MQA vs HiRez: an apples-to-apples comparison

[manisandher]

There's been a lot of discussion about MQA here, but not a lot of apples-to-apples comparisons. I'm hoping to change that with this thread.

 

Below you'll find links to two files:

 

- one is the original 24/88.2 (FLAC file size = 67.9 MB)

- the other is MQA decoded to 24/88.2 (FLAC file size before decoding = 38.6 MB)

 

In both cases, the original and the MQA were played back through Roon, and the output captured using Audacity. So everything remained in the digital domain. You can listen to them and/or analyse them in which ever way you like. If all the MQA talk here is true, it should be trivial to figure out which is which.

 

So... which is the original, and which is the MQA?

 

Edit please use the following files for the comparison:

 

https://drive.google.com/open?id=1eZMPe2cbPgEqKDztx8D2VZgcPOsnEuSy

 

https://drive.google.com/open?id=1_nIJnB5_WspDOMa_-So4LTB-d11b3Qdv

 

Mani.

[Fokus]

Good initiative. But do you know that the MQA file was derived from the same original?

[manisandher]

1 minute ago, Fokus said:

Good initiative. But do you know that the MQA file was derived from the same original?

 

Yep, from the same master. (Any analysis will show that this is indeed the case.)

 

Mani.

[PeterSt]

27 minutes ago, manisandher said:

I'm hoping to change that with this thread.

 

Mani, Let me know if you need more from the same master. I might have run into a few the last time. I'd need to check it for real though.

 

Count me out as a contender (if that would be desirable in the first place) because my ears are stuffed the past few days. Blame the weather^2.

 

Peter

[manisandher]

5 minutes ago, PeterSt said:

Let me know if you need more from the same master. I might have run into a few the last time. I'd need to check it for real though.

 

Hey Peter, let's see how we go with this first, and maybe I'll take you up on it later.

 

5 minutes ago, PeterSt said:

Count me out as a contender (if that would be desirable in the first place) because my ears are stuffed the past few days. Blame the weather^2.

 

Hope you get better soon. But stuffed ears don't preclude doing analysis ?.

 

Mani.

[Miska]

I've been buying bunch of albums in both original hires and MQA format from highresaudio.com for comparison purposes. The MQA version costs 2€ more, so I'm kind of having hard time understanding who would buy MQA versions for anything else except comparison.

 

Anyway, if someone wants to compare, that is a good way.

 

[manisandher]

6 minutes ago, Miska said:

I've been buying bunch of albums in both original hires and MQA format from highresaudio.com for comparison purposes. The MQA version costs 2€ more, so I'm kind of having hard time understanding who would buy MQA versions for anything else except comparison.

 

Anyway, if someone wants to compare, that is a good way.

 

 

The original 24/88.2 file I've used here was from HDTracks. The MQA from Tidal.

 

Mani.

[PeterSt]

2 minutes ago, Miska said:

I've been buying bunch of albums in both original hires and MQA format from highresaudio.com for comparison purposes. The MQA version costs 2€ more, so I'm kind of having hard time understanding who would buy MQA versions for anything else except comparison.

 

Haha, maybe your workflow ain't right. I have a subscription to Tidal and from there I may compare Redbook with MQA. Virtually for free.

I don't think I would be buying MQA when I'd also be in the possession of original Hires ... But, easy to see : there's much more dormant Hires available than what HDTracks offers (which mostly is fake at the same time).

 

PS: Lately I see people write about fake MQA Hires. But the first of that  yet have to run into ... So these people have bad dreams or I am slow.

 

[PeterSt]

10 minutes ago, manisandher said:

But stuffed ears don't preclude doing analysis ?.

 

Mani - I hope you're not referring to me looking into the files ?

(I could do that, but it doesn't look like the good idea to me)

 

Peter

[Miska]

5 minutes ago, PeterSt said:

Haha, maybe your workflow ain't right. I have a subscription to Tidal and from there I may compare Redbook with MQA. Virtually for free.

I don't think I would be buying MQA when I'd also be in the possession of original Hires ... But, easy to see : there's much more dormant Hires available than what HDTracks offers (which mostly is fake at the same time).

 

PS: Lately I see people write about fake MQA Hires. But the first of that  yet have to run into ... So these people have bad dreams or I am slow.

 

 

I don't want to promote Tidal's use of MQA for pseudo hi-res. Rather Qobuz and others who stream real thing. I just don't see point in MQA in first place. It is just harmful thing in so many ways.

 

I've found some fake hires MQA albums. But overall I don't have interest or time to care much about MQA content as long is it is not coming on my face.

 

I don't frequently buy from HDtracks... highresaudio.com has usually better quality control.

[Miska]

By quick glance A looks like MQA and B like original. At least the two things are clearly different. Unless MQA has recently added noise gate to their encoding algorithm. Which would explain why B is missing some features A has. Otherwise MQA is adding some new stuff...

 

[manisandher]

23 minutes ago, PeterSt said:

Mani - I hope you're not referring to me looking into the files ?

(I could do that, but it doesn't look like the good idea to me)

 

Why not? I'd be interested in seeing what you find.

 

Mani.

[manisandher]

15 minutes ago, Miska said:

By quick glance A looks like MQA and B like original. At least the two things are clearly different. Unless MQA has recently added noise gate to their encoding algorithm. Which would explain why B is missing some features A has. Otherwise MQA is adding some new stuff...

 

 

Thanks, that's exactly what I was hoping for.

 

Could you share your analysis? (I'm not saying you're correct though.)

 

Mani.

[manisandher]

58 minutes ago, Miska said:

By quick glance A looks like MQA and B like original. At least the two things are clearly different. Unless MQA has recently added noise gate to their encoding algorithm. Which would explain why B is missing some features A has. Otherwise MQA is adding some new stuff...

 

 

@Miska, I found a mistake in files A and B. Could you redo your "quick glance" with files C and D please? (Again, one is the original 24/88.2, the other decoded MQA.)

 

Thanks.

 

Mani.

[Miska]

40 minutes ago, manisandher said:

 

Thanks, that's exactly what I was hoping for.

 

Could you share your analysis? (I'm not saying you're correct though.)

 

Mani.

 

There's more noise-shaped noise in lead-in and loud-out on sample A. Also the frequency response at high frequencies is more rolled off on A. In lead-out there are some ticks in A though that don't exist in B. And B has some ultrasonic sounds in lead-out that A doesn't have.

 

Since MQA cuts word-length of the source to around 15-bit effectively format becoming like noise-shaped 88.2/15 here. And since their filter is slow roll-off it causes some HF roll-off.

 

However, applying a digital noise gate to a source would cut some of the noise, so you can get dropped noise levels that way too.

 

I didn't listen to the samples yet, just downloaded the files to my laptop and took a quick look while eating breakfast...

 

[PeterSt]

33 minutes ago, manisandher said:

Why not? I'd be interested in seeing what you find.

 

 

1 hour ago, manisandher said:

You can listen to them and/or analyse them in which ever way you like.

 

Oops. I missed that.

 

I've never done such a thing (MQA vs normal analysis) so maybe there should be a first time. What I will do is attempt this analysis, but without ever listening. I mean, stuffed ears aside it would be to easy (for me, and I am sure the same would count for you, Mani).

It would be a thing to really sit down for and with sufficient effort I think it will lead to something. Recall that Sunday afternoon with Take Five ? something like that.

 

I have one very strange request : Assumed that there's no identifying data in those downloads (I did not try it yet), could you link me to a Tidal Redbook link so I can listen to the music itself and be prepared ? So I "must" listen to a neutral product which in this case is neither the Hires nor the MQA. I might even like to compare in-file with the Redbook, unless you tell us that this is not a good idea, just not allowed or unfair etc. But anyway, I must be able to recognize spectral data and for that I first need to know how the data sounds. Cymbals ? Piano ? Singing ? ... and I can't listen to the Hires or MQA because already one of them would unveil which is which. I am sure you agree (, believe) and understand.

 

So my personal fun task is to find out by means of looking into the files only. Not listen ...

[PeterSt]

1 minute ago, Miska said:

while eating breakfast...

 

Shouldn't that be lunch ? ?

[PeterSt]

2 minutes ago, Miska said:

Also the frequency response at high frequencies is more rolled off on A.

 

OK, I wasn't even going to look at that particular level, or "domain" if you want. I am just going to look at the samples ... (sequences of).

[manisandher]

2 minutes ago, Miska said:

 

There's more noise-shaped noise in lead-in and loud-out on sample A. Also the frequency response at high frequencies is more rolled off on A. In lead-out there are some ticks in A though that don't exist in B. And B has some ultrasonic sounds in lead-out that A doesn't have.

 

Since MQA cuts word-length of the source to around 15-bit effectively format becoming like noise-shaped 88.2/15 here. And since their filter is slow roll-off it causes some HF roll-off.

 

However, applying a digital noise gate to a source would cut some of the noise, so you can get dropped noise levels that way too.

 

I didn't listen to the samples yet, just downloaded the files to my laptop and took a quick look while eating breakfast...

 

 

Thanks. As I said, I found a mistake in A and B. C and D are the correct ones. Thanks.

 

Mani.

[manisandher]

19 minutes ago, PeterSt said:

It would be a thing to really sit down for and with sufficient effort I think it will lead to something. Recall that Sunday afternoon with Take Five ? something like that.

 

Your memory is obviously much better than mine... and you out-drink me by 10-to-1!

 

19 minutes ago, PeterSt said:

I have one very strange request : Assumed that there's no identifying data in those downloads (I did not try it yet), could you link me to a Tidal Redbook link so I can listen to the music itself and be prepared ?

 

Here you go: https://tidal.com/track/26265350

 

And remember, it's tracks C and D that are to be compared (I've taken tracks A and B down).

 

Mani.

[Miska]

38 minutes ago, manisandher said:

Thanks. As I said, I found a mistake in A and B. C and D are the correct ones. Thanks.

 

Same comments, just A=C and B=D.

 

[manisandher]

11 minutes ago, Miska said:

 

Same comments, just A=C and B=D.

 

 

Thanks. When you have the time, it'd be great to actually see your analysis.

 

Mani.

[mansr]

Neither file has the MQA rendering data in the LSB. It is not a pure capture.

[manisandher]

5 minutes ago, mansr said:

Neither file has the MQA rendering data in the LSB. It is not a pure capture.

 

Huh???

 

 

Mani.

[PeterSt]

1 hour ago, manisandher said:

1 hour ago, PeterSt said:

It would be a thing to really sit down for and with sufficient effort I think it will lead to something. Recall that Sunday afternoon with Take Five ? something like that.

 

Your memory is obviously much better than mine...

 

It seems it was on a Thursday. The thread ended on Sunday though. Maybe because of being startled I remember that particular day.

 

 

Anyway, the text of one of the posts went like this :

 

A little zoomed out. What we see here, I only saw after a long time as a pattern ... what do you see ? Look at each of the peaks and compare the both takes ... Do you see it ? Somehow there is a consistent "opposite" balance in how the peaks are formed. Just find a peak where one sample bumps out. It doesn't matter where you find it. Now look at its counterpart. See ? the one with one sample jumping out has 3 samples to *let* that sample jump out, where it counterpart has an even number of samples for the width concerned, and it is impossible for one sample to jump out. If you try to reason where the "base" of this emerges, at least it is hard to see, but it must be at the base of the wave somewhere, where the one is wider as the other. This is a most consistent thing throughout the both files ! Will this change sound ? ha, you bet it will.

 

16/44.1 vs. 24/176.4

 

So something like that I have in mind.