MQA is Vaporware

[kumakuma]

削除した。

[mansr]

1 hour ago, Fitzcaraldo215 said:

Sorry, my request was vague.  I intended that he let me know which MQA DAC he was reading the micro instructions from its DSP as it performed the MQA unfolding and processing.  He alluded to knowing the instructions actually performed on the MQA stream.  We know, I guess, he does not have actual source code, since that is proprietary.

I do indeed not have the source code, but even without it, it's still possible to figure out how it works. The Bluesound decoder runs on ARM which is a well-known CPU. The decoder itself is quite generic and not tailored to the device. In fact, the output from the core decoder is bit for bit identical with digital captures from Tidal.

[Fitzcaraldo215]

Just now, kumakuma said:

 

Why does this matter?

Only slightly.  The contention is 2L files recorded at 352k contain no content from the original recording above 88k when processed and put into an MQA 44k/24 bit container file, and that apparent content beyond 88k is all synthesized via upsampling by Bluesound on playback in spite of claims by the recording engineers to the contrary.  So, there is only one MQA unfolding at twice the recorded sampling rate, contrary to everything MQA and 2L have gone to extremely great length to say, over and over.

 

Yes, he tried some 2L recordings and that issue should show up even going from 88k to 176k, which he claims is nothing but an upsampled 88k via Bluesound.  However, I am still very unclear on exactly what was tested and how.  A few 2L recordings were done natively at 44k or 96k, so they would have had no content at higher sampling rates under MQA, and MQA makes this quite clear.  I hope they caused no confusion or errors in his testing.

 

Sorry to get hung up on technicalities, although I think these particular accusations were potentially hugely significant, because they seemingly had very specific technical substance and they involved a third party, 2L.

 

FWIW, I am nether lawyer nor engineer.  I am just an enquiring mind that values  verifiable truth, especially in audio, tough or even impossible as that is to achieve.  Fitzcaraldo is my moniker, but it might as well be Don Quichote or Teiresias, or so my wife says.

 

[mansr]

5 minutes ago, Fitzcaraldo215 said:

A few 2L recordings were done natively at 44k or 96k, so they would have had no content at higher sampling rates under MQA, and MQA makes this quite clear.  I hope they caused no confusion or errors in his testing.

I'm well aware of that. Those use a different upsampling filter with a lot less aliasing issues.

[Bob Stern]

4 hours ago, Don Hills said:

Chris specifically asked him not to post technical details on CA.

 

Chris's original request was based on his erroneous assumption that posting technical details obtained by analyzing the behavior of copyrighted but publicly available software would infringe patents or copyrights.  I jumped in with a post stating that such disclosures would be perfectly legal.  Chris did not respond to my post or object to any subsequent discussion, which I interpret as tacit acquiescence.  

 

More specifically, I stated: (1) The MQA patent claims expressly exclude pure software implementations. So even if someone were reverse-engineering MQA for the purpose of writing MQA decoding software, that would be perfectly legal. (2) Copyright of the software does not protect the data streams produced by the software or any analysis thereof.  Here's my prior post:

 

 

[Bob Stern]

-- 

[mansr]

24 minutes ago, Bob Stern said:

Chris's original request was based on his erroneous assumption that posting technical details obtained by analyzing the behavior of copyrighted but publicly available software would infringe patents or copyrights.  I jumped in with a post stating that such disclosures would be perfectly legal.  Chris did not respond to my post or object to any subsequent discussion, which I interpret as tacit acquiescence. 

Nevertheless, facing a frivolous lawsuit can be very costly, even to the point of bankruptcy, and I wouldn't want to put Chris at any financial risk.

[Fokus]

14 hours ago, Fitzcaraldo215 said:

 

http://www.2l.no/hires/documentation/2L-MQA_Comparisons.pdf

 

They state that the MQA version folds recordings done at 352k down 3 times to 44k. That is considerably different from your version.

 

Your position is based on just that short statement, that a 352k recording is folded three times, in a blurb. That's all.

 

1) if you study the MQA patents, you'll find that there is no room for such 3 times folding.

 

2) if you study MQA's technical literature, as was made public since 2014 or so, you'll find that there is no 3 times folding. What you may find is an allowance to encode and decode a small portion of original signal above 48kHz, in other words, a little bit of a 192k recording (up to 56kHz or so). Not a 352/384k one.

 

3) if you understand how MQA works (which is not that hard at all), then you'll see that thruthful three times folding is not possible, unless the original signal has no content above, say, 48kHz.

 

4) there are a few post-MQA DAC spectral analyses on the web from files that are known to originate from 192k originals, which upon playback through an MQA DAC light the DAC's '192k' indicator. These spectra consistently show the result of lazy upsampling above 48kHz. No actual original content is conveyed.

In order for a recorded spectrum to show the fingerprint of this upsampling the signal has to contain a lot of treble. This can come from the music, but also from high-frequency interference (from CRTs and whatever) originally recorded with the music in the studio. In my experience 2L recordings do not contain much treble (at least the ones I tried), and were cleanly recorded so as not to be contaminated. That, and the typical modulation noise bump above 40kHz of many DAC chips and many ADC chips make these recordings hard to fingerprint. But again, AFAIK, all 192k recordings analysed so far show upsampling as the sole mechanism to get back to 192k upon MQA playback.

 

Of course you have the right to keep on dreaming.

 

[esldude]

1 hour ago, Fokus said:

 

Your position is based on just that short statement, that a 352k recording is folded three times, in a blurb. That's all.

 

1) if you study the MQA patents, you'll find that there is no room for such 3 times folding.

 

2) if you study MQA's technical literature, as was made public since 2014 or so, you'll find that there is no 3 times folding. What you may find is an allowance to encode and decode a small portion of original signal above 48kHz, in other words, a little bit of a 192k recording (up to 56kHz or so). Not a 352/384k one.

 

3) if you understand how MQA works (which is not that hard at all), then you'll see that thruthful three times folding is not possible, unless the original signal has no content above, say, 48kHz.

 

4) there are a few post-MQA DAC spectral analyses on the web from files that are known to originate from 192k originals, which upon playback through an MQA DAC light the DAC's '192k' indicator. These spectra consistently show the result of lazy upsampling above 48kHz. No actual original content is conveyed.

In order for a recorded spectrum to show the fingerprint of this upsampling the signal has to contain a lot of treble. This can come from the music, but also from high-frequency interference (from CRTs and whatever) originally recorded with the music in the studio. In my experience 2L recordings do not contain much treble (at least the ones I tried), and were cleanly recorded so as not to be contaminated. That, and the typical modulation noise bump above 40kHz of many DAC chips and many ADC chips make these recordings hard to fingerprint. But again, AFAIK, all 192k recordings analysed so far show upsampling as the sole mechanism to get back to 192k upon MQA playback.

 

Of course you have the right to keep on dreaming.

 

 

Yes, if you study the patents you'll see everything Fokus said is true.

 

Measured results from Mansr and Miska show nothing to disagree with the above description. 

[Fokus]

More analysis here http://www.pinkfishmedia.net/forum/showpost.php?p=3072610&postcount=619

[Fitzcaraldo215]

17 hours ago, mansr said:

I do indeed not have the source code, but even without it, it's still possible to figure out how it works. The Bluesound decoder runs on ARM which is a well-known CPU. The decoder itself is quite generic and not tailored to the device. In fact, the output from the core decoder is bit for bit identical with digital captures from Tidal.

That your findings were based on your analysis of the Bluesound decoder was a misunderstanding on my part.  Sorry.

 

After sleeping on it, and assuming I now have a slightly better understanding of the data you have obtained, here are my current thoughts, for what they are worth.

 

Your testing was only of the Bluesound software, I believe.  I think without further testing of a wider range of MQA implementations, it is presumptious to say that MQA in toto is a fraud that does not conform to statements made repeatedly by Stuart and others about its capabilities.

 

If you have a provable problem, it seems to me that complaint should go to Bluesound, not 2L as I previously suggested.  It is possible that the Bluesound decoder contains an error or a bug.  But, maybe not, and,  as you suggest, it implements the same deception embodied in all of MQA.  We just do not know for sure.

 

The question of what the 2L MQA files contain in terms of higher resolutions is not known based on your testing.  Your testing may only have established what the Bluesound decoder does on playback, and there may actually be 352k content in higher order folds as stated by Stuart and Lindberg.

[mansr]

11 minutes ago, Fitzcaraldo215 said:

That your findings were based on your analysis of the Bluesound decoder was a misunderstanding on my part.  Sorry.

 

After sleeping on it, and assuming I now have a slightly better understanding of the data you have obtained, here are my current thoughts, for what they are worth.

 

Your testing was only of the Bluesound software, I believe.  I think without further testing of a wider range of MQA implementations, it is presumptious to say that MQA in toto is a fraud that does not conform to statements made repeatedly by Stuart and others about its capabilities.

Repeating a lie doesn't make it true.

11 minutes ago, Fitzcaraldo215 said:

If you have a provable problem, it seems to me that complaint should go to Bluesound, not 2L as I previously suggested.  It is possible that the Bluesound decoder contains an error or a bug.  But, maybe not, and,  as you suggest, it implements the same deception embodied in all of MQA.  We just do not know for sure.

That would be a hell of a bug let slip past MQA's certification testing.

11 minutes ago, Fitzcaraldo215 said:

The question of what the 2L MQA files contain in terms of higher resolutions is not known based on your testing.  Your testing may only have established what the Bluesound decoder does on playback, and there may actually be 352k content in higher order folds as stated by Stuart and Lindberg.

Recordings of the analogue outputs of other DACs, e.g. Meridian Explorer 2 and Mytek Brooklyn, all show the same thing, MQA adding aliases of low frequencies where the original high-res master had nothing at all.

 

Now answer this, is there anyone you'd trust who isn't Bob Stuart or one of his acolytes?

[AJ Soundfield]

Hang on...

[Fitzcaraldo215]

16 hours ago, mansr said:

Nevertheless, facing a frivolous lawsuit can be very costly, even to the point of bankruptcy, and I wouldn't want to put Chris at any financial risk.

A significant new wrinkle has emerged in this question of the handling of hi rez by MQA.  I was totally unaware of it until just now.  Perhaps @mansr was, but I am not sure.  Were you aware, @mansr?

 

I hate to give Lavorgna a plug, but this summarizes the situation, and note Stuart's comments:

 

https://www.audiostream.com/content/mqa-decoding-explained

 

There is a key difference between an MQA Core Decoder and a Full Decoder.  A Core Decoder only does 1 fold to 88 or 96k, exactly as @mansr described.  A Full Decoder can do multiple unfolds up to 384k sampling (possibly more?). 

 

Software, such as Tidal or Bluesound, is only capable of being a Core Decoder, because the DAC properties are not known or controllable by the software.  The Full Decoder capability, including multiple unfolds, exists only in an MQA DAC.  Although, I think possibly Dragonfly's and such may be restricted also.

 

So, it now seems that @mansr did in fact measure only a single unfold to 88 or 96k,  precisely because he was using the Bluesound decoder, which is Core only.  Not in MQA, but elsewhere in his setup, there was also an upsampling to 176/192k -SoX, perhaps?  Testing for the unfolding of higher sampling rates could only be done on an MQA DAC with Full Decoder, which I think we all realize, would be quite difficult.

 

I thought it unlikely or impossible, even, but how could @mansr and Stuart possibly both have been right?  This, I believe, fully answers that question.  They are both right.

 

I have learned something.   I hope there are no hard feelings on @mansr's part. However,  I do believe, in the interest of honesty and fairness,  that one should avoid making statements that say unequivocally MQA cannot handle hi rez above 88/96k except via upsampling.  There is no basis to believe it cannot when using the Full Decoder in an MQA DAC.

 

[Fokus]

Fitz, it is clear that you don't even remotely understand how this works.

 

we have been aware of the precise difference between Core and Renderer from day one.

 

The Bluesound DAC does full decoding. The software does a first unfold, followed with upsampling. The result of that is sent to the DAC chip, whose oversampling is maximally suppressed (as this is now done in the SW). It is all very simple. If you stay clear of MQA Newspeak.

 

[kumakuma]

15 minutes ago, Fitzcaraldo215 said:

A significant new wrinkle has emerged in this question of the handling of hi rez by MQA.  I was totally unaware of it until just now.  Perhaps @mansr was, but I am not sure.  Were you aware, @mansr?

 

I hate to give Lavorgna a plug, but this summarizes the situation, and note Stuart's comments:

 

https://www.audiostream.com/content/mqa-decoding-explained

 

There is a key difference between an MQA Core Decoder and a Full Decoder.  A Core Decoder only does 1 fold to 88 or 96k, exactly as @mansr described.  A Full Decoder can do multiple unfolds up to 384k sampling (possibly more?). 

 

Software, such as Tidal or Bluesound, is only capable of being a Core Decoder, because the DAC properties are not known or controllable by the software.  The Full Decoder capability, including multiple unfolds, exists only in an MQA DAC.  Although, I think possibly Dragonfly's and such may be restricted also.

 

So, it now seems that @mansr did in fact measure only a single unfold to 88 or 96k,  precisely because he was using the Bluesound decoder, which is Core only.  Not in MQA, but elsewhere in his setup, there was also an upsampling to 176/192k -SoX, perhaps?  Testing for the unfolding of higher sampling rates could only be done on an MQA DAC with Full Decoder, which I think we all realize, would be quite difficult.

 

I thought it unlikely or impossible, even, but how could @mansr and Stuart possibly both have been right?  This, I believe, fully answers that question.  They are both right.

 

I have learned something.   I hope there are no hard feelings on @mansr's part. However,  I do believe, in the interest of honesty and fairness,  that one should avoid making statements that say unequivocally MQA cannot handle hi rez above 88/96k except via upsampling.  There is no basis to believe it cannot when using the Full Decoder in an MQA DAC.

 

 

 

[Fitzcaraldo215]

13 minutes ago, Fokus said:

Fitz, it is clear that you don't even remotely understand how this works.

 

we have been aware of the precise difference between Core and Renderer from day one.

 

The Bluesound DAC does full decoding. The software does a first unfold, followed with upsampling. The result of that is sent to the DAC chip, whose oversampling is maximally suppressed (as this is now done in the SW). It is all very simple. If you stay clear of MQA Newspeak.

 

When you say "we" were aware, that apparently does not include @mansr, who implied that both Stuart and 2L were liars, but based on his observations of the Bluesound decoder PC software, not the DAC.  Had he been aware, he would have known the difference.

 

Glad you were aware though.  But, why did you not speak up once Bluesound software came into the picture yesterday?

[mansr]

5 minutes ago, Fitzcaraldo215 said:

When you say "we" were aware, that apparently does not include @mansr, who implied that both Stuart and 2L were liars, but based on his observations of the Bluesound decoder PC software, not the DAC.  Had he been aware, he would have known the difference.

There is no Bluesound PC software, it's all in the DAC firmware. Besides, even if the Bluesound weren't fully featured, the Explorer 2 shows the exact same behaviour. How do you explain away that?

 

[mansr]

41 minutes ago, Fitzcaraldo215 said:

Testing for the unfolding of higher sampling rates could only be done on an MQA DAC with Full Decoder,

Absolutely.

41 minutes ago, Fitzcaraldo215 said:

which I think we all realize, would be quite difficult.

Not in the slightest.

[Fitzcaraldo215]

21 minutes ago, mansr said:

There is no Bluesound PC software, it's all in the DAC firmware. Besides, even if the Bluesound weren't fully featured, the Explorer 2 shows the exact same behaviour. How do you explain away that?

 

Ok.  I know nothing about a lot of things, including Bluesound and Meridian.  Based on the behavior observed, I would say it is likely both only have the Core Decoder in firmware.  I could be wrong.  So could you if you insist you know they have Full Decoder without verification.  Why not ask them?

[Fitzcaraldo215]

17 minutes ago, mansr said:

Absolutely.

Not in the slightest.

OK.  Glad to hear that.  

 

I still think it is premature to brand Stuart a liar without testing more DACs.  

[kumakuma]

3 minutes ago, Fitzcaraldo215 said:

Ok.  I know nothing about a lot of things, including Bluesound and Meridian.  Base on the behavior observed, I would say it is likely both only have the Core Decoder in firmware.  I could be wrong.  So could you if you insist you know they have Full Decoder without verification.  Why not ask them?

 

Full Decoder:

 

https://helpdesk.bluesound.com/discussions/viewtopic.php?f=4&t=4059

[Fitzcaraldo215]

2 hours ago, mansr said:

Repeating a lie doesn't make it true.

Agreed.  

2 hours ago, mansr said:

That would be a hell of a bug let slip past MQA's certification testing.

Agreed.

2 hours ago, mansr said:

Recordings of the analogue outputs of other DACs, e.g. Meridian Explorer 2 and Mytek Brooklyn, all show the same thing, MQA adding aliases of low frequencies where the original high-res master had nothing at all.

Does the alias thing actually prove that they are all only upsampling over 96k, as you stated?  I thought higher than normal aliasing was a fact of life in the tradeoffs made for MQA's deblurring, allegedly as determined by their listening tests.  

 

Were you able to trace through instruction by instruction on the Meridian and Mytek as you had on the Bluesound?

 

2 hours ago, mansr said:

 

Now answer this, is there anyone you'd trust who isn't Bob Stuart or one of his acolytes?

I have known about Bob Stuart and his products for decades.  I see no reason to disbelieve him, any more than I would disbelieve, let's say, Floyd Toole. Neither infallibly walks on water, however.  And, new information can change my view of most anybody.  BTW, I also have very high regard for Morten Lindberg of 2L.  His endorsement of MQA means something to me.

 

Stuart also seems a good businessman, so I doubt he would be telling the biggest audio lies in my memory, because the stakes are too high legally and commercially.  He could lose everything if he got caught in the fraud you allege he is committing. But, you lose nothing, even if your claims are totally wrong.  So, I tend to believe him much more than I believe you.  He has much more at risk, even in terms  of "sweat equity" in his ideas.  Of course, people thought Bernie Madoff was a good, trustworthy guy, too, for many years.  

 

So, my trust for Stuart is high.  I also think his ideas for MQA make considerable sense.  But, the commercial success of MQA is far from guaranteed.

 

 

[Cebolla]

21 hours ago, kumakuma said:

削除した。

サイバーマン？

[kumakuma]

1 minute ago, Cebolla said:

サイバーマン？

 

サラリーマン