MQA is Vaporware

[mansr]

1 hour ago, Fokus said:

Moreover, in practice the patent text does not even have to be correct or complete.

It also doesn't have to be novel. The whole system really ought to be abolished.

[firedog]

21 hours ago, Lee Scoggins said:

 

But this isn't true in the case of Peter McGrath's files.  The mastering is identical except for the MQA treatment.  The treated files sound noticeably better and it's very easy to hear, in part because they are live acoustic performances.

And your sighted listening tests, which prove nothing.

Do the same tests blind without Peter McGrath in the room.

[John Dyson]

17 hours ago, Lee Scoggins said:

 

I wasn't going after Pro-Tools so much as the engineers who just record in 24/44 or 24/48.  Joe Palmaccio in Nashville and many others have been railing about this for over 10 years.  As someone who likes the best possible sound, I have been discussing this on the Hoffman board for a similarly long period of time.  My view is that 24/96 or better is ideal.

For processing --  for the best quality you really need more than 48k for the general case (where you don't know what kind of processing is being done.)  96k (or 88.2k) is the next higher sample rate that makes sense.  48k/44.1k sample rates are okay for delivery or simple applications, but lock you in to some problematical issues.  (Of course, you -- or the processing product -- can do up/down conversion, but that isn't really a good thing to do if can be avoided.)

 

Nowadays, with HW being so common, and plenty of disk space available, I don't see any reason at all for production not to use anything at 88.2k/96k or above @ 24 bits no-lossy compression involved.  Actually, 32bit floating point is better -- opens up some dynamic range mistake recovery during processing.  192k is nice, but mostly overkill for audio (real audible audio) applications, but other than overhead, really cannot hurt.  Also, the integral up/down conversion to 192k isn't too awful -- just best to avoid. There are a few minor additional degrees of freedom when using 192k sample rate (I am speaking of the DSP issues), but I would hope that most competent developers can avoid problems with processing at 96k.  48k is less easy/more trash can be left in the audio.  44.1k is simply ludicrious to try to do certain kinds of processing directly.  Also, there is a VERY VERY slight spectre of some Gibbs effect when up/down converting to/from 44.1k -- but NOT for linear applications.  Things like limiters/compressors have to do their duty carefully to avoid the sidebands wrapping around Nyquist rate.

 

The big problem with 48k (or 44.1k) is NOT problems for listening purposes, but one problem is there are some kinds of processing that produce sidebands in the audio -- those sidebands can easily (very very easily) create aliasing.  Linear processing -- no problems -- even doing up/down conversion (other than the well known issues.)  For nonlinear operatoins (like limiters/compressors/clippers/etc), best to avoid the lower sample rates.

 

[Lee Scoggins]

1 hour ago, firedog said:

And your sighted listening tests, which prove nothing.

Do the same tests blind without Peter McGrath in the room.

 

I've done it unsighted.  The difference is obvious.

[The Computer Audiophile]

1 hour ago, John Dyson said:

Nowadays, with HW being so common, and plenty of disk space available, I don't see any reason at all for production not to use anything at 88.2k/96k or above @ 24 bits no-lossy compression involved.  

The limitation is usually the total bandwidth required. 96 or 192 is great unless one needs a ton of tracks/channels. 

[fung0]

22 hours ago, Lee Scoggins said:

 

I wasn't going after Pro-Tools so much as the engineers who just record in 24/44 or 24/48.

 

An odd admission. As a claimed professional journalist, you might be expected to choose your words more carefully. Dissing on a product you admit is irrelevant to your argument seems like a pretty basic thing to avoid.

[John Dyson]

3 hours ago, The Computer Audiophile said:

The limitation is usually the total bandwidth required. 96 or 192 is great unless one needs a ton of tracks/channels. 

Maybe, in that case upconverison/downconversion might be appropriate.  However, at 192ksamples/sec * 4bytes/sample*16channels is 13Mbytes/sec.  If/when you REALLY need 16 or 24 channels and not just playing around, that isn't all that much.  A computer HDD is capable of 100Mbytes/second, and whenever processing those channels in HW, one would USUALLY do it in parallel, so the bandwidth per channel isn't all that big a deal.

I can acknowledge that 192k is probably excessive, especially for pop music,  so 96k might be a reasonable tradeoff.

 

My software is incredibly 'dense' with signal processing (it is difficult to describe the amount of processing -- but it is written absolutely the most efficiently as possible, and takes 3 cores of a Haswell CPU -- 3.3/3.4MHz because of the SIMD instructions throttling the CPU -- to run realtime.)   Very LITTLE procesisng in audio (unless maybe something neural net based) would need ANYWHERE near as much CPU.

 

For example:

(My software for resources for 2 channels each:  does over 200, different 300-1024 tap FIR filters, 64 different 1024 tap (super accurate) Hilbert transforms using DP math  & BLACKMAN 92 windows to maximize accuracy), and that only takes 3 Haswell cores realtime.   I doubt that many situations require this much processing in REALTIME @ 96k samples/sec for the entire chain!!!  Additionally, the timing is tracked perfectly, so that the input/output files or realtime output is accurate sample per sample!!!

(BTW, the code runs well over 98% of the time in SIMD instructions -- LOTS of math.)

 

So, it seems to me that generally working with 16channels at 96k/floating point samples is not beyond the capability of a workstation.  (Running my software 16 channels in realtime on a normal workstation might not be practical , however running at 48k and less quality only doubles the speed) Additionally, more than 4 cores is regularly available nowadays.  I am, of course, assuming  moderately efficiently written code.

I'd suspect that very few channels for very many applications need nearly as much processing as my software in most cases.

 

John

[botrytis]

5 hours ago, firedog said:

And your sighted listening tests, which prove nothing.

Do the same tests blind without Peter McGrath in the room.

I have heard the MQA MCGrath files and colour me not impressed. They were playing in the big Quintessence Audio room at AXPONA last year. I ran out of the room since they had so much high end noise, I was thinking there was problems with the system. McGrath liked MQA, and he said as much so their is your not so subtle bump to your sub-conscious to fill in the music.

 

I recently, went to a store in Des Moines, IA where I was listening to both MQA and regular FLAC. You could definitely hear the differences and many times is was not subtle. As an example, Melody Gardot's - Live in Europe can be had in both 48/24 FLAC and MQA. Listening to Track 1 in MQA, the drums (using brushes) sounded like noise, while the regular FLAC sounded like brushes on drums. This is only one of the things I noticed.

 

Lee - you can keep your MQA. I will take my regular FLAC files.

[tmtomh]

On 2/20/2019 at 7:37 PM, Lee Scoggins said:

 

A few corrections:

• As for DRM, there is not one single example of it being utilized.

 

Now as for hirez sampling rates having value, I find that pretty basic.  Anyone with decent experience doing recordings understands the sonic improvements in moving from 24/48 to 24/95 or 24/192.

 

HiGH-RES SAMPLE RATES: You need to re-read my comment (which you are replying to in what I've quoted above): When I said there's no evidence that anything more than 48k is necessary or useful, I clearly noted that I was referring to consumer end-product only, not to recording or mixing. 

 

So you need to acknowledge that you erred in responding as if I were including "doing recordings" in my point about sample rate.

 

EVIDENCE OF DRM: MQA is a DRM'd format. The fact that it has a currently unused ability to implement additional forms of DRM is irrelevant to the factual question of whether or not MQA is DRM'd. It is. An MQA file made from a 24/192 PCM original and purchased by a consumer, cannot be unfolded to 24/96 except during live streaming aka playback. As the purchaser and therefore owner of the file, I cannot play that unfolded 24/96 file on my non-MQA-capable equipment. I cannot load it into an audio editor and EQ it if I have an issue with the mastering. (And as I've said before, that's leaving aside the "final render" version that gets upsampled to 192.)

 

That's DRM, and you know it.

 

[crenca]

Update on Roon and MQA:  They admit that MQA should not trump Qobuz hi res (though I think they merely mean MQA's marketing speak resolution when equal to equivalent real hi res).  They called the current behavior a "bug" on their forums.

[Lee Scoggins]

4 hours ago, crenca said:

Update on Roon and MQA:  They admit that MQA should not trump Qobuz hi res (though I think they merely mean MQA's marketing speak resolution when equal to equivalent real hi res).  They called the current behavior a "bug" on their forums.

 

It's hard to do a fair comparison between MQA and Qobuz as there are some differences in how they do the streaming.  I find on some albums I prefer the Tidal MQA version and on others I prefer the Qobuz.  I play both via Roon 1.6.  It's hard to know what mastering differences exist on each file as well and what impact that is having on the final sound quality.

[Lee Scoggins]

5 hours ago, tmtomh said:

 

HiGH-RES SAMPLE RATES: You need to re-read my comment (which you are replying to in what I've quoted above): When I said there's no evidence that anything more than 48k is necessary or useful, I clearly noted that I was referring to consumer end-product only, not to recording or mixing. 

 

So you need to acknowledge that you erred in responding as if I were including "doing recordings" in my point about sample rate.

 

EVIDENCE OF DRM: MQA is a DRM'd format. The fact that it has a currently unused ability to implement additional forms of DRM is irrelevant to the factual question of whether or not MQA is DRM'd. It is. An MQA file made from a 24/192 PCM original and purchased by a consumer, cannot be unfolded to 24/96 except during live streaming aka playback. As the purchaser and therefore owner of the file, I cannot play that unfolded 24/96 file on my non-MQA-capable equipment. I cannot load it into an audio editor and EQ it if I have an issue with the mastering. (And as I've said before, that's leaving aside the "final render" version that gets upsampled to 192.)

 

That's DRM, and you know it.

 

 

1.  On consumer playback, 24/96 sounds way better than 24/48.  It doesn't matter whether it's recording or playback, the differences are real.  I cannot even believe we are arguing this.

 

2.  There is no DRM.  It's an authentication process.  They are trying to ensure a process from start to finish.  No one is forcing you to buy MQA gear.  And the unfolding can be done in inexpensive software.  And why do you need to mess with the mastering anyway?  If you don't like the release, don't buy it. 

 

These are incredibly weak arguments against MQA.

[mansr]

5 minutes ago, Lee Scoggins said:

It's hard to do a fair comparison between MQA and Qobuz as there are some differences in how they do the streaming.

Please do elaborate.

 

5 minutes ago, Lee Scoggins said:

It's hard to know what mastering differences exist on each file as well and what impact that is having on the final sound quality.

Pigs are airborne over a frozen Hell! Lee Scoggins speaks a truthful word!

[crenca]

1 minute ago, Lee Scoggins said:

 

It's hard to do a fair comparison between MQA and Qobuz as there are some differences in how they do the streaming.  I find on some albums I prefer the Tidal MQA version and on others I prefer the Qobuz.  I play both via Roon 1.6.  It's hard to know what mastering differences exist on each file as well and what impact that is having on the final sound quality.

 

It's not that hard.  Mastering differences are relatively easy to spot (instruments and voices move around, obvious level differences, whole new rhythm sections are suddenly heard on 50 year old rock, etc.) When the mastering is the same among an MQA album and the equivalent 16/44 or hi res, now that's hard.  MQA is a pretty transparent codec, with only a touch of digititus  heard on genre's/albums with enough HF content to reveal this.  On albums like Bob James "The New Cool", there is hardly enough so, it comes down to subjective bias.

 

Even on this (the real A/B of MQA vs PCM) you industry shills/writers get it wrong.  It's really amazing, you get it almost all wrong!  You supposedly hear amazing, obvious differences on albums with the same mastering.  Then again, you also hear amazing and obvious differences from a room treated with a handful of thimbles with "nano technology" in their tiny holes.  You guys are in a word, unbelievable.  You can't even do the "golden ear" thing right...

[Teresa]

On 2/21/2019 at 11:31 AM, mcgillroy said:

...why did Lee suddenly go after Pro-Tools?

 

 

On 2/21/2019 at 11:43 AM, The Computer Audiophile said:

That's been a popular product for audiophiles to bash over the years. 

 

Most audiophile recording companies don't use Pro-tools or other types of processing as they are the antithesis to natural recording techniques. I understand most posters here prefer major label recordings because that is where the music they like resides. 

 

Those audiophile labels don't like how Pro-tools changes the sound of their recordings, they believe using it and other such tools degrade the sonics. Music lovers who love the sound quality of audiophile recordings and loathe the sound of major label recordings tend to logically blame Pro-tools.

 

In my experience the most enjoyable classical, jazz and blues recordings are from audiophile labels, and boutique European classical labels.

[MikeyFresh]

1 hour ago, Lee Scoggins said:

These are incredibly weak arguments against MQA.

 

This from the guy whose pro-MQA stances are the poster child for weak "arguments".

[Don Hills]

2 hours ago, Lee Scoggins said:

1.  On consumer playback, 24/96 sounds way better than 24/48.  It doesn't matter whether it's recording or playback, the differences are real.  I cannot even believe we are arguing this.

 

I can believe you believe it, but it isn't true. If it were "way better," there would be overwhelming evidence in the form of passed blind listening tests. Instead, the tests show the difference is, at best, barely noticeable in some cases.  That's why we're still arguing this...

 

2 hours ago, Lee Scoggins said:

2.  There is no DRM.  It's an authentication process.  They are trying to ensure a process from start to finish.  No one is forcing you to buy MQA gear.  And the unfolding can be done in inexpensive software.  And why do you need to mess with the mastering anyway?  If you don't like the release, don't buy it.

 

There is DRM, just not currently implemented. As well as the actual code in the decoders, there's the smoking gun of paying for a license for strong encryption when a simpler, free implementation would have achieved the authentication goal.

As for messing with the mastering, that's exactly what MQA does. For any source not originally mastered with MQA (essentially, everything that exists), applying MQA processing changes the sound from that originally approved and signed off by the artists / producers / labels. You may argue that MQA corrects ADC shortcomings and brings you closer to the sound that was heard in the studio, but that sound is rarely what the aforementioned artists and producer want. It is tweaked and tweezed until it is the way they want it, alleged ADC warts and all.

[tmtomh]

3 hours ago, Lee Scoggins said:

 

1.  On consumer playback, 24/96 sounds way better than 24/48.  It doesn't matter whether it's recording or playback, the differences are real.  I cannot even believe we are arguing this.

 

2.  There is no DRM.  It's an authentication process.  They are trying to ensure a process from start to finish.  No one is forcing you to buy MQA gear.  And the unfolding can be done in inexpensive software.  And why do you need to mess with the mastering anyway?  If you don't like the release, don't buy it. 

 

These are incredibly weak arguments against MQA.

 

On consumer playback of various sample rates: As usual, you are a gigantic and complete un-self-aware hypocrite. You ask for peer-reviewed studies on claims others make, but are content with a bald-faced assertion like "24/96 sound way better than 24/48," which not only has zero peer-reviewed evidence to support it, but also is an assertion you could not back up yourself if subjected to a blind test. "I cannot believe wee are arguing this" - are you f***ing kidding? You couldn't tell 24/48 from 24/96 in a blind test if your life depended on it.

 

As for DRM, you are incorrect and either you don't understand MQA or else you are intentionally obfuscating how it works. Assuming as an example a 24/192 PCM original that gets turned into a 24/48 MQA, there is nothing about authentication that has to preclude the owner of such a file from using MQA software to do a first unfold to 24/96, with the ability to save or store the file in that format. Such a file can be checksummed or use any number of simple authenticating markers to ensure it was in fact unfolded by an actual MQA algorithm; and such a file could then subsequently be played through an MQA DAC, which would do the final (and meaningless) upsampling back to 192k, paired with a supposedly "DAC-specific" MQA filter applied during playback. That would still be an authenticated chain, and it still would require proprietary MQA hardware and software - but it would not be DRM because the purchaser - the OWNER - of the file would be able to see, store, copy, and play back on any equipment the first-unfold 24/96 file, which contains the full sample rate that's in that MQA file (since as you know the 192k version is just a doubled, upsampled creation - the MQA file only contains 96k)..

 

There is no technological or authentication reason that requires the file's owner to be stuck with the 24/48 folded version and unable to access the higher resolution except in playback/live streaming. The only reason the owner of the file can't directly access and make use of the first-unfold 96k file is DRM. And once again, you know that. You are being disingenuous in the extreme.

 

Finally, "why do you need to mess with the mastering anyway"? The reason is, because once I buy the file it's my business what I do with it and.you can kindly go pound sand. This is what I meant in an earlier comment when I said that people would take you more seriously if they thought you actually cared about their concerns about DRM, which you clearly don't.

[Ralf11]

Lee, you should get a job working for Putin

[FredericV]

6 hours ago, Lee Scoggins said:

2.  There is no DRM.  It's an authentication process.  They are trying to ensure a process from start to finish.  No one is forcing you to buy MQA gear.  And the unfolding can be done in inexpensive software.  And why do you need to mess with the mastering anyway?  If you don't like the release, don't buy it. 

 

These are incredibly weak arguments against MQA.

Your arguments are very weak, and very similar to those of the banned PV:
 

Another #6

Now convince me you are not being paid to be a key opinion maker.

[KeenObserver]

Watching Lee Scoggins continued promotion of the MQA agenda strikes me as being profoundly sad.  It makes me think of someone performing CPR on a close family member who is in rigor.

Lee,  you need to give it a rest.  The patient was fraught with cancer and disease.  It is a blessing to lay it to rest.

No one can fault you for not promoting it 100%.  But the patient is dead.  Let it rest in peace.

[Norton]

1 hour ago, KeenObserver said:

Watching Lee Scoggins continued promotion of the MQA agenda strikes me as being profoundly sad

 

Maybe he’s trying to bring it back from the Brink?

Or just sees it as a path towards identifying  “Sonic Hedonism For Everyone”?

[opus101]

4 hours ago, Ralf11 said:

Lee, you should get a job working for Putin

 

Way too cheap a jibe at Putin - he does answer tricky questions without deflection or obfuscation.

[mansr]

6 hours ago, Don Hills said:

There is DRM, just not currently implemented. As well as the actual code in the decoders, there's the smoking gun of paying for a license for strong encryption when a simpler, free implementation would have achieved the authentication goal.

The authentication is quite simple. All it does is compute a hash over the PCM data and compare it to an RSA-encrypted reference.

[John Dyson]

7 hours ago, tmtomh said:

 

On consumer playback of various sample rates: As usual, you are a gigantic and complete un-self-aware hypocrite. You ask for peer-reviewed studies on claims others make, but are content with a bald-faced assertion like "24/96 sound way better than 24/48," which not only has zero peer-reviewed evidence to support it, but also is an assertion you could not back up yourself if subjected to a blind test. "I cannot believe wee are arguing this" - are you f***ing kidding? You couldn't tell 24/48 from 24/96 in a blind test if your life depended on it.

 

 

I agree with you.  I do find this argument about 48K not being 'good enough' for consumer delivery to be distasteful  (44.1k might have a few troubles, but is also 'good enough' most of the time) .   The reason for my distaste  is that the matter of adequate sample rate is pretty much settled and has good basis in mathematics & science.  (I'd even argue that 16bits is good enough FOR DELIVERY also -- but that argument is weaker in my opinion -- because of the dithering and possibility after minor processing to dither again...  16bits is on the margins, but okay. )  (I was going to discuss the various reasons and when for more than 16bits, but that would only further confuse matters.)

 

It is an embarassment to humanity & intellect that someone would claim that greater than 48k is needed for delivery of audio for humans to listen to.  Any day -- I'd prefer 48k/24bit over 96k/16bit FOR LISTENING -- not because all of the 24bits is needed, but simply I like it better :-).  The 96k is specious for human listening.  (We all know that for other purposes, 96k can be useful or even necessary to maintain quality, just like 24bits or FP area also necessary or helpful in some cases.)

 

44.1k is a different matter...  From an emotional standpoint, I do not like it.  Perhaps the more important reason is the various kinds of tighter (more delays) filtering needed to make sure that there isn't aliasing, and it gives almost ZERO room for extra dynamic processing while maintaining the entire 20+kHz BW.  48kHz gives that extra 2-3kHz which really makes a difference for certain kinds of processing.  44.1k JUST BARELY FITS -- no extra room. 

 

John