Trying out MQA for classical listening, questions and observations

[crenca]

6 minutes ago, Markpd said:

 

So if I listen to two tracks on Tidal both released at the same time, same album artwork, and same track timing, and the MQA version, software decoded, sounds clearly better, what should I assume about the masters used?

 

 

I would check the wax in your ears ?

 

 Seriously though, when it is clear the masters are the same only a handful of audiophile journalist and audiofools dare use the words  "clear difference".  The difference is subtle at best, and a matter of preference as to whether it is even "better".  

[mav52]

I find it interesting two MQA albums, same artist, same cover art but when played my device notes

 

Miles Davis Quintet '' Cookin"

 

one :  24/192 Studio

the other:   24/96 Studio

 

Then you have a non MQA remastered which plays at 16/44

 

So one ask a question, what master did MQA use.

[knickerhawk]

34 minutes ago, crenca said:

 

 

I would check the wax in your ears ?

 

 Seriously though, when it is clear the masters are the same only a handful of audiophile journalist and audiofools dare use the words  "clear difference".  The difference is subtle at best, and a matter of preference as to whether it is even "better".  

“Subtle” and “clear difference” are not mutually exclusive. In fairly extensive comparisons I’ve done over the past month, I’ve come to the conclusion that the difference is all three: subtle, clearly audible and better.

[crenca]

39 minutes ago, knickerhawk said:

“Subtle” and “clear difference” are not mutually exclusive. In fairly extensive comparisons I’ve done over the past month, I’ve come to the conclusion that the difference is all three: subtle, clearly audible and better.

 

You know you're right.  I forgot this is Audiophilia we are speaking, where "subtle", "clear difference", and "a birth of a new world" all mean the same thing 

 

You know, the kind of differences that lead to coin flip chance in a blind... 

 

[knickerhawk]

14 minutes ago, crenca said:

 

You know you're right.  I forgot this is Audiophilia we are speaking, where "subtle", "clear difference", and "a birth of a new world" all mean the same thing 

 

You know, the kind of differences that lead to coin flip chance in a blind... 

 

That’s all you’ve got? A couple of emojis and some tired anti-audiophile-double-blind smack? My guess is that you haven’t even done any comparisons for yourself because, well...real audiophiles know a priori that MQA is just hype. No auditioning required...

[Kal Rubinson]

2 hours ago, knickerhawk said:

“Subtle” and “clear difference” are not mutually exclusive. In fairly extensive comparisons I’ve done over the past month, I’ve come to the conclusion that the difference is all three: subtle, clearly audible and better.

I'd say subtle but consistent.

[rodrigaj]

3 hours ago, firedog said:

The bridge is a card in the box of the DAC is it not? So I think it’s merely a semantic distinction. As far as I’m concerned it’s still in the DAC. It depends how you define DAC-as merely a chip or an FPGA, or as the box of which it is a part.

 

Even with the DS, the full unfold and filtering requires a HW solution, and that solution Is only found in the DS.

 

 

 

You can buy a PS Audio DS DAC without the Bridge II in which case you would not get the full unfold. The hardware solution is in the Bridge II not the DAC.

 

Those who bought a DS thinking they would use a renderer such as a Sonore rendu in place of the BridgeII would not get the full unfold.

 

That was the only point I was trying to make. As you said, to most it's a semantic distinction.

[crenca]

3 hours ago, knickerhawk said:

That’s all you’ve got? A couple of emojis and some tired anti-audiophile-double-blind smack? My guess is that you haven’t even done any comparisons for yourself because, well...real audiophiles know a priori that MQA is just hype. No auditioning required...

 

Compared, reviewed here, etc. .  Look, we can go back and forth all day long Mr. 27 posts, but the consensus here is that the difference is subtle, subjective, unclear (seems to matter on some recordings, not on others) etc. etc.  Thus, even as a SQ tweak MQA is a failure.  Your subjectivist, art and wine, audiophile  testimonial hits all the right buttons in the evangelist tent (and in that sense it IS "real" audiophiledom - fits the script exactly)....so no, I don't believe you.  I know you believe you, but then those who tape crystals on their cables are true believers as well and as such real audiophiles...   

[wklie]

6 hours ago, mav52 said:

Miles Davis Quintet '' Cookin"

 

one :  24/192 Studio

the other:   24/96 Studio

 

According to meridianunplugged forum, such duplicates could have different regional availability and could have come from different releases with different original sample rates of the masters.

 

By the way, with duplicates included we now have 10000+ MQA albums on Tidal.

[PeterSt]

11 hours ago, firedog said:

You don't get the deblurring - for that you need an MQA DAC.

 

Try to avoid spreading misinformation; this is not correct at all.

The deblurring happened in a process(ing software) and is contained in the file (you obtain from Tidal or from elsewhere).

 

11 hours ago, firedog said:

I'd be quite surprised if they are upsampling Redbook in order to make hi res MQA.

 

From all what I (explicitly) see, I can only agree with this.

 

16 hours ago, Norton said:

Or in either case am I just hearing an upsample of the RBCD release?

 

So no, not that.

 

16 hours ago, Norton said:

For example, where I'm listening to a legacy recording only issued digitally to date as RBCD, am I now hearing a fresh master at the resolution of the original recording?

 

My personal take on this :

The resolution as given to you by the MQA incarnation of the album (or track) exists somewhere in reality. Thus, if you receive a 192, then the album exists for example on DVD Audio in 192. Remember, you can't easily see this without MQA DAC, but software could tell you (like I explained/showed about my own). BUT :

I don't think any rate higher than 96 will give you that resolution for real via MQA. IOW, *that* will be an upsampling step from 88.2 or 96. The genuity of highes higher than 96 is also not promised by MQA (if you read all closely). So any 176.4, 192, 352.8 are upsampled from 88.2 / 96 and that by means of the by MQA provided filters applied by the MQA DAC (renderer). And to come back to Firedog before het gets too confused  this is not deblurring - instead this is preventing from blurring again (don't let the DAC apply ringing, while the ringing was just taken out (in the file given to you - this is the prospect of MQA ltd).

 

The conclusion of this latter and your question, Norton :

- If you receive a 88.2 or 96 (as how the file natively is, as told by MQA decoding), this will be genuine (for its source hence base of encoding, which is not the same as lossless compared to this original source).

- If you receive a 44.1 or 48, this also is genuine (to the master it was taken from).

- If you receive (render !) a 176.4/192 or higher, this is upsampled from 88.2/96 and possibly the 88.2 or 96 never existed as a master (possibly : the MQA encoding process may have downsampled from 176.4/192).

 

There is a lot more to this, because all the hoopla from the not far past of what is hires and what is fake to begin with, is in order. And Norton, I reckon you feel this (without expressing it really). So the 176.4 might be from DSD (which was PCM'd somewhere) and a 96 may be a downconverted 5.1 DVD, same for a 48. And so much more which always has been the issue with Hires. One thing :

I can almost guarantee that none of the HD Tracks upsampled sh*t is taken for the base for MQA encoding. So if you exclude all that fake stuff, what remains is what exists somewhere for real (Hires) and that can have been given to the MQA encoder. This includes masters which never say daylight so far (for whatever reason).

 

 

[PeterSt]

4 hours ago, crenca said:

Look, we can go back and forth all day long Mr. 27 posts, but the consensus here

 

Look mr junior objectivist, maybe it is time to get yourself some education. You already told me to get out of your life because of my poor English, but if now you are telling that you just be infinitely better because of your 1730 posts opposed to 27 of someone else, then WHAT are you. And no, this was not half-baked English.

 

The consensus here is nowhere, except for your theorist terrorism without even listening.

[PeterSt]

9 hours ago, mav52 said:

So one ask a question, what master did MQA use.

 

Generally the one with 100% the same track times, up to 1/100 of a second. This is easier said than done of course, as you need to have the references to compare with.

I have quite many albums and often for an MQA in my genre, I have 10 other versions at hand. Say that in 10% of all cases I try to compare, I find a matching album ("master") and thus know the source.

 

In not any occasion I found no differences to be there. The difference is always major for the perception you get from it. I must emphasize once again ... the perception. This is not about technical SQ as such. Thus, not better highs or better bass or better resolution or whatever. It is about how the music speaks to you (addresses you). It works differently. This is not subtle at all.

- SQ as such if often (technically) worse (this is major for differences, like less refined highs);

- The perception of the lot is liked better (the "subtlety" of this is sneaky but overwhelming in the end and therefore major again).

"Is liked" : by subjective me.

 

Not to forget please : I use a NOS DAC (the perfect match for MQA to begin with) and I use my own "rendering filters" (say that these are subjective to our likings and they replace the "weird" filters of an MQA DAC).

So it is all not easy to compare (unless we use a NOS 768 capable DAC and same in-software filters).

[firedog]

2 hours ago, PeterSt said:

Try to avoid spreading misinformation; this is not correct at all.

The deblurring happened in a process(ing software) and is contained in the file (you obtain from Tidal or from elsewhere).

 

We are talking about 2 different things: I was referring to the deblurring/blurring prevention on the DAC end. You are referring to the deblurring on the ADC end. You don't get the full "benefit" of MQA without the MQA DAC- so you aren't hearing the same thing you would hear if your DAC was an MQA DAC.

[firedog]

Peter: question about MQA and sample rates of 176 and above. 

It's been stated/implied that no MQA file actually gives you a file of this resolution, even after full unfolding. Is this correct?

Can you explain a bit more what is going on? What happens when they apply the MQA process to a 176 or 192 file? 

[GUTB]

Okay guys, I took a stab at it.

 

I'm tired of the 2L tracks. We've all heard them a billion times. Besides, I'm not a huge fan of how they do recordings. So I went looking for something different. I found this on hiresaudio:

 

https://www.highresaudio.com/en/album/view/3jihm2/unamas-strings-septet-p-i-tschaikovsky-op-70-souvenir-de-florence

 

 

Tschaikovsky op - 70 Souvenir de Florence by a Japanese audiophile label, UNAMAS. This was a fairly recent release meaning they are going to use a newer MQA encoder. Their specialty is doing nine-channel recordings, but according to the booklet they also release a stereo version and an MQA version -- this provides us a reasonable level of assurance that the stereo FLAC and the MQA version both comes from the same source, and that the same mastering engineer is doing both. The equipment, venue, and setup are all documented in the booklet. Finally, they appear to master in 24/192; this is important because I wanted to be able to compare something better than 24/96.

 

I bought both of them:

 

 

For this critical listening session I sit down at my headphone setup consisting of Pro-Ject S2, the best native MQA-capable DAC I have. The Pro-Ject is connected to a Mjolnir 2 headphone amp with a pair of nice Amaprex Bugle Boys via a set of Belden+Neutrik RCA to XLR cables (World's Best Cables from Amazon)-- I do this because the Mjolnir's single ended input is compromised in terms of SQ I believe. Headphones are my very fine Fostex TH900, superb headphones for classical. 

 

First I listened to the last track, Allegretto moderato in the 24/192 version. I pick the last track because the last track in a score is usually the most impressive. Very nice recording. I can tell there is a decent amount of stage present, even though the sense is severely limited by the fact that its headphones. Very much this is a  high-definition audio recording; high separation of instruments, micro-detail and harmonic structures are present. The Pro-Ject is weak in dynamic force / transient swings, but I'm aware of those shortcomings so I don't hold that against the recording; where the Pro-Ject is strong is in clarity and detail retrieval, and there's no hint I'm listening to anything less than a high resolution audio file. After a minute or so I start thinking to myself that it's unlikely the MQA will be noticeably better.

 

So then I move onto the same track in the MQA version. Interestingly, the Pro-Ject lights up green, not blue. Well, whatever, I guess that's fine. Roon is showing my lossless output stream, so the Pro-Ject should be doing the regular MQA unfolding processes. Listening to it, my first thought was: yup, it's the same. But, when I reached the first big movement, I went "huh?" Was it really the same? I listened to the end, becoming increasingly sure they WEREN'T the same, so I went back to the 24/192 version and queued up at a spot I thought would be especially revealing -- and DAMN was the difference obvious! It was if the violins were leaning into the mics in the MQA version -- they popped out with much greater clarity and distinction from stage, more vibrancy and details. The SAME DAMN phenomena I've noticed with some other MQA vs non-MQA comparisons I've done on Tidal -- it's as if there is someone playing with the EQ on the MQA version. The 24/192 version seems bland and shallow in comparison now that I understood what I was listening for.

 

Actually, I should have gone with the first track -- it starts off strong with a violin, and it's immediately apparent how much better the MQA version is! Is this really time domain correction in action?

[Norton]

4 hours ago, PeterSt said:

 

The conclusion of this latter and your question, Norton :

- If you receive a 88.2 or 96 (as how the file natively is, as told by MQA decoding), this will be genuine (for its source hence base of encoding, which is not the same as lossless compared to this original source).

- If you receive a 44.1 or 48, this also is genuine (to the master it was taken from).

- If you receive (render !) a 176.4/192 or higher, this is upsampled from 88.2/96 and possibly the 88.2 or 96 never existed as a master (possibly : the MQA encoding process may have downsampled from 176.4/192).

 

Thanks Peter,

 

So to take a possible analysis based on some recordings I am familiar with outside of MQA,:

 

the Daniel Barenboim Elgar symphonies 1 and 2, when unfolded, are likely to be the genuine 24/96 files just as I bought previously as downloads, but with "added MQA".  I think some have suggested that files which are "natively" 24/96 are the sweet spot of MQA - maximum resolution and fidelity without recourse to the latter MQA stages. If this is the case then MQA makes sense, as a tool to allow glitch free streaming at this resolution.  But I guess Qobuz can also do this with std flac?

 

Mozart Violin Concertos or Ole Bull violin pieces on 2l, which are sold as DXD downloads ( and presumably recorded as such) , when processed by a MQA DAC will also unfold to 384.  However in this case the MQA file may be different, the master possibly created by downsampling  to 24/96 and then upsampled to 384 in the DAC?  

[PeterSt]

1 hour ago, firedog said:

You don't get the full "benefit" of MQA without the MQA DAC- so you aren't hearing the same thing you would hear if your DAC was an MQA DAC.

 

The "benefit" emerges when you tell your DAC not to apply its usual filtering = ringing. This already happens by feeding it the higher resolution instead of Redbook. So if all is right, the DAC will now skip its internal filtering (= ringing) and just pass through the samples.

Now we must watch out, because when the DAC is 192 capable it is already sufficient to not let it apply ITS filtering by means of giving it a 192 file. Feel the trick ?

What happens further is that MQA applies their thought best filters because else no filter was applied and it should be better with filter. Watch out again, because *if* such a filter is in order, it will be in order on anything below 176.4 hence exactly what MQA presents you (44.1 up to 96).

 

All what remains now is thinking how that filtering could be applied. Well, this is in-DAC by MQA firmware or in-software by our own. And not to forget, filtering always happens, be that by your non-NOS DAC or be that by your playback software and its filtering - or a bit of both.

Where it goes sub-optimal is when MQA presents you the 88.2/96 unfolded file and your DAC is 96 capable only. Now nothing can happen further and the last piece of rendering (filtering !) which is "mandatory" anyway, lacks. What's happening extra-wrong with MQA is that the transients/frequency rings extra-few while nothing is capable to reconstruct it to mere normal behavior (this requires upsampling but no upsampling headroom is present in this case (DAC is 88/2/96 capable only)). So even MQA ltd intends that (it first goes back to base and removes all the blurring as far as possible and FWIW) while next it depends on your DAC (MQA ltd says) what must be done to normalize it again (and this differs per DAC, again, as MQA ltd says).

There's more to say about this, but it is sufficiently difficult to follow as it is, I'm afraid (and yes, this involves my English).

[PeterSt]

17 minutes ago, Norton said:

will also unfold to 384.  However in this case the MQA file may be different, the master possibly created by downsampling  to 24/96 and then upsampled to 384 in the DAC?

 

Yes, that's what I was saying.

 

Btw DXD is 352.8. So if we receive the 2L files as 96 (do we ? ... which will "unfold" to 384), then also resampling took place. I am not 100% sure of this, but I think I have seen examples of Redbook which is 44.1 which was transferred to 48 by MQA. If I observe this correctly then other (good) trickery is at hand, but this is not for this thread (and no secret, I talked about this elsewhere). This is not to be confused with "The Nightfly" example, because this exists on DVD in 48 with a 100% match with that as the used master.

[semente]

1 hour ago, GUTB said:

Okay guys, I took a stab at it.

 

I'm tired of the 2L tracks. We've all heard them a billion times. Besides, I'm not a huge fan of how they do recordings. So I went looking for something different. I found this on hiresaudio:

 

https://www.highresaudio.com/en/album/view/3jihm2/unamas-strings-septet-p-i-tschaikovsky-op-70-souvenir-de-florence

 

 

Tschaikovsky op - 70 Souvenir de Florence by a Japanese audiophile label, UNAMAS. This was a fairly recent release meaning they are going to use a newer MQA encoder.

 

Wait, there is a "newer" MQA encoder?

Does this mean that MQA files will sound different/better depending on the age of the encoder?

[PeterSt]

2 hours ago, firedog said:

It's been stated/implied that no MQA file actually gives you a file of this resolution, even after full unfolding. Is this correct?

 

I can not state this is correct, but I would agree with it, yes.

 

2 hours ago, firedog said:

Can you explain a bit more what is going on? What happens when they apply the MQA process to a 176 or 192 file?

 

Again I will not state this as the truth, and can only give my perception from the consistent whole :

 

The 176.4/192 is downsampled to 88.2/96 so it can be made ready for transport (streaming) of 24/88.2 - 24/96. This is packed into 24/44.1 - 24/48 and there will be no more bits utilized than 17 for the unfolded to 24/88.2 - 24/96. So indeed, the file which could have been e.g. 24/96 now is 17/96 only. Technically it is 24/96 but functionally it is 17/96. So this is not Hires ? -> MQA ltd does not care and does not say it is. All MQA ltd says is that for temporal resolution it is so. And since this is about the "96" part of it, this is no lie.

 

Because what happens after the first unfold is an upsampling step, anything could happen. What in any event does *not* happen is that this is a to the frequency domain decent filter. Of course this fits MQA ltd's strategy because they like the time domain.

Why do I say this ? well, because it is so easy for us all to think that genuine Hires should avoid all the hassle, WHILE it is not about Hires at all. It is about that deblurring thing ...

Of course nobody believes him, but this should be an example of it :

 

1 hour ago, GUTB said:

Okay guys, I took a stab at it.

 

I think he really tried there as a sort of confirmation testimonial for himself.

[PeterSt]

9 minutes ago, semente said:

Wait, there is a "newer" MQA encoder?

 

My take : Possibly this has been stated elsewhere (Gearslutz perhaps) and otherwise it could be a conclusion. Anyway what has been proven at least to me, is that over time different MQA versions emerged of the same album. As if it had to be re-done and now is better. But please, this would only be my personal conclusion about what I see happening.

It would be a typical streaming thing as the version you can download from hiresaudio.com would (for you) be that one and only you received and play. With Tidal they can just replace it and give you the best version (for whatever that's worth). Odd would be that the older (virtually obsolete) versions remain, for what I can see. In rare occasions albums / MQA versions get deleted from MQA, the Purple Rain compilation being an example (maybe something failed) which was withdrawn and put back a week after or so.

 

If there really was a new MQA decoder, I'd hope that the Led Zeppelin's were re-done. To me they are still the example of failed MQA. Or, Barry Diament made them too well. 

Btw, I bought one from hiresaudio.com. Will I receive an email when a better version is available ? doubtful.

[Norton]

8 hours ago, wklie said:

 

By the way, with duplicates included we now have 10000+ MQA albums on Tidal.

 

Worth noting too that, for Classical at least, in a number of cases one title is actually a "box set" , i.e. the complete symphonies of x.  I think I counted over 250 tracks within one Classical title (not the Ring).  There is some interesting stuff there, classic DECCA recordings and also transfers from the Everest master tape for LSO.  However I reckon there are probably no more than 100 MQA titles to interest me at the moment.  Foghat  look to be well represented though...

 

 

[bhobba]

22 hours ago, Norton said:

in terms of the stated  objectives for MQA , what advantages should I be enjoying from the initial decode?  The claimed temporal de-blurring for example?

 

To understand that you need to go into its detail:

http://www.aes.org/tmpFiles/elib/20180107/17501.pdf

 

Have a look at figure 13.  Its advantage as far as blurring goes lies in the triangle sampling - have a look at what happens to a transient with that sampling - even between samples you get some of the transient.  Also the  effect is a very shallow filter that creates less damage to a spike (technically called a Dirac impulse) than a sharp brick-wall filter that's normally used.  That's the blurring advantage.

 

Notice the line drawn at 16 bits TPDF  - nothing really above about 48k.  MQA just has those 16-17 bits sent with a lot more sophisticated dithering than TPDF - the point though its just 16-17 bits.  Any thing below 16-17 bits isnt even transmitted ie only 48k is transmitted - and that transmits everything at 16 bits anyway - no use going any higher unless you transmit more than 16 bits.  So any reasonable up-sampling will be fine.  But what it does is the reverse of the sampling it used to create it in the first place - for triangle sampling it's simply linear interpolation.  However in practice they use  something more complex than a triangle.  This is theoretically the ideal and why they do it, but it has a disadvantage of creating what are called aliasing components about 30-40db below max level.   Its ultrasonic above 48k - most above 60k - so inaudible - but may cause issues with an amps slew rate and beats due to non-linearity.   Other methods such as up-sampling algorithms like SOX don't have this issue - but are not theoretically optimum.  But since those frequencies are chopped anyway its not particularly important.  MQA claims it is but I have done listening tests and it isn't in what I listened to ie I compared 96K from the first unfold up-sampled in a Direct Stream to the one up-sampled with MQA - not much difference I can tell.   The one that really makes the difference is the up-sampling in HQPplayer - that sounds the best of the lot ie 96k sources up-sampled via HQPlayer sounds better than either MQA up-sampling of that done in the DS.

 

So if you use say Audirvana and use its SOX up-sampling I don't think you will notice much if any difference.  Roon keeps promising MQA decoding and there is a way to feed that into HQPlayer - when that finally comes I expect it to sound best of all.

 

Why is HQPlayer so much better - beats me - ask Miska.

 

Thanks

Bill 

 

 

[bhobba]

1 hour ago, semente said:

Wait, there is a "newer" MQA encoder?

Does this mean that MQA files will sound different/better depending on the age of the encoder?

 

When you read articles about MQA they explain it in terms of triangle sampling.   In fact they use a more sophisticated sampler called a B-Spline.   Also it can be applied to PCM of virtually any resolution eg 12mbs.  They would obviously be experimenting with higher sampling and different splines.  They may have found some combination that sounds better - who knows - this is cutting edge stuff.

 

Thanks

Bill

[mansr]

4 hours ago, firedog said:

Peter: question about MQA and sample rates of 176 and above. 

It's been stated/implied that no MQA file actually gives you a file of this resolution, even after full unfolding. Is this correct?

Can you explain a bit more what is going on? What happens when they apply the MQA process to a 176 or 192 file? 

There are obviously things about the encoder we don't know. One thing we do know is that an input of higher than 96 kHz is downsampled to 88.2 kHz or 96 kHz using a rather leaky anti-alias filter. After this, the signal is split into two bands, 0-24 kHz and 24-48 kHz. The high half is analysed to determine how many bits are needed to encode it. At least 8 bits are always used, a few files I've seen use 9. The high band is encoded and placed in the low bits of the output stream. A 1-bit control/metadata stream is added above this, and the remaining bits (up to 15) are used for the baseband frequencies.