MQA is Vaporware

[lucretius]

2 minutes ago, Cebolla said:

 

Odd that given the many of those tests & the conclusions were made by mansr himself, you doubt his observation in tests of never having seen an MQA-CD file with more than 1 bit for MQA authentication and further, prefer to believe that this contradicts his observation in tests of never having seen an MQA file with more than 2 bits for MQA authentication.

 

It is unclear to me what you are saying.  Mansr's 14-bit comment appears to contradict his 15-bit comment.  Nonetheless, I believe his testing specifically involved identifying non-music bits.  So where he could only identify 1 bit of the 16, he concluded the other 15 bits where PCM music bits?  In any case, within these 16 bits there must be coded the authentication, filter selection, and sample rate.  Can 1 bit accommodate this?  

[lucretius]

7 minutes ago, Cebolla said:

 

Goodness, please go back & actually read about some of those 'tests', eg:

 

 

What are you referring to exactly?  Of course, the lower 8 bits can be dropped from a 24 bit MQA file without loss of authentication. This is what happens when 16 bit MQA files are made from 24 bit MQA files.

 

 

 

[Cebolla]

For goodness sake, actually read the whole test mentioned in the posts that followed that post! 

 

Just some of the phrases to out for:

"the control stream has a packed structure" - so very much capable of containing original sample rate, MQA filter selection, etc; 

"Corrupting the control bit (counting from 0, bit 8)" - singular, not the 3 control bits!;

"contrary to corrupting the bit just above it (counting from 0, bit 9)" - obviously referring to adjoining 0-24kHz PCM bit;

"when I corrupt the control bit" - ditto singular.

 

[UkPhil]

42 minutes ago, lucretius said:

 

There are two sources for MQA-CD. The first is where the master is 44.1/16 bit. The second is where an MQA 16-bit file was made by removing the lower 8 bits of an 24 bit MQA file (this appears to be the case for the file you noted). And in this latter case, it would appear that there are no unfolds or upsampling. I could be wrong -- there definitely is no unfold to 88.2/96k but there could be straight-up upsampling?

 

 

[PeterSt]

33 minutes ago, lucretius said:

In any case, within these 16 bits there must be coded the authentication, filter selection, and sample rate.  Can 1 bit accommodate this?

 

Of course it can !

If you lay out all those one bits and for example observe chunks of 8 bits, then each such trunk can hold 256 "situations". Need more ? then make 16 bit chunks (65536 "situations" now) (or 11 bit or 25 bit or ...).

[Cebolla]

53 minutes ago, lucretius said:

 

It is unclear to me what you are saying.  Mansr's 14-bit comment appears to contradict his 15-bit comment.  Nonetheless, I believe his testing specifically involved identifying non-music bits.  So where he could only identify 1 bit of the 16, he concluded the other 15 bits where PCM music bits?  In any case, within these 16 bits there must be coded the authentication, filter selection, and sample rate.  Can 1 bit accommodate this?  

 

1 bit can certainly accommodate that, plus I cannot see any contradiction & share @danadam's thoughts:

 

 

[lucretius]

2 hours ago, Cebolla said:

For goodness sake, actually read the whole test mentioned in the posts that followed that post! 

 

Just some of the phrases to out for:

"the control stream has a packed structure" - so very much capable of containing original sample rate, MQA filter selection, etc; 

"Corrupting the control bit (counting from 0, bit 8)" - singular, not the 3 control bits!;

"contrary to corrupting the bit just above it (counting from 0, bit 9)" - obviously referring to adjoining 0-24kHz PCM bit;

"when I corrupt the control bit" - ditto singular.

 

 

I wasn't sure whether "control stream" (as opposed to the "control bit 9") was referring just to bit 9 (counting LSB from 0). And where does filter selection and sample rate fit in?

 

[bogi]

5 hours ago, lucretius said:

 

There's not even a second unfold. There's one unfold (to 96k) with a x2 (times 2) or x4 (times 4) or x8 (times 8) upsampling. In MQA speak, x2, x4, and x8 are misleadlingly characterized as a second, third, and fourth unfold, respectively.   

If it is done in hardware, it could be about how DAC chip or other chip does it. It is known about older DAC chips (the BurrBrown ones used in iFi Micro etc. were discussed here years ago) do oversampling in stages or levels, where each stage doubles the sample rate. So for example from 44.1 to 352.8 it is done in 3 stages.

Or it may be only an unfolding propaganda, who knows. 😏

[danadam]

1 hour ago, lucretius said:

And where does filter selection and sample rate fit in?

Here is MQA bitstream description from mansr: https://code.videolan.org/mansr/mqa/-/blob/master/mqa.rst

I'm guessing filter selection and sample rate would be in "render_filter" and "orig_rate" fields of datasync packet.

[bogi]

Here mansr really writes that bit 8 (counting from bit 0 - the least significant one) is used for control stream. That's what Cebolla says.
 

This is by design. The MQA identification and authentication data is embedded in bit 8 of a 24-bit PCM stream. Dropping bits 0-7 thus leaves it untouched. The authentication works by computing a Blake2s hash over the top 15 bits (the plain PCM portion) and parts of the control stream in bit 8. This hash is then verified against a cryptographic signature extracted from the control stream. The public key is stored in the decoder. If the signature matches, the blue light goes on. The low 8 bits encoding the high-frequency content are not covered by the authentication.

 

[KeenObserver]

The bottom line is that MQA damages the original recording.  If you play it on a non MQA playback system you are getting a very damaged copy of the original.  If you play MQA on a MQA playback system you are still getting a damaged copy of the original.

 

MQA tells you that MQA is better than the original master. That is pure marketing BS.  The only way you could convince people that MQA is to their benefit is to tell people that it is "better".  PURE BS. 

 

Sadly, there are a number of people that actually believe the MQA marketing BS.

[Cebolla]

1 hour ago, lucretius said:

I wasn't sure whether "control stream" (as opposed to the "control bit 9") was referring just to bit 9 (counting LSB from 0). 

 

No idea where you got that from. Also, just look a few posts above, in the same test conversation, where mansr clearly mentions that bit 9 is "just above the MQA control stream":

Which, given that bits 0 to 7 are the eight LSB 24kHz-48kHz 'folded' bits, just leaves the single bit 8 as the "control stream" mentioned by mansr, ie, the exact same single "control bit (counting from 0, bit 8)" mentioned later in the test conversation by @FredericV. 

 

 

 

1 hour ago, lucretius said:

And where does filter selection and sample rate fit in?

 

What not convinced by the potential vast capacity implied by stream (as in "control stream") and packed structure (not to mention @Don Hills's & @PeterSt's posts above), eh? 🙂

 

How about this mansr post on the first page of that test thread?

"The displayed rate is whatever the control stream says. That's the orig_rate value in the mqascan output."

 

[bogi]

And in the next post mansr writes:
The displayed rate is whatever the control stream says.
So that's the source of misleading information discussed yesterday.
That seems to be result of stripping 8 lower bits form 24bit MQA content. The control stream is not affected by this stripping so it will show for example 88.2 with 16bit content. Nobody at Tidal or MQA cares ...

[KeenObserver]

Let me see if I get this. 16 bit MQA is not sixteen bit.  24 bit MQA is not 24 bit. And the higher play rates are just up sampled from the corrupted MQA file?

Does that about explain the situation?

 

The bottom line is that the music consumer is being screwed and is being forced to pay for it.

[bogi]

8 minutes ago, KeenObserver said:

Let me see if I get this. 16 bit MQA is not sixteen bit.  24 bit MQA is not 24 bit. And the higher play rates are just up sampled from the corrupted MQA file?

Does that about explain the situation?

 

The bottom line is that the music consumer is being screwed and is being forced to pay for it.

 

16bit MQA is probably 15bit PCM content + 1bit control stream.

24bit MQA is 16bit MQA and 8 bits in group 4+4: 4 bits for lossy encoding of high frequency content and another 4bits to control "unfolding" = upsampling, filtering.

[Cebolla]

42 minutes ago, bogi said:

And in the next post mansr writes:
The displayed rate is whatever the control stream says.
So that's the source of misleading information discussed yesterday.
That seems to be result of stripping 8 lower bits form 24bit MQA content. The control stream is not affected by this stripping so it will show for example 88.2 with 16bit content. Nobody at Tidal or MQA cares ...

 

That Roon example is showing Authentication MQA 44.1kHz, which is the MQA control stream's actual original sample rate value - so that original sample rate isn't what Roon's MQA Core Decoder is actually upsampling to, ie, 88.2kHz!

[lucretius]

52 minutes ago, Cebolla said:

No idea where you got that from. Also, just look a few posts above, in the same test conversation, where mansr clearly mentions that bit 9 is "just above the MQA control stream":

 

Bit 8 makes more sense.  I got confused with what FrederickV wrote:  "Setting the 9th bit (starting counting from bit 0) via one of these 2 Perl operators, de-authenticates the file, and mytek says it's 24/44.1"

 

If we can get 15 bits of music out of 16 bit MQA, then why is 24 bit MQA limited to only 17 bits (instead of 19 bits) of music?

 

[UkPhil]

15 minutes ago, bogi said:

 

16bit MQA is probably 15bit PCM content + 1bit control stream.

24bit MQA is 16bit MQA and 8 bits in group 4+4: 4 bits for lossy encoding of high frequency content and another 4bits to control "unfolding" = upsampling, filtering.

So where does the MQA CD version sit as this version shows upsampling 8x from an alleged 16/44.1 file when played ? 

[lucretius]

1 hour ago, bogi said:

Here mansr really writes that bit 8 (counting from bit 0 - the least significant one) is used for control stream. That's what Cebolla says.
 

This is by design. The MQA identification and authentication data is embedded in bit 8 of a 24-bit PCM stream. Dropping bits 0-7 thus leaves it untouched. The authentication works by computing a Blake2s hash over the top 15 bits (the plain PCM portion) and parts of the control stream in bit 8. This hash is then verified against a cryptographic signature extracted from the control stream. The public key is stored in the decoder. If the signature matches, the blue light goes on. The low 8 bits encoding the high-frequency content are not covered by the authentication.

 

 

I get that bit 8 is used for the MQA identification and authentication data.  I wasn't sure about whether that same bit was also used for the filter selection and sampling rate.  The consensus here is that bit 8 it is used for that.  This leaves me wondering why we can get 15 bits of music out of 16 bit MQA but we cannot get 19 bits (24 -4 -1) of music out of 24 bit MQA (putting aside the fact that much of what is in the 24-48kHz range is noise).

 

 

[lucretius]

12 minutes ago, UkPhil said:

So where does the MQA CD version sit as this version shows upsampling 8x from an alleged 16/44.1 file when played ? 

 

Actually 352.8k is 4x upsampling (i.e. 4 x 88.2).  Apparently, the DAC just picks up the rate from the control stream.

[bogi]

15 minutes ago, Cebolla said:

 

That Roon example is showing Authentication MQA 44.1kHz, which is the MQA control stream's actual original sample rate value - so that original sample rate isn't what Roon's MQA Core Decoder is actually upsampling to, ie, 88.2kHz!

So then upsampling to 88.2/96 is possible also without those 4 bits which are used to choose a filter?
The picture shows 24bit result from 16bit source content - that's in principle not needed for 2x upsampling of 15bit content (since 2x upsampling adds 1bit of resolution), but more added bits can be used for dithering.

[bogi]

50 minutes ago, UkPhil said:

So where does the MQA CD version sit as this version shows upsampling 8x from an alleged 16/44.1 file when played ? 

Every delta sigma DAC is oversampling ... and every current one is capable of filtered 8x oversampling, usually with possibility to choose a filter. If an MQA DAC contains filters from MQA Corp. and teh control stream tells to use them to uspsample 8x ... I didn't say that it works so (and I really didn't read the 800+ pages here), but technically it could.
If so, then those 8x are a kind of propaganda ... every delta sigma DAC is oversampling type of DAC.

[KeenObserver]

Tell me something.  If you took a 12/32 recording from an old computer file and ran it through MQA "authentication", when it was unfolded and up sampled, would it light up the blue light and the 24/192 light?

[bogi]

See here
The display tells you what went into the MQA encoder, not what is coming out of the decoder.
So for example 352.8k master is degraded to 15bit PCM and control stream is added to the 16th bit - that's 16bit MQA. Effectively you have 15bit PCM, but display will show 352.8k. Master quality authenticated how the artist wanted. :D

[KeenObserver]

Perhaps Robert Harley and John Atkinson were right.  Perhaps we are at a paradigm shift.  Perhaps true Hi-Rez will no longer be available in the future.  Perhaps we will only be able to get MQA garbage.

 

We can thank The Absolute Sound and Stereophile for being in the forefront of promoting this new paradigm.