JVS Cheerleads an MQA CD..Sis Boom Bah!

[FredericV]

If labels are selling CD's encoded with MQA, without an MQA logo on it, you're not even getting CD quality without an MQA dac.

Any cd-player with SPDIF output will pass the 16 bits bitperfect to an MQA dac, where the MQA dac will see the control data in the 16th bit (LSB bit) and try to reconstruct the lossy highres part with just a few bits borrowed. They get away with this as explained earlier, because the dynamic range of real music is not 16 bits. So the baseband signal is not lossless with MQA cd. There's buried data within redbook, not in the noise floor below redbook as that part is not available with MQA CD. The reason that the control bit signalling the presence of MQA is at the 16th bit, is because MQA expects in some applications that a 24 bit stream is truncated to 16 bits.

So placing it at the 16th bit is a convenient place, compatible with both 24 bits, 24 bits truncated to 16 bits, and MQA CD.

24 bit MQA files have something like 17 bits of actual resolution after MQA decoding.
If only 16 bits are available, I expect the resolution to be far worse, and below CD quality in terms of noise floor and dynamic range, with and without MQA decoder.

Related experiment:

One experiment I intend to do is to truncate 24 bit MQA files to 16 bits, and then send them to my Mytek, and check if the MQA light is active. In the official documentation, they stated that truncating was supported.

Then I can run audiodiffmaker to see the difference between fully decoded MQA from full entropy, vs reduced entropy.

[FredericV]

Let's test what the MQA dac does, when it receives a truncated version of 2Lno's test files:

 

[root@mqb]# sox -D -V 2L-050_01_stereo_DXD_WAV.mqa.flac -b 16 2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav
[root@mqb]# file 2L-050_01_stereo_DXD_WAV.mqa.flac 2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav
2L-050_01_stereo_DXD_WAV.mqa.flac:              FLAC audio bitstream data, 24 bit, stereo, 44.1 kHz, 8015028 samples
2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav: RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, stereo 44100 Hz

the -D disables dithering in sox

-V to see what sox does, including the processing chain
 

[FredericV]

40 minutes ago, FredericV said:

Let's test what the MQA dac does, when it receives a truncated version of 2Lno's test files:

 

[root@mqb]# sox -D -V 2L-050_01_stereo_DXD_WAV.mqa.flac -b 16 2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav
[root@mqb]# file 2L-050_01_stereo_DXD_WAV.mqa.flac 2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav
2L-050_01_stereo_DXD_WAV.mqa.flac:              FLAC audio bitstream data, 24 bit, stereo, 44.1 kHz, 8015028 samples
2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav: RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, stereo 44100 Hz

the -D disables dithering in sox

-V to see what sox does, including the processing chain
 

 

The truncated version does not light up the blue quack light on my DAC.

[FredericV]

27 minutes ago, mansr said:

No green light either?

Original MQA file 2L-050_01_stereo_DXD_WAV.mqa.flac shows blue light and renderer upsample resolution:

 

2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.wav
+ 16 bits packed in a 24 bit file (2L-050_01_stereo_DXD_WAV.mqa.truncated-24b.wav) both show:



Note that the MQA light is off and not green.
 

[FredericV]

Comparing the truncated version of an MQA file encoded back to flac, and the original encode in flac (16 bit file vs 24 bit file), we can see that the non-nyquist parts don't compress at all:

-rw-r--r-- 1 root root  31928793 Apr 26  2016 2L-050_01_stereo_DXD_WAV.mqa.flac
-rw-r--r-- 1 root root  14503274 Feb 20 18:40 2L-050_01_stereo_DXD_WAV.mqa.truncated-16b.flac

The 24 bit version is more than 2x the size of the 16 bit version. So the 8 extra bits with the secret crypto MQA drm sauce don't compress at all.

 

 

 

[FredericV]

7 minutes ago, mansr said:

Did you check with mqascan that the MQA data is still there?

 

Those tools are compiled on a raspberry somewhere, need to find it and boot it. But when I find it I will check it.

[FredericV]

2 hours ago, hsmeets said:

 

And what happens if you only truncate/mask 1 bit (the least significant) from the file? 

 

And what happens if you mask e.q. the 20th bit to zero (thus outside the encrypted MQA part).

 

Aside from that a file maybe becomes unlistenable, does the MQA authentication detect this? 

I used the wrong truncation method. New research available, which shows MQA files can be manipulated and the blue light will still shine:
 

 

[FredericV]

6 hours ago, Rt66indierock said:

Back to the topic I typed MQA cd in Google and sorted by past week. First is Stereophile of course but second is this thread. Carry on.

MQA CD will be even more rare than DTS CD

There's a chance MQA CD behaves like how it behaves in our 24 bit to 16 bit truncating topic:

1. the data for the first unfold is no longer there (as it has been truncated)

2. MQA decoding of the truncated version will still generate some fake ultrasonic content between 20 and 26 Khz because of their leaky filters, which some may misread as actual content. We measured the truncated MQA files after decoding, audio from the baseband is aliased into the ultrasonics.

3. it will still upsample to the original resolution as dictated in the control stream, so customers will believe they get 24/352.8K from a redbook CD

We would need to get our hands on an MQA CD and then the 24 bit MQA distribution file of the same track, to see if the redbook version was basically truncated, and their truncating behaves like our own truncating, or if they use a different encoding scheme for MQA CD.

The fact that there are so few MQA CD's may suggest they took the very easy route: truncating 24 bits to 16 bits.

[FredericV]

There are probably more DTS CD's than MQA-CD's .... which is to my surprise quite large:

http://www.surrounddiscography.com/dts/dtsdisc.htm

3 google hits for MQA-CD on bol.com which is targeted to Belgium & The Netherlands:

 

0 hits on amazon.fr



only 1 hit on amazon.com
3 hits on amazon.de

This one is interesting:

https://www.amazon.de/Inside-Moment-Mqa-CD-Camille-Thurman/dp/B06ZZ6GTBV

 

Quote

MQA ist eine preisgekrönte Technologie, die den Klang des Studios liefert. MQA-CDs sind vollständig kompatibel und laufen auf jedem CD-Player in normaler CD-Auflösung. Wird der Player an einen MQA-tauglichen DAC angeschlossen, entfaltet dieser die volle Auflösung von bis zu 352 kHz.

 

The description is again misleading. MQA can't do 352.8 Khz content (but it can do upsampling of the first unfold to 352.8 Khz). "Auflösung" means resolution, so the description is certainly misleading.

The first unfold when the 24 bit MQA file has been stripped to 16 bits does not happen, except for some fake content because of their leaky upsampling with aliasing. Will MQA-CD be a different beast with a different bit allocation, or does it behave like a truncated 24 bit MQA file to 16 bits?

If MQA would change "resolution" to "upsample resolution", it would reflect the truth in case claimed resolution is higher than 24/96 or 24/88.2. Now it's false advertising.
 

[FredericV]

MQA CD is not highres

 

 

On 3/16/2018 at 3:22 PM, testikoff said:

A FLAC snippet of track 01 Pulse taken off Reich: Pulse/Quartet MQA-encoded CD...

MQA CD 16/44.1, undecoded



 

MQA CD decoded to 88.2K, twice the bins needed (otherwise the plot looks competely different), no highres part added, no ultrasonics added, just some garbage at 24 Khz below the treshold of human hearing (probably some aliasing due to their leaky filters):



 

[FredericV]

14 minutes ago, adamdea said:

Have you set  audacity to 145dB scale? 

The main issue here is that we don't have the original true 24 bit files to compare with.
So going to compare their leaky upsampler with our sox config, and see if the aliasing overlaps.
Putting the scale to 24 bit is no issue.

Coming back on this post, based on earlier truncation experiments:

 

[FredericV]

39 minutes ago, Brinkman Ship said:

I thought that might be it..but for some reason I thought MQA was not compatible with S/PDIF..

 

The output of the renderer is not supported via SPDIF. So digital outputs from a device decoding MQA will always be limited to 24/88.2 or 24/96. The core decoding aka first unfold.

But SPDIF can carry the 16 or 24 bit distribution files bitperfect (e.g. streamer playing 24 bit distribution file, or redbook transport playing 16 bit truncated file with MQA ), which an MQA dac with MQA enabled SPDIF input, can decode.