MQA files.

 

So, for what reasons other than faith, would a listener be compelled to think that the MQA version is better, more accurate, or otherwise sonically beneficial? This is an honest question, as no white papers or other technical bases are available from MQA themselves at this time.

[PeterSt]

8 hours ago, watercourse said:

So, for what reasons other than faith, would a listener be compelled to think that the MQA version is better, more accurate, or otherwise sonically beneficial? This is an honest question, as no white papers or other technical bases are available from MQA themselves at this time.

 

I'll honestly try ...

 

Accurate does not exist in this world. Unless ... unless it is proposed the same as I do it myself and we relate it to time accuracy (at the cost of frequency accuracy). I don't say this since yesterday but for 10 years by now.

Time accuracy would be crucial when we observe MQA, because white papers or not, MQA provokes exactly that. So this is about the famous "deblurring" and despite no CA reader believes it exists, take it from me that it does. To judge this, never look at any output which people create from it, because all what can be done is looking at output with MQA filters active (also see below) which are of nobody's choice, or with no filters active which is even more poor. So what to look at then ? well, a net result we created ourselves. What goes in is "deblurred" music data (say it counteracted A/D necessary filtering) and what goes out is determined by ourselves. With this approach you may start to see how things can improve for real.

May it help : you are talking to the person who undertakes everything and all to let the Operating System process nothing whatsoever, if possible, because all such processing influences SQ. So go figure how this same person would even be capable of digesting the processing involved here.

 

Interesting for you could be the notice that when MQA is played back natively (so no unfold) and where MQA makes us believe that this is already "better", this gives a painstaking result. Horrible. BUT, use a playback chain which does not do anything to the stream, like already your DAC will do (mine does nothing to it - see below). Point here is, this "is" better indeed, but does not sound better at all. The file (stream) has been prepared for us, but we need to decode it and decently filter it.

This thinking is our base.

 

So we don't listen to the same (Mani btw does listen to the same as I do). You are not using XXHighEnd which is crucial because of the matching filtering and you are not using the same D/A converter which does not contain any filtering (but accepts our filtered output up to 768K).

FYI : You most probably use Audirvana's  iZotope filtering and there is no single way that can output the same as XXHighEnd's filtering (with the explicit notice that we tried to make it the same to the benefit of the Phasure NOS1 DAC).

 

MQA's output shows more highs, especially in a more "firm" fashion. Do not confuse this with the graph testikoff showed (which shows a roll off towards the highs) because it does not work like that at all. Btw, yesterday in the Lush thread I tried to elaborate on this somewhat. Highs (so to speak) are set by the reconstruction filter and how that works out (for perception) 100% depends on the working of that filter. And for example, if we take an apodizing filter, which is a filter which rolls of more than a non-apodizing filter, it will show you MORE highs. This is all related how transients (but micro transients) work out, opposed to smear of a not-so-good filter, still assumed we take time accuracy for a base. It's a bit of an upside down not so intuitive world.

I think it is fair to say that MQA applies the apodizing filter in the file (stream) already, but this would only be a logical conclusion from my hand. I don't think this can be tested (look at the not unfolded file gives unreconstructed garbage (I say) and when we look at the unfolded version de decoder can have applied the apodizing filter on the fly).

 

To end a long story so it remains short : I myself make the filters and I make them match to MQA. Now at least on my own side nothing remains for having faith. Of course it is quite a thing to now have faith in me. And trust my ears etc. (and my system).

But we are not there yet. I mean, don't try Led Zeppelin because you will run away screaming. This looks to be related to "first trials" (via Tidal and HiresAudio.com). But it is also allowed to think that what doesn't work out today, may work out next year (improve the payback software).

 

Peter

[soxr]

MQA files vs DXD source files from which they were derived, download from the 2L.no site

- MQA: illusion of more echo & reverb, voices loose their raw quality, become polished. Everything more thinned.
- DXD: more real, more body

Tested on Mytek Brooklyn + Vitus RI-100 amp + Amphion studio gear (Two 18) in nearfield setup.

[watercourse]

10 hours ago, PeterSt said:

 

I'll honestly try ...

 

Accurate does not exist in this world. Unless ... unless it is proposed the same as I do it myself and we relate it to time accuracy (at the cost of frequency accuracy). I don't say this since yesterday but for 10 years by now.

Time accuracy would be crucial when we observe MQA, because white papers or not, MQA provokes exactly that. So this is about the famous "deblurring" and despite no CA reader believes it exists, take it from me that it does. To judge this, never look at any output which people create from it, because all what can be done is looking at output with MQA filters active (also see below) which are of nobody's choice, or with no filters active which is even more poor. So what to look at then ? well, a net result we created ourselves. What goes in is "deblurred" music data (say it counteracted A/D necessary filtering) and what goes out is determined by ourselves. With this approach you may start to see how things can improve for real.

May it help : you are talking to the person who undertakes everything and all to let the Operating System process nothing whatsoever, if possible, because all such processing influences SQ. So go figure how this same person would even be capable of digesting the processing involved here.

 

Interesting for you could be the notice that when MQA is played back natively (so no unfold) and where MQA makes us believe that this is already "better", this gives a painstaking result. Horrible. BUT, use a playback chain which does not do anything to the stream, like already your DAC will do (mine does nothing to it - see below). Point here is, this "is" better indeed, but does not sound better at all. The file (stream) has been prepared for us, but we need to decode it and decently filter it.

This thinking is our base.

 

So we don't listen to the same (Mani btw does listen to the same as I do). You are not using XXHighEnd which is crucial because of the matching filtering and you are not using the same D/A converter which does not contain any filtering (but accepts our filtered output up to 768K).

FYI : You most probably use Audirvana's  iZotope filtering and there is no single way that can output the same as XXHighEnd's filtering (with the explicit notice that we tried to make it the same to the benefit of the Phasure NOS1 DAC).

 

MQA's output shows more highs, especially in a more "firm" fashion. Do not confuse this with the graph testikoff showed (which shows a roll off towards the highs) because it does not work like that at all. Btw, yesterday in the Lush thread I tried to elaborate on this somewhat. Highs (so to speak) are set by the reconstruction filter and how that works out (for perception) 100% depends on the working of that filter. And for example, if we take an apodizing filter, which is a filter which rolls of more than a non-apodizing filter, it will show you MORE highs. This is all related how transients (but micro transients) work out, opposed to smear of a not-so-good filter, still assumed we take time accuracy for a base. It's a bit of an upside down not so intuitive world.

I think it is fair to say that MQA applies the apodizing filter in the file (stream) already, but this would only be a logical conclusion from my hand. I don't think this can be tested (look at the not unfolded file gives unreconstructed garbage (I say) and when we look at the unfolded version de decoder can have applied the apodizing filter on the fly).

 

To end a long story so it remains short : I myself make the filters and I make them match to MQA. Now at least on my own side nothing remains for having faith. Of course it is quite a thing to now have faith in me. And trust my ears etc. (and my system).

But we are not there yet. I mean, don't try Led Zeppelin because you will run away screaming. This looks to be related to "first trials" (via Tidal and HiresAudio.com). But it is also allowed to think that what doesn't work out today, may work out next year (improve the payback software).

 

Peter

 

@PeterSt: Thanks for your honest input. There is still a large proportion of "trust me, it's better" that remains, when I would prefer MQA to be more transparent about the why and how of deblurring. That said, I can see how some people could simply prefer the MQA processing and be satisfied without further explanations. I don't fall into that group, but I certainly don't expect you or anyone else that is not behind MQA to do any more explaining.

 

There are inconsistencies in the results of the current MQA catalog, e.g. some albums simply sound inferior to non-MQA'ed albums, that I feel deserve some explanation. Is it poor quality control? Who does the processing, and how are results checked? Will there be some sort of quality standard that is achieved at some point in the future? These factors should be understood by consumers, IMHO.

 

And I agree with @soxr that I've heard some tracks that sound more like movie soundtracks than their non-MQA counterparts. Reminds me of a well-implemented version of Q Sound processing (which I'd call gimmicky, but again, it's implemented well). I don't necessarily want my music on a big screen...

[PeterSt]

@watercourse, I noticed in one of the other threads that it requires emphasis on the fact that I am quite explicitly not talking about Hires material. So if you see me explain how the MQA workflow looks like (or should look like) like I did in here :

https://www.computeraudiophile.com/forums/topic/37552-mqa-off-topic-spinoff/?do=findComment&comment=746622

then when you once know I am talking about Redbook you can easily see it by the logic of what's being done. In other words :

 

I don't think I'd dare say that MQA will surpass Hires. Very maybe when it is 24/96 but I don't think so. Point is also that I don't know of any published 24/96 (or 24/192 for that matter) which are genuine to begin with and which exist in MQA. And those I do know I find failed to begin with. Otoh I now recall some MQA's passing by which I did not know before, which sound good and which I maybe have or can find in normal Hires. Alice Cooper comes to mind with Peter & The Wolf in Hollywood .... but I already see I don't have it in normal Hires. Maybe I must give Welcome to my Nightmare another try (I have that and it exists in MQA too), but I recall it sounds bad to begin with.

 

Anyway I can't repeat enough that the 2L examples are the worst because they are just too good, an almost exclusively recorded in DXD (24/352.8). There's a universe of distance between those and any other Hires that I know of (which are almost always "wrong" for one or the other reason). "Too good" in this sense means : of too high genuine sampling rate which does not require any improvement that I can think of (please read into the link I gave) BUT it will be different. It will be different already because 2L explicitly does not filter anything in their DXD recordings (again see that link and see me telling about filtering from ADC to digital which 2L thus not does) while the MQA process requires filtering and with that always destroys what 2L wanted to preserve (justified).

[Archimago]

12 hours ago, PeterSt said:

...

I don't think I'd dare say that MQA will surpass Hires. Very maybe when it is 24/96 but I don't think so. Point is also that I don't know of any published 24/96 (or 24/192 for that matter) which are genuine to begin with and which exist in MQA...

 

@PeterSt

 

Interesting discussions. Obviously I cannot speak to your filtering techniques and your DAC.

 

But I would argue that MQA cannot surpass the quality of high-resolution PCM including 24/88, 24/96, etc. 

 

MQA reduces the bit-depth resolution. MQA also reconstructs frequencies >22/24kHz with a lossy variant when unfolded. Sure, we can argue that humans might not be able to hear these low levels of degradation. But there's ZERO evidence as well that what MQA purports to improve - time domain "deblurring" even works or makes any difference.

 

Hilarious that supposedly impartial listening tests at McGill University isn't even scheduled to begin until January 2018... Gee, when will the data be analyzed and presented???

 

There're really nothing interesting going on with MQA technically IMO. And if Bob Stuart actually had something interesting, he could argue for it as a separate DSP/filtering algorithm without need for all this "unfolding", hardware firmware upgrading, and DRM nonsense.

 

We argue now about politics, ethics, questionable marketing, and ridiculousness of an incompetent audiophile press incapable of serving the interests of "audiophiles" or providing leadership to advocate.

[PeterSt]

7 minutes ago, Archimago said:

But I would argue that MQA cannot surpass the quality of high-resolution PCM including 24/88, 24/96, etc.

 

I read this as "MQA can surpass the quality of high-resolution PCM etc." (makes it better).

But you can't be saying that ... Must be my English.

?

So if you don't say that indeed, I fully agree. But then I already told so.

 

Thanks.

[Archimago]

19 minutes ago, PeterSt said:

 

I read this as "MQA can surpass the quality of high-resolution PCM etc." (makes it better).

But you can't be saying that ... Must be my English.

?

So if you don't say that indeed, I fully agree. But then I already told so.

 

Thanks.

 

IMO MQA CANNOT be better than PCM 24/88, 24/96, etc... from the perspective of resolution provided by the file type.

 

Of course one could subjectively prefer the sound of a remastered track that went through the MQA encoder, and some people might even prefer the resolution reduction / dithering / noise shaping / filtering that MQA performs (just like many people prefer the sound of vinyl even though technically there are clear limitations). Again, technically, MQA still reduces potential resolution regardless of whether someone prefers the sound. That is simply a fact which IMO cannot be disputed.

[PeterSt]

@Archimago, why all the emphasis while we agree and I was first with it ?

 

On 22-11-2017 at 5:34 PM, manisandher said:

but ask that people refrain from turning this into yet another ' MQA is evil' or 'Bob Stuart is the devil' type of thread. Much appreciated.

 

[Archimago]

8 minutes ago, PeterSt said:

@Archimago, why all the emphasis while we agree and I was first with it ?

 

 

 

Because I want to be clear in no uncertain terms - sometimes I cannot understand what you're saying and hope readers do not have misconceptions especially when the paragraphs are dense and seem complex.

[Archimago]

Just as an addendum to the statements...

 

Sure, we could "make" MQA sound good. But why should we?

 

We know that MQA actually tarnishes the quality of a pristine 24/88 or 24/96, etc. Why should we even bother entertaining this monstrosity at this "early" state when all the company has is the smallest paid streaming service (TIDAL) supporting it?

 

Truly, nothing good can come out of MQA adoption. Fight to abort this nonsense in the embryonic phase. Encourage people who record/mix/master to do the job well at the highest fidelity they're aiming for and ignore this MQA silliness. Keep the message simple and the vision of what hi-fi is about as clear as possible...

  

[PeterSt]

You wanted to emphasize what ?

[Archimago]

9 minutes ago, PeterSt said:

You wanted to emphasize what ?

 

Emphases can be used for clarity.

[PeterSt]

I am thinking of the best way to blur you. But my English isn't good enough.

I'll call in the MQA's. 

[Archimago]

56 minutes ago, PeterSt said:

I am thinking of the best way to blur you. But my English isn't good enough.

I'll call in the MQA's. 

 

Yet again, I have no idea exactly what you mean to say...

 

Given all the uncertainties, innuendos, claims from MQA Ltd. and Bob Stuart, my feeling is that when discussing MQA, we have to be clear about what we're talking about. MQA has been vague from the start and if we are to add to the discussions, meanings, intentions, procedures, etc. need to be precise especially when making claims about "listening impressions" and generalizations we're trying to make. De-blur the message.

[Fair Hedon]

2 hours ago, Archimago said:

 

Yet again, I have no idea exactly what you mean to say...

 

Given all the uncertainties, innuendos, claims from MQA Ltd. and Bob Stuart, my feeling is that when discussing MQA, we have to be clear about what we're talking about. MQA has been vague from the start and if we are to add to the discussions, meanings, intentions, procedures, etc. need to be precise especially when making claims about "listening impressions" and generalizations we're trying to make. De-blur the message.

Precisely!!

 

The amount of "output of the bull: that Stuart and his minions have slung around is astounding.