MQA technical analysis

[Shadders]

3 minutes ago, abrxx said:

 

Remember that MQA is a process and that there are three potential places a file may have filters applied:

 1) The post mastering, remove "ADC blur" stage

 2) The SRC stage (i.e. converting a 192kHz MQA file into a 96Khz core)

 3) The DAC stage

 

I believe mansr is describing the filter which is applied by the MQA DAC. The majority of the deblurring already happened at stage 1 (and optionally 2).

 

My own guess is that the selection of the filter may depend on the content itself (i.e. what type of music) as well as any technical criteria (original sample rate etc). Probably selected by ear ("this ones sounds the best"). But just my guess.

Hi,

Thanks.

If the tracks are from the same album, then i assumed the same recording process. This is why i questioned the filters used.

I can see what you are saying - the filter depends on the music type etc.

So, is it really about de-blurring, or is it about making it sound nice ?

Is this why each track has to be manually processed, as the encoding affects the bit rate, or artefacts etc ?

This will imply that the processing is subjective ?

Why not just de-blur the album..... (which would mean we don't need MQA....... ???)

Regards,

Shadders.

 

[abrxx]

I don't think the MQA claim is just "deblur the music". Otherwise like you say there is no need for a lot of the process. I believe the MQA claim that they are trying to deliver "the sound of the studio" (given that you have a suitable playback system). Since a certain filter was already used in the studio to preview the sound, in theory that same filter (in the DAC) should be used in the playback at home. Otherwise, you are not getting the same sound. This makes sense to me in theory. Whether we can hear such differences, over such large anomalies such as the effect of the playback room itself (assuming stereo speaker playback), is another question.

[mansr]

16 minutes ago, abrxx said:

My own guess is that the selection of the filter may depend on the content itself (i.e. what type of music) as well as any technical criteria (original sample rate etc). Probably selected by ear ("this ones sounds the best"). But just my guess.

A simple experiment would be to take an actual MQA file, switch the rendering filter, and see if anyone can hear a difference. Who wants to pick the test music?

[abrxx]

Just now, mansr said:

A simple experiment would be to take an actual MQA file, switch the rendering filter, and see if anyone can hear a difference. Who wants to pick the test music?

I am certain I don't own a system with enough resolution to hear the difference. Plus I don't actually have a MQA rendering capable DAC either. But I like the idea in principle!

[mansr]

Just now, abrxx said:

I am certain I don't own a system with enough resolution to hear the difference. Plus I don't actually have a MQA rendering capable DAC either. But I like the idea in principle!

You wouldn't need an MQA DAC. I could do the rendering in software.

[abrxx]

2 minutes ago, mansr said:

You wouldn't need an MQA DAC. I could do the rendering in software.

Maybe I missed something, but if you do that then both files would have an additional filter applied by my own DAC. Doesn't that invalidate the experiment? Perhaps the additional filter makes the previous filter inaudible.

[mansr]

6 minutes ago, abrxx said:

Maybe I missed something, but if you do that then both files would have an additional filter applied by my own DAC. Doesn't that invalidate the experiment? Perhaps the additional filter makes the previous filter inaudible.

All DACs have their own filters, and there's nothing MQA can do about that.

[Shadders]

12 minutes ago, abrxx said:

I don't think the MQA claim is just "deblur the music". Otherwise like you say there is no need for a lot of the process. I believe the MQA claim that they are trying to deliver "the sound of the studio" (given that you have a suitable playback system). Since a certain filter was already used in the studio to preview the sound, in theory that same filter (in the DAC) should be used in the playback at home. Otherwise, you are not getting the same sound. This makes sense to me in theory. Whether we can hear such differences, over such large anomalies such as the effect of the playback room itself (assuming stereo speaker playback), is another question.

Hi,

My interpretation that MQA is to deliver the sound of the studio, is a subjective marketing statement. The deblur (correcting temporal smear) is the key aspect of MQA that is supposed to present what they heard in the studio. Other aspects in the MQA processing are the reduction of the file size, which is a lossy process. So, not really the "sound of the studio".

Why they have a different filter for the 3 groups of tracks is unknown, if it is the same DAC playing back all tracks. As you have indicated, if we have the same filter in the studio, and hence in the DAC, then we hear the same, but did the studio use three different filters when listening ?

My interpretation is that the filters are changed to make it sound nice.

Regards,

Shadders.

[abrxx]

2 minutes ago, Shadders said:

Hi,

My interpretation that MQA is to deliver the sound of the studio, is a subjective marketing statement. The deblur (correcting temporal smear) is the key aspect of MQA that is supposed to present what they heard in the studio. Other aspects in the MQA processing are the reduction of the file size, which is a lossy process. So, not really the "sound of the studio".

Why they have a different filter for the 3 groups of tracks is unknown, if it is the same DAC playing back all tracks. As you have indicated, if we have the same filter in the studio, and hence in the DAC, then we hear the same, but did the studio use three different filters when listening ?

My interpretation is that the filters are changed to make it sound nice.

Regards,

Shadders.

I hear what you are saying.

 

On a slightly different note, are all the filters minimum phase? If not, any MQA tracks out there not using minimum phase filters? I've heard some people say that piano (for example) does not sound the best when using a minimum phase filter.

[mansr]

16 minutes ago, abrxx said:

On a slightly different note, are all the filters minimum phase?

Yes. See the impulse response plots earlier in this thread.

[soxr]

3 hours ago, mansr said:

A simple experiment would be to take an actual MQA file, switch the rendering filter, and see if anyone can hear a difference. Who wants to pick the test music?

 

I have a very revealing system to do this. And I almost got your tools fully working

Wednesday one speaker distributor even preferred minimum ohase upsampled MQA over DXD.
The world is a strange place.

[The Computer Audiophile]

There's no accounting for taste. 

[mansr]

2 minutes ago, soxr said:

And I almost got your tools fully working

If anything is unclear, don't hesitate to ask.

[Jud]

3 hours ago, mansr said:

You wouldn't need an MQA DAC. I could do the rendering in software.

 

If you can then convert it to DSD256 without monkeying with the minimum phase....

 

I don't know what's in MQA besides my memory that I liked Joni Mitchell's "Blue."  (Felt that I liked it slightly better non-MQA.)

 

In fact if you could do three files, MQA, non-MQA with minimum phase filtering, and non-MQA with linear phase filtering, all converted to DSD256, that would be just about perfect.  If you don't like Joni, anything sparsely produced, with vocals, acoustic guitar, perhaps piano, would be nice.

 

Just my tuppence.

[mansr]

30 minutes ago, Jud said:

If you can then convert it to DSD256 without monkeying with the minimum phase....

 

I don't know what's in MQA besides my memory that I liked Joni Mitchell's "Blue."  (Felt that I liked it slightly better non-MQA.)

 

In fact if you could do three files, MQA, non-MQA with minimum phase filtering, and non-MQA with linear phase filtering, all converted to DSD256, that would be just about perfect.  If you don't like Joni, anything sparsely produced, with vocals, acoustic guitar, perhaps piano, would be nice.

My idea was to compare the various MQA render filters. That would mean decoding an MQA track to 96 kHz, then changing the filter setting before feeding it to the renderer for conversion to 192 or 384 kHz. Playing the resulting files (with DSD conversion if you wish) would reveal whether the filter selection actually makes any audible difference.

[Jud]

4 minutes ago, mansr said:

My idea was to compare the various MQA render filters. That would mean decoding an MQA track to 96 kHz, then changing the filter setting before feeding it to the renderer for conversion to 192 or 384 kHz. Playing the resulting files (with DSD conversion if you wish) would reveal whether the filter selection actually makes any audible difference.

 

That sounds fine.  You doing the DSD conversion of a 384kHz file would save my 2009 MBP from having to wheeze along doing it on the fly.  I would prefer not to have a 192kHz file, just to ensure my own A+ filter settings play no role.

 

Edit: Actually, I suppose even with a 384kHz file, conversion to DSD256 on a Mac would involve my A+ filter settings.  So another reason to prefer DSD256, at least in my view.  However, if this a PITA for you, as you are doing this on a volunteer basis, I would be rude to demand anything different than what you would choose to supply.

[soxr]

On 14/7/2017 at 11:00 AM, mansr said:

Yes, it looks like your system libraries aren't compatible with the Bluesound one. You can always use chroot or other tricks to pick up everything from the Bluesound rootfs.

 

Regarding armv7 on raspberry 2 and 3, I found armv7 images into which I can do a chroot into the untarred version instead of flashing another raspberry SD card as they keep piling up or get lost easily otherwise:

https://archlinuxarm.org/about/downloads

 

http://os.archlinuxarm.org/os/ArchLinuxARM-rpi-2-latest.tar.gz
http://os.archlinuxarm.org/os/ArchLinuxARM-rpi-3-latest.tar.gz

 

raspberry can now see the correct deps:
 

root@mqb:~# chroot /mnt/armv7/
[root@mqb /]# ldd /usr/lib/libbluos_ssc.so
        linux-vdso.so.1 (0x7ecb5000)
        libasound.so.2 => not found
        libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x76dd3000)
        libm.so.6 => /usr/lib/libm.so.6 (0x76d4f000)
        libc.so.6 => /usr/lib/libc.so.6 (0x76c0b000)
        libgcc_s.so.1 => /usr/lib/libgcc_s.so.1 (0x76bde000)
        /usr/lib/ld-linux-armhf.so.3 (0x54ad0000)

Let's compile those tools ...

Update: everything compiles on arch linux for raspberry pi II.
 

[soxr]

This library is acceptable fast. Less than 8 secs, but it seems I selected the smallest demo file from 2L.no

# time ./armv7l/mqadec 2L-056_03_stereo_DXD_mqa.flac mqa.flac
mqaAuthored/88200

real    0m7.975s
user    0m7.600s
sys     0m0.340s

# time ./armv7l/mqadec 2L-056_03_stereo_DXD_mqa.wav mqa.wav
mqaAuthored/88200

real    0m7.770s
user    0m7.160s
sys     0m0.590s

The flac vs wav overhead is very small.

The longest MQA file, which is 2L-038_01_stereo_DXD_FLAC.mqa.flac with duration of  09:14, takes 1m50.181s

[soxr]

All 32 impulse responses have been dumped.
Great work again from Archimago and mansr

http://archimago.blogspot.be/2017/07/measurements-audioquest-dragonfly-black.html

[FredericV]

Just analyzed 2L-053_04_stereo-DXD.flac and it's MQA counterpart.

Dark purple is what MQA keeps after full unfold, light purple is the original file.

Some reviewers still believe MQA encodes information above 24/96 ..... they are clearly misinformed and spreading these false claims. Like spreading fake news.

[FredericV]

My overlay was not 100% correct in the gimp, here's the correct one now.

It shows the energy in the ultrasonics has mostly gone.
 

[FredericV]

I'm having a bad gimp overlay day. My previous overlays don't combine well.

So here are the two.

DXD:



Second unfold:


 

Second unfold creates content that was not even in the original.

[mansr]

Here's a better plot:

[mansr]

Here are the actual upsampling filter coefficients from the Bluesound library:

https://code.videolan.org/mansr/mqa/blob/master/render-filters.txt

[witchdoctor]

On 7/14/2017 at 4:48 PM, abrxx said:

I don't think the MQA claim is just "deblur the music". Otherwise like you say there is no need for a lot of the process. I believe the MQA claim that they are trying to deliver "the sound of the studio" (given that you have a suitable playback system). Since a certain filter was already used in the studio to preview the sound, in theory that same filter (in the DAC) should be used in the playback at home. Otherwise, you are not getting the same sound. This makes sense to me in theory. Whether we can hear such differences, over such large anomalies such as the effect of the playback room itself (assuming stereo speaker playback), is another question.

You hit the nail on the head, they are trying to deliver the sound of the studio. The suitable playback system is a HUGE part of the chain. I think you can replicate the sound in the studio without a lot of effort or budget. 
1) a Mytek Brooklyn DAC because the Mytek pro ADC and DAC are what the labels are using to convert their catalogs and the studios are using to record MQA tracks.

2) A good pair of active speakers/montors for two channel.

3) Mogami cables

4) A Tidal subscription.

The sound of the studio may not win best room at an audio show but if you want accuracy not euphony is a good choice.