Recent Tidal Masters

[coliny]

Hi Peter

When I said "Even 44/16 MQA albums work." in an earlier post I only meant the files played OK I recognized then that they just upsampled and did not unfold.

I have done more tests by playing tracks using my DAC into soundcard and generating FFT spectrums and averaging about 10s music. With no upsampling applied by XXHE. Comparing with decoding on and off. I can quite clearly see if they unfold to get data above 22KHz

Results are attached, my conclusions relating to Tidal albums so far are:-

48/24 MQA all decode correctly

44/24 MQA only 1 found so far that decodes this one from 2L.

44/16 MQA not found any that unfold.

Next I will try measuring some 44/24 files by playing from the Tidal App .

Regards
Colin

[coliny]

I have done a lot more investigation of Tidal Masters & exchanged emails about this with Peter.
The big find was a utility freely available from MQA called MQA TagRestorer. It allows one to check if a music file is MQA and if it is restores the OrignalSampleRate Tag to the FLAC file which MQA deleted from all Tidal Master downloads a few years back. Its the OrignalSampleRate which enables us to get the right MQARATE for the .MTT files in XXHE.

There is a big BUT however. Many of the recent Tidal Masters reveal that the OriginalSampleRate is either 44100 or 48000 so there can't have been any music info to fold/unfold above 22 or 24KHz. So it looks like MQA does not fold these albums and I suppose it only De blurs. Measuring the Tidal App output I only see upsampling to 88200 or 96000 of these offending albums.

Back to XXHE and it always tries to unfold MQA and if you enter 44100 or 48000 as MQARATE it won't work.
So as suggested in an earlier post entering 88200 or 96000 is best way to make them playable. But this does not upsample them, XXHE will show 88200 or 96000 but the Unicode...wav file stays at 44100 or 48000. You can use XXHE's own ArcPredict upsampling but its actual output rate will always be half the value displayed. Custom Filters don't work because of this issue.

I don't want to get into a Sound Quality debate some of the low rate Masters sound ok. I don't understand why MQA have accepted 44100 or 48000 masters for encoding when higher rate files will be available in most cases.

You can get the MQA TagRestorer here:-
https://www.mqa.co.uk/tag435sdf43te

Happy Pondering
Colin

[PeterSt]

Thank you so much for all you hard work, Colin.

All is a bit pitfall-ish, already because I don't think that MQA claims that CDs can "unfold" (but you refer(red) to links some times from BobS which need a subscription which I have not - so I can't really read what is in there). What I know is :

In 2016 or so I was told (by telephone) that at some stage normal CDs would be MQA'd too. In by book this won't and can't mean that they will be able to unfold, as there is nothing to unfold. Now, it could be hoped that it does (unveils a higher sampling rate), but I don't have these hopes at all because it just can't be. And as you saw, that is not so indeed.
The real pitfall of course is the "law" of upsampling to 88.2. Thus, Roon must do that (governed by MQA Ltd), Tidal player must do that, hardware must do that, all must do that. And XXHighEnd does not.

The fact that XXHighEnd can not deal with it, is because at the time it was not recognized that it should be decoded (it did not exist). Most probably if I would hand it properly to the decoder, it indeed "deblurs" as you tell it, but it may also depict to upsample, and the combination could work BUT is worse than the original CD upsampled by known XXHighEnd means.

While the deblurring could be a pile of BS (in my view), the 24 bit files at least comprise highres inside, and they *can* unfold to normal 88.2 / 96. It is 99 times better than all the fake highres around, as I don't recall any fake MQA highres. So for that matter alone it may sound better, and about the deblurring I don't know. Mind you, the way MQA (via hardware) depicts the upsampling, is very much alike Arc Prediction with no ringing and such. So that part is ... well ... uhm ... the good part according to MQA (they may call that deblurring because actually there is less blurr by no-ringing (much much more to say about this), but if that were all there is to it, then MQA is a pile of air, except they offer highres while others do not. And mind you, I know for a fact that MQA has access to originals, e.g. the big producers had/have not.

One step more to unfold : If MQA would not present the LOWER res files, there would nothing left to be upsampled per the means MQA wants (which takes care of the not-ringing filtering).
So you see, MQA could be the biggest hoax ever, because it does not even deblur. It does one thing only (as how I see it) :

1. Bring a highres file in low reaolution (downsample it !)
1. Offer a couple of low-ringing filters which blur less according to what we are used to (DAC filters and such).

Yes, that is 2x 1.

The coincidence with XXHighEnd we have is
a. The adjacent DAC is NOS(1) and does not do a thing for filtering;
b. The filtering applied by software is not-ringing even one sample.

So that is 2x something (of also nothing).
Thus the bad luck for MQA is that I already do exactly the same, with one difference : I can't magically create highres from CD's (can only upsample them), while MQA has access to the originals often, which are highres and don't exist anywhere else (for those occasions that they don't). Again, this story is longer, but I know.
We would be better off if we had those originals with all there 24 bits in e.g. 96 or 192 sampling rate, but they just don't exist outside of what HDTracks et all may uncover.

We are also better off with the original CDs and this is exactly what happened as long as e.g. Tidal let them be and not molest by MQA encoding. Maybe, very maybe, that is just a hoax, and for that I should alter the software instead of writing posts like this. But still, a post like this takes 20 minutes, while finding where to be in the code (because this is about tweaking) is easily a day, and I don't have that day.
But I will. I mean, now I know that the 16bit CDs could be unmolestedly in there (apart from here and there one bit used to turn on a green or blue light), I should listen to those and especially not decode them and normally upsample them.

Best regards,
Peter

[coliny]

Hi Peter

I do have some time these days so if there is anything I might be able to do to help just ask.

I looked at one MQA track from Hiromi which is 24/48 and Tagrestorer says ORIGINALSAMPLERATE was 48.
I compared the Unicode...WAV files with decoding on & off but they look identical to me, see attached.

https://tidal.com/browse/album/276031302

Colin

[coliny]

Further to reply#16 I have worked through my MQA downloads with the TagRestorer and find many have an originalsamplerate of 44100 or 48000 so nothing to unfold even though they are recognised as MQA.

Also the 44100 ones are 16bit and decode to 24bit.

I find the 24bit ones  (before decoding) which don't double their frequency when decoded play ok with MQARATE set to 44100 or 48000 as appropriate so have attached macros for this.

The 16bit ones jump to next track prematurely with MQARATE of 44100 0r 48000 so do require 88200 or 96000 to play properly.

To differentiate various possibilities the TagRestorer can be used but needs something like dBpoweramp installed to show the metadata:- see attached.

Alternative is install mp3tag free version. This can be adjusted to show Frequency & Bit Depth. Attached example is for a downloaded album showing all tracks are 48000/24. also it shows how to edit a column to show bit depth.

Next start Mp3tag and change directory to your XXHE folder and in XXHE (can be Full OS Attanded) play a track with decoding turned ON. When play starts click refresh in Mp3tag and the unicode.... file will show the decoded frequency & bit depth. See attached.

Now you have necessary info to set MQARATE in the MTT files.
I determine there a 3 options:-

1. Unicode...file frequency is twice the undecoded frequency then set MQARATE to the Unicode...file frequency.

2. Unicode...file frequency is same as the undecoded frequency and undecoded bit depth is 16 bit then set MQARATE to the twice Unicode...file frequency.

3. Unicode...file frequency is same as the undecoded frequency and undecoded bit depth is 24 bit then set MQARATE to the Unicode...file frequency.

Hope this is useful

Colin




 

[pedal]

Streaming has become a necessity.
But why not switch to Qobuz? It has less titles than Tidal, but is hassle-free. No decoding.
I have been a Qobuz-subscriber for 2 years and number of titles are increasing all the time.

[PeterSt]

I have been more than busy this week, but ran coincidentally on the albums you see below - the rightmost three.

They don't unfold at all, although the pre-process marked them as MQA anyway (MQA44). They play as 16/44.1, but are of course upsampled like everything. I can see that they are from June 2021 and I will most probably be able to find more what I prepared from that time period.

Point is : these sound just fine to my ears, though special for their era (something like end 60s I'd say). *That* they still sound special is a good sign. It somehow (wants to) tells me that this was not tampered with. And so indeed, there will be some bit (now and then) used for the MQA detection, but that does not disturb (my ears).
Keep in mind : these are not decoded. XXHighEnd just leaves them alone, and this is in the XXHighEnd version you use too (the latest). Where these still can be found - I don't know (I did not look either) but it could be worth trying. But more importantly : I should have a reference by these. IOW, what is the software NOT doing to them and why. Probably today I can not test that (because all what seems fake I skip), BUT I know the dates of these plus I will be able to trace back by date what I all changed since (ahem) April 2021 (latest public version).

So right; I now can search / sort by date for MQA 16/44.1 from around that time (June 2021) and listen to more of them ...

Colin, again thank you very much for the spent effort. I will try to make that fruitful.
Peter

[PeterSt]

A limited result ...

When I go 6 months beyond that end date, I have some more (2nd screenshot). Btw, mind the blue icons only (44). But I already know by heart that most I see are 24/44.1. They also don't unfold but they are decoded. This is fine with me and not of my interest today. The interest is : find the software means I must have (had !) in there to leave the 16/44.1's alone for decoding. The the Deep Purple ones.

Peter

[coliny]

Hi Peter

Downloaded Deep Purple/Hard Road/CD01 - It is 44/16 Tagrestorer says ORIGINALSAMPLERATE is 44100. It decodes to 44/24 just like all the other 44/16 MQA albums I have.

In the period 1.4.21 to 1.9.21 I have 4 MQA albums 3 of which are 44/16 :-
K.D.Lang - Sing It Load
Pink Floyd - Animals Remastered
Joan Armatrading - Consequences

The other album is Gretchen Parlato - Flor but this is 44/24 with ORIGINALSAMPLERATE 88200 so full MQA

I also see that that the Unicode.... file decodes to same rate and depth whatever MQARATE you put in the MTT files or even if its blank. So I am guessing that the MQARATE just influences whether tracks play to the end or not.

Regards
Colin

[PeterSt]

So I just obtained this one and a couple more. I see that I denote these "formally" MQA16 (see in the title bar below). This does not unfold or act strange or anything; it is just not dealt with by the MQA decoder. Also see the 1411 kbits/sec in the top-middle.

Does it sound optimally ? I don't know - I didn't know it. And because I didn't have it plus I like to have it "all" (from Gruppo Sportivo) I am a happy camper. It better had been 100% untampered with, but alas.

Please notice that there are also "opportunities" so to speak; I mean, this album was not there at Tidal a while ago. So maybe because of MQA it now is ?

Regards,
Peter

[PeterSt]

Addendum :

Notice that I use 2.11 just like you for the Audio PC. But it is the updated version on the Music Server PC that will cause this.

Maybe the below is telling ?

Stupid thing is that I now don't know what I am "skipping" (don't show) as such. So XXHighEnd just finds these and deals with them exactly how I would want. Or maybe a year ago I thought to go skip them because worthless ? ... But I did not do that ? and now insights change and I am even happy with that because I/we can choose (to skip the MQA16 denotes ones) ?

Peter

PS: The Quality : HI_RES is not correct of course. But this is officially in the data provided by Tidal - just saying ...

[coliny]

Hi Peter

What bit depth is the unicode.... file ?. If it stays at 16 then it is just like switching decoding off in Settings.

With my settings for MQA16 (decoding ON, MQARATE 88200) I get 24bit and this gives a measurable lower noise even though XXHE upsampling ultimately outputs 24bit with or without decoding. Both options giving 44100.

I slightly favour my settings for sound quality but I might be imagining it. I will get someone with younger ears to see if they hear a difference.

Anyway good news if you are able get the MQARATE automatically now.

Best Regards
Colin

[PeterSt]

Hi Colin,

Maybe it wasn't entirely clear : those MQAs are 16 bits and 44.1. There can't be anything in there to do with. And thus it is good that nothing is done with it because it would be illegal if something *is* done with it (like Roon and Tidal Player do).

You can always look in the middle-top of the CD cover (the one shown during playback). If that shows 1411 it is 16 bits and 44.1 KHz.
If it is 24 bits withoutunfold (your last textual example I think) it will show 1411/16*24 = 2116 (2117).
Is it 24 bits unfolded, then 2116 * 2 = 4232. More is not possible because more never unfolds (the made up story/lie). This doesn't prevent that it can be upsampled from there. This shows in the left pane in the XXHighEnd gui. There can also show the native 24/192 if it would be that. Etc. ...

Maybe superfluously : the 24 bits MQA's which don't unfold exist from almost the beginning. Electrified from Boris Blank is such an example (the 24/44.1 version - a 24/96 version exists too). Thus that is an example which will show 2216 in the top of the Coverart. Let's say that these are "deblurred" just the same, but not HiRes.

The ones I show in my last post(s) are 16/44.1 and whether deblurred or not, I thus don't do anything with them because there is nothing to do**. In my version that is ! And they are denoted MQA16. Ad oh, since I now know that (recall it), I now see that I can look for them by searching for MQA16 - see below above the mouse pointer.

**): Deblurring and such does not happen on your side; it happens at the encoding server-side somewhere (ehh, if at all). The decoding is only for the unfolding. Unfolding means : expanding to Hires - which requires 24 bits first.

Peter

[coliny]

Hi Peter

Something happens when decoding MQA16, attached is 1st track of the Gruppo Sportiv album and it shows decoding to 24bit and lower noise above about 20KHz

Good album
Colin