MQA Deep Dive - I published music on tidal to test MQA

[pjug]

I am not the best person to do this but I thought the following might jumpstart someone to do better. I ran the @2L-053 track through Deltawave, in MQA, DXD, and 17/96. The 17/96 was made from the 2L 24/96 file and truncated with WinWavBitFreezer (I don't know why DXD and MQA are a multiple of 44.1 and the other PCM is 96 and 192).

You can see that the statistics (-67dB vs -98dB, although I don't know exactly what these mean) and waveform tracking, compared to DXD, is better with the bitfrozen 17 bit than with MQA. In both cases DeltaWave downsampled the DXD so I assume this is why the two DXD spectra are not exactly the same at the high end. Also both tracks required Deltawave to adjust time to match the files.

This is just a test of the first MQA unfold. Really it would be helpful if @mansr were allowed to give us the MQA files since he also has the rendered output in addition to the decoded, where he tapped into Bluesound to get the rendering without any affects added by an ADC.

Assuming no errors on my part, is this evidence that the decoded MQA is not as good as 17 bit PCM at reproducing the DXD, or has MQA intentionally done something to the MQA file (for goodness sake)?

Maybe it is time to invite @mansr back. Time served?

 

[amirm]

macOS certainly has BSD components - the XNU kernel is an odd composition of the Mach/NeXTSTEP micro-kernel and FreeBSD/NetBSD network stack, filesystems, user-space utilities, etc. Can't do very much with just a (micro-)kernel...

To also be pendantic.

Well yeh, in those days we all took bits and pieces from every version of Unix. No commercial implementation was ever pure. But when asked, you always responded with the baseline code base and for MacOS, it is Mach.

 

[DimitryZ]

I am not the best person to do this but I thought the following might jumpstart someone to do better. I ran the @2L-053 track through Deltawave, in MQA, DXD, and 17/96. The 17/96 was made from the 2L 24/96 file and truncated with WinWavBitFreezer (I don't know why DXD and MQA are a multiple of 44.1 and the other PCM is 96 and 192).

You can see that the statistics (-67dB vs -98dB, although I don't know exactly what these mean) and waveform tracking, compared to DXD, is better with the bitfrozen 17 bit than with MQA. In both cases DeltaWave downsampled the DXD so I assume this is why the two DXD spectra are not exactly the same at the high end. Also both tracks required Deltawave to adjust time to match the files.

This is just a test of the first MQA unfold. Really it would be helpful if @mansr were allowed to give us the MQA files since he also has the rendered output in addition to the decoded, where he tapped into Bluesound to get the rendering without any affects added by an ADC.

Assuming no errors on my part, is this evidence that the decoded MQA is not as good as 17 bit PCM at reproducing the DXD, or has MQA intentionally done something to the MQA file (for goodness sake)?

Maybe it is time to invite @mansr back. Time served?

View attachment 133233
View attachment 133234
View attachment 133235
View attachment 133236

This is similar results that @Archimago did in 2017. 67dB is a very good null and is not achievable with even very good lossy codecs. Especially since the deviation is mostly in higher ultrasonics.

 

[pjug]

This is similar results that Archimago did in 2017. 67dB is a very good null and is not achievable with even very good lossy codecs. Especially since the deviation is mostly in higher ultrasonics.

I don't have a feel for the numbers but apparently it is achievable with a lossy codec if I haven't made a mistake.

 

[blueone]

As long as we are being pedantic , MacOS was built on Mach OS variant of Unix, not BSD.

Yup, and there's proof of the Mach origin in every MacOS release. Bring up the Mac Activity Monitor, double click on a process name, and a window will come up and click on "statistics". Another window comes up that lists "Mach messages in", "Mach messages out", "Mach system calls", and "Unix system calls". I've never seen a definitive history of the MacOS Unix code origin, though a common explanation I've heard from a few Apple employees who weren't involved in the original development project is that the Unix code portions originated in BSD.

 

[DimitryZ]

I don't have a feel for the numbers but apparently it is achievable with a lossy codec if I haven't made a mistake.

@Archimago had similar results in his strong work on this subject from 2017. In his results gives -70dB to MQA and -55dB to 320 MP3.

He writes this:

Obviously we can see that the MP3 version is less accurate compared to the MQA decode above. I believe for those who have tried to blind-test the sound (like what we did a few years ago), LAME MP3 encoding at 320kbps for the vast majority of folks would be "transparent" compared to lossless 16/44. Compared to the "Software MQA" decoding, the difference is a significant 15dB represented by the difference in average RMS power in the "null" file (remember, the dB scale is logarithmic).

This IMO is meaningful in terms of recognising how subtle any effect is! If this is all you remember from this blog post, it might in fact be enough .

https://archimago.blogspot.com/2017/02/comparison-hardware-decoded-mqa-using.html?m=1

 

[tmtomh]

I don't have a feel for the numbers but apparently it is achievable with a lossy codec if I haven't made a mistake.

Indeed. He's just trying to insinuate that MQA isn't lossy, or is, as MQA's recently altered FAQ page now states, "better than lossy!"

Out of curiosity, I just tried nulling a lossless redbook track with a 320k mp3 of the same track, generated in Audacity (which I think uses the LAME encoder). I set it to constant bitrate and turned off joint stereo. This is a super quick-and-dirty comparison, and I make absolutely no general claims based on it, but I thought it might be of some interest. Here's the difference file:

Difference ranges between -61.5dB and -64dB from 20Hz to about 300Hz, and from -65 to -70dB from about 300Hz to what appears to be about 19.9kHz.

So I guess the average would likely be around -64 or -65dB?

 

[pjug]

In his results gives -70dB to MQA and -55dB to 320 MP3. He writes this:


Obviously we can see that the MP3 version is less accurate compared to the MQA decode above. I believe for those who have tried to blind-test the sound (like what we did a few years ago), LAME MP3 encoding at 320kbps for the vast majority of folks would be "transparent" compared to lossless 16/44. Compared to the "Software MQA" decoding, the difference is a significant 15dB represented by the difference in average RMS power in the "null" file (remember, the dB scale is logarithmic).

This IMO is meaningful in terms of recognising how subtle any effect is! If this is all you remember from this blog post, it might in fact be enough .

https://archimago.blogspot.com/2017/02/comparison-hardware-decoded-mqa-using.html?m=1

So the goal is just to be better than 320kbps lossy? The 17 bit file looks to have much better tracking of the DXD. I didn't compress it, but the FLAC will be a good bit smaller than the MQA.

 

[DimitryZ]

So the goal is just to be better than 320kbps lossy? The 17 bit file looks to have much better tracking of the DXD. I didn't compress it, but the FLAC will be a good bit smaller than the MQA.

Sure, downsampling is a valid way to create distribution from large masters. That's what CD does.

 

[pjug]

Indeed. He's just trying to insinuate that MQA isn't lossy, or is, as MQA's recently altered FAQ page now states, "better than lossy!"

Out of curiosity, I just tried nulling a lossless redbook track with a 320k mp3 of the same track, generated in Audacity (which I think uses the LAME encoder). I set it to constant bitrate and turned off joint stereo. This is a super quick-and-dirty comparison, and I make absolutely no general claims based on it, but I thought it might be of some interest. Here's the difference file:

View attachment 133247

Difference ranges between -61.5dB and -64dB from 20Hz to about 300Hz, and from -65 to -70dB from about 300Hz to what appears to be about 19.9kHz.

So I guess the average would likely be around -64 or -65dB?

That seems pretty good.

 

[DimitryZ]

Indeed. He's just trying to insinuate that MQA isn't lossy, or is, as MQA's recently altered FAQ page now states, "better than lossy!"

Out of curiosity, I just tried nulling a lossless redbook track with a 320k mp3 of the same track, generated in Audacity (which I think uses the LAME encoder). I set it to constant bitrate and turned off joint stereo. This is a super quick-and-dirty comparison, and I make absolutely no general claims based on it, but I thought it might be of some interest. Here's the difference file:

View attachment 133247

Difference ranges between -61.5dB and -64dB from 20Hz to about 300Hz, and from -65 to -70dB from about 300Hz to what appears to be about 19.9kHz.

So I guess the average would likely be around -64 or -65dB?

@Archimago is comparing max power nulls. Average power null for MQA was 88 with most delta in the ultrasonic. That is hard to do.

 

[tmtomh]

That seems pretty good.

It does indeed - I was surprised that it was so close to the -67dB achieved with MQA.

On the notion that downsampling high-res masters to CD quality is just another method of reducing file size like MQA (which I recognize is not your notion), we should all remain clear that downsampling to CD resolution is not lossy, because the 44.1kHz sample rate has a Nyquist frequency that's above the limit of human hearing.

MQA, mp3, AAC, Vorbis, and other lossy codecs, while they differ from each other of course, all use some form of perceptual encoding. It's a qualitatively different kind of data compression than downsampling an 88.2kHz or 96kHz file to redbook/CD. It might sound just fine - but it's lossy.

(This is not a comment on bit depth, only on sample rate.)

 

[pjug]

Sure, downsampling is a valid way to create distribution from large masters. That's what CD does.

Of course. Why are you shrugging this off? I thought you wanted a comparison. The 17/96 version seems better and it is smaller.

 

[tmtomh]

Of course. Why are you shrugging this off? I thought you wanted a comparison. The 17/96 version seems better and it is smaller.

He's shrugging it off because he doesn't really want the comparison - he just used it as an opportunity to try to insinuate a false equivalence between CD/redbook and MQA under the red herring claim that CD is lossy. If you look at the pinkfishmedia.net MQA thread he was banned from, you'll get a clearer picture of his objectives and argumentative methods, and perhaps be able to more quickly come to your own conclusion about the value of engaging with him.

 

[pjug]

He's shrugging it off because he doesn't really want the comparison. If you look at the pinkfishmedia.net MQA thread he was banned from, you'll get a clearer picture of his objectives and methods, and perhaps be able to more quickly come to your own conclusion about the value of engaging with him.

If you eyeball the MQA tracking and call it -120dB, then the 17/96 must be pushing two hundred!

 

[DimitryZ]

Of course. Why are you shrugging this off? I thought you wanted a comparison. The 17/96 version seems better and it is smaller.

I am not. I listen to Qobuz and like it. I was even a subscriber when it wasn't officially available in the US.

However, Tidal/MQA were first to the market in the states and I really enjoyed listening. Continue to do so.

 

[tmtomh]

Maybe it is time to invite @mansr back. Time served?

Fully acknowledging the full authority of @amirm and the mods to make such decisions, I would respectfully echo this request.

 

[DimitryZ]

It does indeed - I was surprised that it was so close to the -67dB achieved with MQA.

On the notion that downsampling high-res masters to CD quality is just another method of reducing file size like MQA (which I recognize is not your notion), we should all remain clear that downsampling to CD resolution is not lossy, because the 44.1kHz sample rate has a Nyquist frequency that's above the limit of human hearing.

MQA, mp3, AAC, Vorbis, and other lossy codecs, while they differ from each other of course, all use some form of perceptual encoding. It's a qualitatively different kind of data compression than downsampling an 88.2kHz or 96kHz file to redbook/CD. It might sound just fine - but it's lossy.

(This is not a comment on bit depth, only on sample rate.)

It's been pointed out that the 64-65dB average for MP3 (average power) should be compared to 88dB for the same metric in MQA results. In the MP3 baseband, MQA is better still.-100dB or better.

If this is a lossy performance, I suggest a different designation - perceptually lossy to your stereo equipment, which makes it lossless to the listener by definition.

 

[lucretius]

Another observation and conclusion:
MQA is THE solution against loudness war . Because heavily clipped tracks contain square waves which cannot be MQA encoded. So the mastering is forced to avoid clipping, otherwise the track may not find its way to Tidal.

I could see some clipping in the Eagles Hotel California album -- MQA 192. I didn't unpack it, but I doubt that would change anything. Nonetheless, I don't believe the MQA encoder avoids clipping. It also doesn't appear to improve DR scores either.

 

[pkane]

I am not the best person to do this but I thought the following might jumpstart someone to do better. I ran the @2L-053 track through Deltawave, in MQA, DXD, and 17/96. The 17/96 was made from the 2L 24/96 file and truncated with WinWavBitFreezer (I don't know why DXD and MQA are a multiple of 44.1 and the other PCM is 96 and 192).

You can see that the statistics (-67dB vs -98dB, although I don't know exactly what these mean) and waveform tracking, compared to DXD, is better with the bitfrozen 17 bit than with MQA. In both cases DeltaWave downsampled the DXD so I assume this is why the two DXD spectra are not exactly the same at the high end. Also both tracks required Deltawave to adjust time to match the files.

This is just a test of the first MQA unfold. Really it would be helpful if @mansr were allowed to give us the MQA files since he also has the rendered output in addition to the decoded, where he tapped into Bluesound to get the rendering without any affects added by an ADC.

Assuming no errors on my part, is this evidence that the decoded MQA is not as good as 17 bit PCM at reproducing the DXD, or has MQA intentionally done something to the MQA file (for goodness sake)?

Maybe it is time to invite @mansr back. Time served?

View attachment 133233
View attachment 133234
View attachment 133235
View attachment 133236

This actually looks pretty good between DXD and MQA. The average spectrum isn't bad at all, and seems to start to diverge around 24kHz or so.

For fun and amusement, try the FFT Scrubber plot in DeltaWave after you match MQA and DXD/17 files. Scrubber uses a short span (400ms) FFTs and lets you move around the file sequentially, as fast as you want, or jump to random points in it, so you are more likely to catch short-lived or occasional differences that might be hidden by the average spectrum view.

Also, the spectrum of delta and delta of spectra plots make it easier to see the differences between the two files in the frequency domain.