MQA vs HiRez: an apples-to-apples comparison - FINAL

[manisandher]

Thanks to the 9 people who posted their thoughts.

 

I've given a score of '2' for your preference, and a score of '1' for your second choice. Also, you'll see some ticks to the right-hand-side. A black tick shows that you explicitly identified all 3 tracks correctly. A grey tick shows that you probably identified all 3 tracks correctly, but didn't explicitly say so.

 

And here are the results:

 

 

(With my scores added, the totals become 6, 10 and 10).

 

I think we can conclude that the truth between hires and MQA does indeed lie somewhere in the middle!

 

Mani.

[manisandher]

On 10/27/2018 at 1:10 AM, austinpop said:

 

Could you also post a Tidal (or Qobuz) link to the track? 

 

The track was Anouar Brahem - Blue Maqams - Opening Day (MQA on Tidal: https://tidal.com/track/79927158).

 

Mani.

[Rexp]

Good game Mani, thx for organising! My take away is neither hi res nor MQA is the answer to providing (much needed) better sound than redbook. 

[manisandher]

4 minutes ago, Rexp said:

Good game Mani, thx for organising!

 

You're welcome.

 

4 minutes ago, Rexp said:

My take away is neither hi res nor MQA is the answer to providing (much needed) better sound than redbook. 

 

?

 

Mani.

[PeterSt]

 

Mani, 

@manisandher, thank you for this (again). I am not glad that I had it right, but I am with joy that I had it right for the reasons I gave. They were quite explicit and probably show that I recognize the matter indeed. But really, the Hires among it made it difficult, or at least I thought so in advance. That, in aftermath, I appeared to be right about my observations, only testifies that I really don't like Hires most of the time.

Small addition in an upcoming post.

 

[PeterSt]

2 hours ago, manisandher said:

 

What troubles you about this?

 

Thank you for this.

What troubles me is that there is no decay in the highs (towards the higher frequency). This can't exist.

 

Maybe you forgot to show the end cymbals. Of course you could argue that the ride on the (ride-)cymbal shows ever so much more frequency, but ... 45KHz ? No way. And at this (soft ticking) level ? triple no.

The end cymbals oddly enough show about right to me.

https://www.musical-u.com/learn/percussion-frequencies-part-2-cymbals/

 

I did not emphasize anything with n dB and I always compare at this level. So the Greendale I showed - same thing. I never touch the level (in no 10 years of time) thus both are comparable. 

 

Also, I looked into this this morning because if I had to be correct on the how I came to my judgment on the Hires (sounds the worst) then first I had to be correct on it being fake, right ?

 

Peter

 

 

[Le Concombre Masqué]

6 hours ago, PeterSt said:

 

Maybe it is hard to believe for you, but when I was as far as reading the text above (not a word further) I had my answer ready :

 

Probably my filtering, which is a means of transferring the sound to some other SQ level, is doing something which in Hires has not been done (because it is hires already and the filtering is about reconstruction of the band limited signal as such) ... and now can not be done any more by my filtering.

And then I read on, and you said quite the same (with a very different filter, but alas).

 

Although this is an answer, it is also too far out in my own view. Hires just should sound the best, if the hardware is running at the same frequency everywhere, which is so in my case (I always output at 705600 or 768000, depending on the base).

 

 

This is not exactly what I am saying. What I mean is : the Hires made available is most often faked. It is not real hires. It is either or upsampled or it is hires done in an awful way (like tape with so much noise that nothing above 12KHz is there anyway *and* denoisers have been applied which only make it (far) worse). Examples of fame which come to mind : King Crimson - In the court of the Crimson King, or Bill Evans with Waltz for Debby. This, while both are of nothing less than superb quality if you only get hold of the correct Redbook (which is quite a task for either, especially the Bill Evans).

An other category would be the wrong downmixes with the good example of Dire Straits - Brother in Arms. I forgot whether it is the DVD or the SACD (but likely the DVD) which just lacks a couple of channels. Hires is OK, but "mix" is obviously failed.

Similar with the pile of existing quadraphonic recordings, like from earlier Grand Funk (yes, that old quadraphony already is). Many, many exist, but AFAIK all with the wrong downmix. Or this :

 

 

I have been working on downmixing some Grand Funk original myself, but it is obviously impossible because the headroom is lacking. So with 4ch we need to get rid of 6dB while no 6dB headroom (which is half of the available digital headroom) is available. Solution : compress it, or leave out two channels. But normally thinking of the former : now you'll have Hires in a way poorer incarnation. Thus, real hires all right, but the original recording which was mastered for stereo, just utilizes the full headroom. You wouldn't see that on the spectrograph readily, but this time on the DR figure. Plus 6dB is really a lot to "compress out".

All DVD 5.1 which is available in Hires Stereo has been subject to this. I think it was at this forum (maybe 8 years ago) that someone who mastered for DVD in some department, always got the order to give that now (IIRC) 9dB of extra headroom but they never did it for various (logical) reasons. And so we ended up with hires DVD 5.1's which utilized all the headroom, never suitable for downmixing again.

 

FYI : I have a huge pile (think 1000++) of whatever so-called downmixes from DVD which I all had prior to the existence of HDTracks. HDTracks, one by one as we are used from them, put these up, spread over many years (but now they ran out on that pile I already had for many years). They were *all* fake (I still have them) and 100% exactly the same as those put up by HDTracks. And notice that I even talked on the phone with the person who created them and admitted they were a quickie, back at the time. They are official though, unlike what I thought at first. They are NOT suitable for MQA, which is why we see different masters for MQA so often.

Or what about the former producer of Turtle Records over here in Holland, who en public thought he was having fun with the audience with his : hahaha hohoho when someone asks to dig up the master tapes I have no time for that, obviously *if* I know where they are in the first place, so we upsample them and the customer is happy. Ha !! 

It's plainly outrageous.

Btw, Chris C. witnessed that too.

 

The one subject to this thread is also doubtful to these eyes. Straight to 48KHz for this track throughout ?? How ? where ? what exactly ?

 

 

 

This is Hires. 06-Bandit from Neil Young - Greendale DVDA (96).

 

 

So that could be justified.

 

Want to have a fair chance on genuine Hires ? then look at DVDA 192. This won't exist in multi channel (IIRC this just does not fit on the DVD) and not really many of it exist. But those existing are always the real thing that I have seen.

 

Regards,

Peter

 

Thank you Peter,

 

Which versions do you recommend for Court and Waltz?

 

I own a Pink original Island LP of Court and always marveled at its sound, especially of cymbals, and am OK with the 24/96 remaster that you quote as a negative case for hires ; I own a test pressing of the Analogue Productions 45 RPM of Waltz and have always been puzzled that it doesn’t sound as good as Sunday (I also own a test pressing of the Analogue 45 RPM) ; is it the Stubblebine 24/192 that you criticize ? how about the xrcd or the Analogue SACD mastered by Doug Sax?  Those 3 sound quite different but none is bad IMO

 

As of Brothers, I recommend the MFSL SACD

 

@Mani : a good case for MQA might be Amy Winehouse : sounds better (through HQP’s mqa filter ) than the original LP and 24/96, those 2 never satisfied me ; so maybe something has been made right with the MQA ; a different master?

 

Regards,

Le Concombre

[austinpop]

2 hours ago, manisandher said:

 

The track was Anouar Brahem - Blue Maqams - Opening Day (MQA on Tidal: https://tidal.com/track/79927158).

 

Mani.

 

Thanks Mani,

 

I never did get the chance to get the listening done on your samples, but I'll just listen sighted now, and see if I hear the same.

 

Oh, the confirmation bias!

[austinpop]

2 hours ago, manisandher said:

 

I think we can conclude that the truth between hires and MQA does indeed lie somewhere in the middle!

 

 

I have always felt that more important than the encoding (redbook, hi-res, MQA) was the actual care and effort taken on the (re)mastering. Case in point - the Steven Wilson Yes remixes. These sound equally good in CD and hi-res, and both better than the earlier masterings, also released in CD and hi-res.

 

I'd take the Wilson redbook version over the earlier hi-res version.

[manisandher]

15 minutes ago, PeterSt said:

Also, I looked into this this morning because if I had to be correct on the how I came to my judgment on the Hires (sounds the worst) then first I had to be correct on it being fake, right ?

 

The label is ECM, which I felt wouldn't engage in such things. It's a real shame if indeed it is 'fake' hires. But perhaps it was out of their control?

 

Mani.

[PeterSt]

9 minutes ago, austinpop said:

I never did get the chance to get the listening done on your samples, but I'll just listen sighted now, and see if I hear the same.

 

Rajiv,

Let someone rename them for you, while taking notes of what was done. Then it is your turn to listen ...

After your writing out of what you think of it, all the A's B's and C's should be changed back accordingly.

[Le Concombre Masqué]

11 minutes ago, manisandher said:

 

The label is ECM, which I felt wouldn't engage in such things. It's a real shame if indeed it is 'fake' hires. But perhaps it was out of their control?

 

Mani.

we can't rule out it was tape rather than digital or 192 rather than 96 to start with but I bet it's been done very seriously : http://www.josephbranciforte.com/james_farber.html

[austinpop]

13 minutes ago, PeterSt said:

 

Rajiv,

Let someone rename them for you, while taking notes of what was done. Then it is your turn to listen ...

After your writing out of what you think of it, all the A's B's and C's should be changed back accordingly.

 

Excellent suggestion.

 

I should have known it was Anouar Brahem, Mani!

[PeterSt]

45 minutes ago, Le Concombre Masqué said:

I own a Pink original Island LP of Court and always marveled at its sound, especially of cymbals,

 

Exactly. With I Talk to The Wind ahead. And man, this is old.

But the one I have as "good" is unknown for its source. Some Torrent, I'm afraid. I was glad I found it anyway, at first only the lousy Hires being available (and MP3) for what I could find.

All the covers of the various versions look the same.

The other day I found an HDCD (Japanese).  So that exists too and that one has a different cover ("moon face"). I don't have it noted as "Nice Stuff" so I guess the other one is still the best.

 

Debby is really difficult (and even ~10 years older of course). This is the combination to watch for :

  

 

Mind the track times. Mind the number of tracks. Mind the sequence of them. Mind the position of the "Riverside" and also mind the yellow square.

DR is 21, but is maybe not reliable because I have it as "Cue File" only (this is one large file, although the tracks can be differentiated by the Cue Sheet).

 

Yes, it is the Stubblebine Hires transfer I objected to. FWIW.

 

[Le Concombre Masqué]

6 minutes ago, PeterSt said:

 

Exactly. With Talk to The Wind ahead. And man, this is old.

But the one I have as "good" is unknown for its source. Some Torrent, I'm afraid. I was glad I found it anyway, at first only the lousy Hires being available (and MP3) for what I could find.

All the covers of the various versions look the same.

The other day I found an HDCD (Japanese).  So that exists too and that one has a different cover ("moon face"). I don't have it noted as "Nice Stuff" so I guess the other one is still the best.

 

Debby is really difficult. This is the combination to watch for :

  

 

Mind the track times. Mind the number of tracks. Mind the sequence of them. Mind the position of the "Riverside" and also mind the yellow square.

DR is 21, but is maybe not reliable because I have it as "Cue File" only (this is one large file, although the tracks can be differentiated by the Cue Sheet).

 

Yes, it is the Stubblebine Hires transfer I objected to. FWIW.

 

Thank you Peter, looks like the Riverside 20 bits remastered

[esldude]

Self serving interpretation follows.

 

I correctly identified hires.  I correctly chose Redbook as sounding more similar. I correctly commented that mqa was more different than the others. As mqa is lossy with some low level aliasing that makes sense.  Therefore I actually chose 100% correctly.  My mistake was thinking mqa would manage to outdo Redbook. Some of you were fooled by mqa.

 

I'll leave it to you to come up with your own self serving interpretation.

[manisandher]

@Arpiben and anyone else who's interested, here's the original hires file (for a limited time only) to help with any analysis you'd like to do:

 

https://drive.google.com/open?id=14Xau4aJ5ms391itiAYbEQLt6R3SAaDwt

 

Mani.

[manisandher]

5 minutes ago, esldude said:

I correctly identified hires.

 

Uh no, you didn't - you thought A was the hires:

On 10/26/2018 at 7:10 AM, esldude said:

If I were going with conventional ideas, Hi-res, MQA and CD.

 

You said this about the actual hires (that you thought was MQA):

On 10/26/2018 at 7:10 AM, esldude said:

B sounds a bit artificially airy, and soft imaging.

 

You said this about the actual MQA (that you thought was CD):

On 10/26/2018 at 7:10 AM, esldude said:

C is missing some low end, and has unnatural hardness on the upper mids.

 

I agree with your descriptions in both cases though.

 

5 minutes ago, esldude said:

I correctly commented that mqa was more different than the others.

 

Yes, you did.

 

Mani.

[esldude]

36 minutes ago, manisandher said:

 

Uh no, you didn't - you thought A was the hires:

 

You said this about the actual hires (that you thought was MQA):

 

You said this about the actual MQA (that you thought was CD):

 

I agree with your descriptions in both cases though.

 

 

Yes, you did.

 

Mani.

Hey, you weren't supposed to notice that.   Where's a little confusion when you need it?

[esldude]

New self serving interpretation. 

 

I correctly chose A as the best which was redbook. 

 

I said B was artificially airy and soft imaging.  No doubt the result of less precise timing of the high sample rate upon playback, and possibly some IMD from ultrasonics down into the audible band. 

 

I correctly said C was least like the other two.  It was lossy MQA with aliasing to boot. 

 

Again 100% correct.

 

 

[pkane2001]

1 hour ago, manisandher said:

@Arpiben and anyone else who's interested, here's the original hires file (for a limited time only) to help with any analysis you'd like to do:

 

https://drive.google.com/open?id=14Xau4aJ5ms391itiAYbEQLt6R3SAaDwt

 

Mani.

 

Here are some analysis/comparisons using DeltaWave:

 

First, the original Hires vs MQA

 

Average spectrum, looks nearly the same with a very tiny rise at the end in the MQA, barely noticeable:

 

 

Spectrogram of the difference of the two waveforms is a bit strange, with some visible artifacts like that bluish line that seems to increase in frequency with time. That line actually indicates a better null (smaller difference) than the brighter areas. I guess it could be due to some clock drift when MQA encodes/decodes the file, which also shows up in the measurements below:

 

Measurements of Hires vs MQA:

• Volume diff= 0.000dB (1.000x) Phase offset=6627.45ms (636234.934 samples)
• Difference (rms) = -60.77dB
• Correlated Null Depth=71.63dB
• Phase drift before correction: -0.0004 ppm
   

And now, Hires vs 16/44.1:

 

 

And the measurements (Hires vs 16/44.1):

• Volume diff= 0.000dB (1.000x) Phase offset=6627.47ms (636237.000 samples)
• Difference (rms) = -69.56dB
• Correlated Null Depth=90.87dB
• Phase drift before correction: 0.0000 ppm

The below 22KHz difference is much greater in MQA. Those orange/red colors below 22Khz indicate a greater difference. That also explains the larger RMS difference and lower correlated null value in Hires vs MQA compared to Hires vs redbook.

[esldude]

3 hours ago, pkane2001 said:

 

Here are some analysis/comparisons using DeltaWave:

 

First, the original Hires vs MQA

 

Average spectrum, looks nearly the same with a very tiny rise at the end in the MQA, barely noticeable:

 

 

Spectrogram of the difference of the two waveforms is a bit strange, with some visible artifacts like that bluish line that seems to increase in frequency with time. That line actually indicates a better null (smaller difference) than the brighter areas. I guess it could be due to some clock drift when MQA encodes/decodes the file, which also shows up in the measurements below:

 

Measurements of Hires vs MQA:

• Volume diff= 0.000dB (1.000x) Phase offset=6627.45ms (636234.934 samples)
• Difference (rms) = -60.77dB
• Correlated Null Depth=71.63dB
• Phase drift before correction: -0.0004 ppm
   

And now, Hires vs 16/44.1:

 

 

And the measurements (Hires vs 16/44.1):

• Volume diff= 0.000dB (1.000x) Phase offset=6627.47ms (636237.000 samples)
• Difference (rms) = -69.56dB
• Correlated Null Depth=90.87dB
• Phase drift before correction: 0.0000 ppm

The below 22KHz difference is much greater in MQA. Those orange/red colors below 22Khz indicate a greater difference. That also explains the larger RMS difference and lower correlated null value in Hires vs MQA compared to Hires vs redbook.

Which corroborates my hearing A and B as similar and C being much more different. 

 

BTW, when will Deltawave be available for others to use?  Hint, hint.  

[pkane2001]

52 minutes ago, esldude said:

Which corroborates my hearing A and B as similar and C being much more different. 

 

BTW, when will Deltawave be available for others to use?  Hint, hint.  

 

Yes, same in my testing, A and B were hard to distinguish, C sounded different.

 

Deltawave is pretty much done, I just need to clean up all the experimental code I’ve been adding to it, and then it should be ready for a small beta test

 

[Arpiben]

4 hours ago, pkane2001 said:

 

Here are some analysis/comparisons using DeltaWave:

 

First, the original Hires vs MQA

 

Average spectrum, looks nearly the same with a very tiny rise at the end in the MQA, barely noticeable:

 

 

Spectrogram of the difference of the two waveforms is a bit strange, with some visible artifacts like that bluish line that seems to increase in frequency with time. That line actually indicates a better null (smaller difference) than the brighter areas. I guess it could be due to some clock drift when MQA encodes/decodes the file, which also shows up in the measurements below:

 

Measurements of Hires vs MQA:

• Volume diff= 0.000dB (1.000x) Phase offset=6627.45ms (636234.934 samples)
• Difference (rms) = -60.77dB
• Correlated Null Depth=71.63dB
• Phase drift before correction: -0.0004 ppm
   

And now, Hires vs 16/44.1:

 

 

And the measurements (Hires vs 16/44.1):

• Volume diff= 0.000dB (1.000x) Phase offset=6627.47ms (636237.000 samples)
• Difference (rms) = -69.56dB
• Correlated Null Depth=90.87dB
• Phase drift before correction: 0.0000 ppm

The below 22KHz difference is much greater in MQA. Those orange/red colors below 22Khz indicate a greater difference. That also explains the larger RMS difference and lower correlated null value in Hires vs MQA compared to Hires vs redbook.

 

Using Sonic Visualizer & Audacity I do not reproduce the spectogram singularities of your MQA vs Original HiRes.

 

[pkane2001]

2 minutes ago, Arpiben said:

 

Using Sonic Visualizer & Audacity I do not reproduce the spectogram singularities of your MQA vs Original HiRes.

 

 

How was this produced?