Everything you wanted to know about MQA

[mgbpuff]

www.theabsolutesound.com/articles/the-politics-of-mqa/

[mgbpuff]

Sept 22, 2017 12:01:10 GMT -5 DYohn said:

Sept 21, 2017 21:07:57 GMT -5 mgbpuff said:

People seem to think MQA is a lossy compression scheme, but from what I've read it seems to me to be an expansion scheme instead which maintains the original file, adds corrective new information to that original file in a noise hidden codec and then upon play the original file is expanded to a higher sample rate that now contains all the new information. It could have been decoded into a higher sample rate file before transmission, but why do that since it is easier to transmit the lower undecoded sample rate file and the final result using MQA decoding DACs is the same.

This is incorrect. MQA is a lossy compression scheme. en.wikipedia.org/wiki/Master_Quality_Authenticated

Oh, come on! You're referencing Wikipedia? This is nothing but copies of anonymous postings and reeks of the same disdain as exhibited in most of the postings in this thread. It may be lossy for the undecoded playback, but the coded part may include the original borrowed LSB information plus even more so that when it is unfolded into a higher sampling rate, all of the old info is there plus some new corrective information.

[DYohn]

"Unfolded at a higher sampling rate" means what? Zeroes are added to fill in the missing information? Pointless, and adds noise. And yes I used Wikipedia but only because it was the first thing that came to mind. Here's a different source:

www.stereophile.com/content/meridians-mqa-one-listeners-impression

I stand by my opinion of MQA as posted earlier.

[LCSeminole]

Sept 22, 2017 13:21:09 GMT -5 DYohn said:

"Unfolded at a higher sampling rate" means what? Zeroes are added to fill in the missing information? Pointless, and adds noise. And yes I used Wikipedia but only because it was the first thing that came to mind. Here's a different source:

www.stereophile.com/content/meridians-mqa-one-listeners-impression

I stand by my opinion of MQA as posted earlier.

While I've remained quiet and not said my .02 cents worth on this subject, I couldn't agree more with your assessment of MQA.

[KeithL]

Here's a VERY detailed explanation of the factors involved (although it's a bit dated in terms of the operational details).
www.soundonsound.com/techniques/mqa-time-domain-accuracy-digital-audio-quality

This, and the other specific explanations I've seen, all seem to agree that the "fully unfolded" MQA file should end up at the original resolution.
I'm not 100% sure that this "original sample rate" refers to the the sample rate of the original original master file.
(There is precedent for this question. The apodizing filter in Dolby's Professional Encoder takes in a 48k file and outputs a 96k file as part of the "improvement" process.)
However, I've seen nothing to suggest that the MQA software itself will ever "upsample" to any sample rate beyond the one specified as being "the original one".

As far as I can tell, Cary Audio has for some time offered various types of upsampling options on their digital equipment.
They seem to simply offer the user the option of upsampling the digital audio to a variety of sample rates (pick the one you like for whatever you're playing).
(Many manufacturers do this.)

Just to expand on something I talked about in another post....

Let's assume you have a 44k file and upsample it to 96k....
- if your 44k file was originally produced from a 96k master, the upsampling process does NOT specifically replace any information that was removed during the original downsampling process
- if your file started life at 44k, the upsampling process does NOT magically figure out what information would have been there if it was recorded at 96k and add or replace that information
- the upsampling process CAN, if configured to produce the most accurate conversion possible, create a 96k file that sounds virtually the same as the original 44k file (but is a lot bigger)
- the upsampling process could actually enable the DAC to do a better job of converting your 44k file (most DACs do this internally already, so doing it again is just redundant)
- (even when attempting to produce an accurate upsample, different choices, like different sample rates, will probably produce tiny differences anyway; they are NOT improvements; just differences)
- if configured to do so, the upsampling process CAN be used to deliberately produce a variety of alterations in the sound based on the filters and options chosen by the operator
- the alterations CAN be based on some sort of educated guess of what SHOULD have been there - based on an analysis of the existing content
(somewhat like a restoration artist may replace chipped sections in a painting - by painting what he or she thinks belonged in those areas that are now missing)
- the alterations can simply be based on deliberately producing variations (to give the user choices to pick from)

Please note that none of this in any way suggests that, if they've offered you several different options that sound slightly different, you may or may not prefer one of their variations to the unaltered original.
(In other words, they're simply offering you several slightly different sounding choices, so you can pick the one you like; of course the marketing department wants you to think that those choices are also "better".)

Sept 22, 2017 13:21:09 GMT -5 DYohn said:

"Unfolded at a higher sampling rate" means what? Zeroes are added to fill in the missing information? Pointless, and adds noise. And yes I used Wikipedia but only because it was the first thing that came to mind. Here's a different source:

www.stereophile.com/content/meridians-mqa-one-listeners-impression

I stand by my opinion of MQA as posted earlier.

[Boomzilla]

Well - Success? Maybe...

After triple-checking Audirvana 3 Plus configurations, after rebooting the server computer, and after hopping up and down on my left leg while rubbing my stomach counter-clockwise, I may (not positive, but MAY) have gotten MQA to work. A FIRST!!

So here's what I got - when having Audirvana play downloaded MQA tracks from 2L's website, the LED on the Dragonfly turns violet (is it "purple," or is it "violet?" Search me...). If it's really "purple," in "Audioquest-speak," then it's (finally) decoding and rendering MQA. If not, then I'm just hearing PCM audio at some higher-than-red-book sample rate. I'm suspecting that it IS actually MQA because there is an MQA header that's lit on the banner of the Audirvana player.

And this heralded MQA sounds like? How the heck should I know? This is demo music from 2L and although it sounds just fine, so does lots of my "normal" 44.1 stuff through jRiver.

And, further, try as I might, I could NEVER get my Audirvana / Dragonfly combo to play my "MQA CD" in anything but 44.1, non-MQA format. Now this was confusing, because I directed Audirvana to access the MQA-CD directly (and it did). If Audirvana can decode, and the Dragonfly can render, then the gang's all here and the combo should just twist the panties clean off the CD. But no such exposures resulted - just 44.1 PCM. WTF?

So MAYBE I've actually heard my first MQA (and maybe not). If what I heard was MQA, then I can say that there's no "night-and-day" difference or improvement. But OTOH, I still understand that It's possible that I'm close, but no cigar.

I've got 14 more days on my free Audirvana trial, so I'm open to suggestions...

Thanks for being patient with me - Boomzilla

[garbulky]

Sept 22, 2017 14:27:47 GMT -5 Boomzilla said:

Well - Success? Maybe...

After triple-checking Audirvana 3 Plus configurations, after rebooting the server computer, and after hopping up and down on my left leg while rubbing my stomach counter-clockwise, I may (not positive, but MAY) have gotten MQA to work. A FIRST!!

So here's what I got - when having Audirvana play downloaded MQA tracks from 2L's website, the LED on the Dragonfly turns violet (is it "purple," or is it "violet?" Search me...). If it's really "purple," in "Audioquest-speak," then it's (finally) decoding and rendering MQA. If not, then I'm just hearing PCM audio at some higher-than-red-book sample rate. I'm suspecting that it IS actually MQA because there is an MQA header that's lit on the banner of the Audirvana player.

And this heralded MQA sounds like? How the heck should I know? This is demo music from 2L and although it sounds just fine, so does lots of my "normal" 44.1 stuff through jRiver.

And, further, try as I might, I could NEVER get my Audirvana / Dragonfly combo to play my "MQA CD" in anything but 44.1, non-MQA format. Now this was confusing, because I directed Audirvana to access the MQA-CD directly (and it did). If Audirvana can decode, and the Dragonfly can render, then the gang's all here and the combo should just twist the panties clean off the CD. But no such exposures resulted - just 44.1 PCM. WTF?

So MAYBE I've actually heard my first MQA (and maybe not). If what I heard was MQA, then I can say that there's no "night-and-day" difference or improvement. But OTOH, I still understand that It's possible that I'm close, but no cigar.

I've got 14 more days on my free Audirvana trial, so I'm open to suggestions...

Thanks for being patient with me - Boomzilla

I think your MQA files are indeed playing MQA as indicated by the dragonfly. However a bit of a trick.
Red: Standby Green: 44100.0 Hz Blue: 48000.0 Hz
Amber: 88200.0 Hz Magenta: 96000.0 Hz
Unfortunately Magenta isn't a whole lot further away from purple.
So Magenta means the first unfold - of the MQA done by Tidal. 96 Khz.
The PURPLE which I think is MQA is when the dragonfly is using its renderer to do the second unfold.
So the question is ....is it magenta or purple.
These are the normal non MQA colors. (note the one on the bottom right -magenta- is so similar but it's NOT MQA).



This is the MQA color. Confusing enough? It's called "the deeper shade of magenta". How's that for user friendliness!


As an aside to how confusing this is....here is a demo from audioquest ALSO thinking they are playin MQA when they are not. (The light is blue - which is not MQA).


I guess MQA CD's are a new thing though it should be playing MQA, it looks like it isn't.

Now here's something you can do - rip your MQA CD - as a wav file with no effects applied like replaygain or anything - and then try to play back that file. If I'm correct, this should allow audirvana to detect the MQA stream and play it back for ya.

[KeithL]

THIS POST IS ABOUT GENERAL THEORY AND *NOT* ABOUT MQA
STOP NOW OR RISK BECOMING EVEN MORE CONFUSED FOR NOTHING

There is a sort of precedent for making a claim like that.... although I do not believe it applies to MQA.

There are compression methods that rely on creating a "model" rather than directly storing or compressing the data.
One of these is what is known as "fractal compression" used in certain types of image processing.
What happens is that, when you apply fractal compression to a picture of a furry bunny rabbit, the algorithm essentially creates a model for how to draw furry bunny rabbits.
Rather than containing data, the model contains instructions on "how to draw that particular bunny rabbit".
(Those instructions MAY include samples of images of fur and textures - among other stuff.)
Then, when you "decode" the image, rather than simply retrieving data, the decoder creates a new image based on the model.

The nature of this process leads to two very interesting things:
1) you will NEVER get back the original picture; but you can get arbitrarily close to it (which is effectively the same thing)
2) you actually CAN create a new version of the picture that contains MORE DETAILS than the original (although they will be "arbitrary details" - based on the model rather than the original data)

Think of it like looking at a slightly blurry picture, then describing it carefully to a sketch artist.
It's quite possible that the sketch artist could produce a drawing that's actually sharper and more detailed than the original.
(Note that "sharper" and "more detailed" are somewhat different than "more accurate".)

However, to answer DYohn's question:

(The following is a really non-technical explanation .... but it gets the basic idea across.)
When this sort of thing is done with digital audio they normally start either with zeroes, or with points derived by some sort of interpolation.
This is normally followed with some sort of filtering chosen to "squeeze the samples closer to the correct values".
For example, if I'm upsampling 44k to 96k, I could start by simply adding new samples numerically half-way between each pair of original samples.
(If you picture this in your head, the result will be a waveform with a few sharp edges and odd corners - because I've obviously put a few straight lines where there should be curves.)
However, now I apply a sharp filter at 50 kHz (just above the Nyquist frequency for a 96k sample rate).
Since most of the "errors" in my new waveform occur at higher frequencies, when I filter out the higher frequencies, I remove the errors.
And what I end up with is a new 96k file that is as close as it can get to the original 44k file - only it's 96k.
(Note that we haven't created more information.... we've instead created a 96k file that, in terms of the actual waveform contained, is as close as possible to the original 44k file.)

Sept 22, 2017 13:21:09 GMT -5 DYohn said:

"Unfolded at a higher sampling rate" means what? Zeroes are added to fill in the missing information? Pointless, and adds noise. And yes I used Wikipedia but only because it was the first thing that came to mind. Here's a different source:

www.stereophile.com/content/meridians-mqa-one-listeners-impression

I stand by my opinion of MQA as posted earlier.

[KeithL]

I just wanted to put a few more facts (and non-facts) out there that might be helpful.....

1)
As far as I know, ANY player software that can deliver a bit-perfect audio stream SHOULD be able to play an MQA file, and pass it on to an MQA DAC to decode.
It shouldn't require anything special - although it will only work if there is NO processing (NO DSP; NO volume control; NO ReplayVolume; etc.)
(They may not be able to SAY that "they support MQA" without paying a licensing fee.)
Note that this would allow your MQA DECODER DAC to decode the MQA file; it won't do anything for your MQA renderer DAC.

2)
Some of the limitations are strictly software.
For example, with Tidal, ONLY the COMPUTER CLIENTS (Apple and Windows) support MQA.
So streamers with Tidal, and Tidal on your phone do not - but simply because the client software doesn't.

3)
Some software may require what seem to be weird settings for it to work.
For example, in Tidal, on a Windows 10 computer, if you want Tidal to do the first unfold and send 96k to your DAC for MQA masters.....
- you must go into the Tidal Setup screen.....
- then go into Stream
- then configure your particular DAC (output device) with "Exclusive Access" (check the box)
With other software, you may need to disable MQA to get the player to pass the un-decoded stream on to your DAC.

4)
The whole concept of "an MQA CD" is interesting.......
In an MQA file, the "extra stuff" is "folded" into the area of the spectrum that occupies the bottom few bits of a 24 bit file.
In other words, all MQA files I've EVER seen described are 24 bit files.
However, CDs only contain 16 bit audio, so they cannot "stash the extra information below the 16 bit floor".
To me, this suggests that MQA CDs may be similar, but use a few bits above the bottom 16 bits to store data, which would suggest that they would deliver lower quality than 24 bit MQA files....
(The only alternative I can see is that they're doing something else entirely - which could be why only some players will play them - perhaps they DO require some sort of special support.)

I haven't looked, but I haven't seen ANY details about how MQA CDs are encoded.
(Again, everything I've seen talks about how "it all folds together" with a 24 bit file..... and not a 16 bit CD.)

[brutiarti]

KeithL the filters on some dacs can be equivalent to some sort of equalization presets?

[Boomzilla]

Sept 22, 2017 14:38:29 GMT -5 garbulky said:

...Now here's something you can do - rip your MQA CD - as a wav file with no effects applied like replaygain or anything - and then try to play back that file. If I'm correct, this should allow audirvana to detect the MQA stream and play it back for ya.

I first tried just copying the CD files to a folder using Apple's "Finder" app. The only files that the machine could see to copy were AIF format, but once copied, they did show a "MQA" flag in the destination folder. But Audirvana won't play them as MQA.

I then tried just directly importing the CD into the Audirvana Library folder, using only Audirvana. Same thing - I get AIF format with a "MQA" flag, but no MQA output.

I could use jRiver to rip the CD to WAV format, but it would seem that this would only be another conversion (and less likely to preserve the MQA data). But I will try it.

I actually DO think that the 2L files I've downloaded ARE playing in MQA. The Dragonfly's LED does seem to be "bright violet" aka "purple" (in AudioQuest speak). I could try playing the same files via the Emotiva XDA-1 DAC to compare to the Dragonfly's output in MQA, but it's still an apples-to-oranges comparison. The DACs themselves may cause more difference than the MQA vs. non-MQA.

So even having heard it, I'm ambivalent about whether it's "better" or not. My best 44.1 PCM CDs sound as good as the MQA files, I think, but until I hear some music that I'm really already VERY familiar with, I can't say for sure.

But my initial impressions are that MQA is NOT any "night-and-day" improvement over the rest of my library. In fact, the only thing that I have heard that is consistently "more impressive" has been Q-Sound. The few discs that I have that were mastered in Q-Sound have effects (perhaps not entirely natural, but certainly impressive on the right music) that one just doesn't get without Q-Sound mastering.

Material available in Q-Sound:

Amused to Death by Roger Waters
The Immaculate Collection by Madonna
Pulse by Pink
The Soul Cages by Sting
and others.

You pays your money, you takes your choice...

[novisnick]

Look who’s getting on board!
Another manufacturer adopting MQA 😁

www.technics.com/us/products/grand-class-g30/network-audio-amplifier-su-g30.html

            

[novisnick]

 
Can I have a hallelujah?

we have ignition,,,,,,,,,,,,,

[novisnick]

This gentleman has plenty of videos, this is a year old but will give some background and youll be able to find updates to it as well. 
Very informative 

                       

[mgbpuff]

Hard to understand and he's political. No thanks!

[novisnick]

Oct 5, 2017 11:07:19 GMT -5 mgbpuff said:

Hard to understand and he's political. No thanks!

I hadn’t noticed any political talk, sorry! Thats not what I was going for but he does a good job of explaining MQA.  Of cource this along with many others ive viewed on the topic have explained a lot to me. Im just not that technical guy that can explain it. 

[novisnick]

Oct 5, 2017 11:07:19 GMT -5 mgbpuff said:

Hard to understand and he's political. No thanks!

OK, please give me the time stamp of the “political talk” i must be blind or your seeing things. Txs.

You do need to listen closely but ive had proffessors like that before.

[mgbpuff]

It's very near the beginning, I didn't watch much because of it.

[novisnick]

Oct 5, 2017 12:37:02 GMT -5 mgbpuff said:

It's very near the beginning, I didn't watch much because of it.

Still don’t see it. What does he say that makes you think any of it is political?

[novisnick]

Oct 5, 2017 12:51:46 GMT -5 novisnick said:

Oct 5, 2017 12:37:02 GMT -5 mgbpuff said:

It's very near the beginning, I didn't watch much because of it.

Still don’t see it. What does he say that makes you think any of it is political?

Just rewatched again, your seeing and hearing things! Please share with the rest of us.