MQA Users & Discussion Thread

[sir sanders zingmore]

2 hours ago, legend said:

 

Can't really compare because they are doing different, though parallel, things.

I'm interested in this. What is your hearing doing that is different from your measurements?

 

[Guest rmpfyf]

5 hours ago, legend said:

I think the most important difference between MQA and PCM should be in the time domain due to the 'temporal deblurring' and I doubt whether my microphones would be sensitive enough to determine a difference - even assuming I had test tracks in each that were suitable eg impulse responses.

 

Hang on Rod - MQA is PCM, let's not go confusing things. It's an interesting way of storing PCM, it allows for effective dynamic changes in filtering based on the entropy of discrete passages of content, but it is PCM. It gets to your DAC IC as PCM. There's nothing magical here and much of the ire at the technical/royalty implementation is here.

 

If you want to take the argument to the nth degree, if there's any audible difference between a DxD master and unfolded MQA it's not likely to favour MQA as it's a very smart lossy method... though it is ultimately lossy. 

 

'Temporal deblurring' is a BS term for 'picks better filtering when it needs it, bandwidth permitting'. (FWIW a lot of people trying to make sense of what are sometimes expensive purchases in audio effing hate BS terms.)

 

It has great advantages in streaming through compression. Winning product right there, and many people will love it just for that. 

 

Rights to 'unfold' content are being managed digitally. This is the definition of DRM. You can theoretically circumvent this (some have) though Meridian takes a dim view of as much and it's a legally ambiguous area for now. Yes, the intent is in part good - Meridian wants to ensure filters applied both ways are consistent, which is the only way to ensure the original waveform is recreated - but they want to charge for this, and have made business of stating that they need to work with the DAC manufacturer inherent to ensure this is so.... when there are honestly quite a few ways to ensure this could happen correctly that are lower cost and permit greater distribution - you're an engineer and you'd have to realise this much - they've chosen not to, and so Bob Stuart doesn't get to sit around wondering why some people aren't enthused by the notion of there being content in media they've just bought that they can't access without paying Meridian a second royalty. 

[bhobba]

1 hour ago, Sir Sanders Zingmore said:

I'm interested in this. What is your hearing doing that is different from your measurements?

Well here is one thing.  An electronics magazine published an amp design with distortion so low they had to buy a special instrument to measure it - even the output terminals affected the incredibly low distortion.   They said it sounded clearer than any amp they had ever heard.  Well I know some people that built it - as advertised it sounded clear as a bell - but the music was as if all the life was sucked out of it.  Why is that?  Well I don't think anyone knows for sure, but the conjecture is it takes a small amount of time for the feedback to actually lower the distortion - this creates a sort of time smear - that led to the life being sucked out.   These days most modern amp designers don't do large amounts of global feedback - rather use small amounts of local feedback at various stages.

 

Nobody is claiming there isn't a reason why measurements don't tell the whole story - if we measured a LOT more things we would probably get closer to what people hear and whats measured eg rise time is important but only sometimes measured.

 

But even aside from that we have preference issues.   Even order harmonic distortion sounds euphonic - small amounts are usually judged nice.  Whats the correct amount - that varies from person to person and even between material.   An amp designed like that deliberately is Hugh Deans NAKSA.  At an amp get-together on certain material like my favorite - Fever by Peggy Lee it was the best amp there - obtaining high marks from everyone - on other highly dynamic material like Private Investigations - not so hot.   I think it was rated something like 3rd or 4th overall.   

 

Thanks

Bill

[bhobba]

5 hours ago, legend said:

Re Bhobba's comment that he (and others) likes MQA even though it sounds thin, I like it for the opposite reason - to me it sounds slightly  'warmer' but in a natural more detailed way, as opposed to the usual euphonic distortion that reduces clarity.

Interesting isn't it.  It may be a bit warmer - by thinner I mean its more delicate than 44.1 Tidal which is less delicate and more ballsy to me - some seem to like that - I don't like Tidal 44.1 that much BTW - don't know why for sure - but prefer my own disk of rips to that.   But then again I play my ripped stuff through high quality software players like my new favorite - HQPlayer which could explain it.  I wont name who they were but I was with two other experienced audiophiles both of which thought I was mad.  BTW that's fully unfolded MQA - just the first unforld is virtually preferred by everyone I know - but over on Computer Audiophile there are people that even like 48/16 to 98/17 - which is what unfolding does - strange isn't it.  Full unfolding applies a filter that up-samples it to its original recorded rate which creates alias components as explained here:

http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html

 

See figure 5 where the lower frequency components have been reflected into the upper frequencies.  You cant hear that high but its presence seems to create effects that are audible - for the worse - or better - as the article explains.

 

Thanks

Bill

Edited December 2, 2017 by bhobba

[sir sanders zingmore]

Thanks @bhobba for your very detailed explanation. No disrespect intended but my question was actually for @legend

[bhobba]

On 12/2/2017 at 7:24 PM, rmpfyf said:

'Temporal deblurring' is a BS term for 'picks better filtering when it needs it, bandwidth permitting'. (FWIW a lot of people trying to make sense of what are sometimes expensive purchases in audio effing hate BS terms.)

 

With the current state of the art its utter BS.   Here is why.  If you place a linear phase brick-wall filter above the highest frequency of music what does it do?   A big fat zero.  Even their published diagrams of typical recordings show it rarely goes above 48k before its basically all noise.   That being the case normal 96/24 does nothing as far as time smear goes and MQA is useless.  I have looked at a lot of currently available HD material - mostly its below 30k - very few material is much above 50k.   But that's precisely the material we need to judge MQA claims.  The 2l recordings are basically a crock:

http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html

 

The maximum that material goes to is 60k - mostly it's lower than that.   We need material that goes to 176K to judge it.   I have one recording like that - is it available in MQA - no.  Until such material is we cant really judge their claims sound-wise.

 

Thanks

Bill

 

 

Edited December 4, 2017 by bhobba

[Guest rmpfyf]

27 minutes ago, bhobba said:

 

With the current state of the art its utter BS.   Here is why.  If you place a linear phase brick-wall filter above the highest frequency of music what does it do?   A big fat zero.  Even their published diagrams of typical recordings show it rarely goes above 48k before its basically all noise.   That being the case normal 96/24 does nothing as far as time sear goes and MQA is useless.  I have looked at a lot of currently available HD material - mostly its below 30k - very few material is much above 50k.   But that's precisely the material we need to judge MQA claims.  The 2l recordings are basically a crock:

http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html

 

The maximum that material goes to is 60k - mostly it's lower than that.   We need material that goes to 176K to judge it.   I have one recording like that - is it available in MQA - no.  Until such material is we cant really judge their claims sound-wise.

 

Thanks

Bill

 

Bill, I think we're all quick to confuse two things here:

 

1. Does MQA capture content we're missing, whether this is able to be characterised by spectral methods or not, and
2. Does MQA provide a means to ensure waveforms as recorded are recreated faithfully through a process whereby all filters from mastering through reproduction at the DAC IC are known and controlled

 

All this talk of 'warm', 'thin' etc is a bit misplaced. I'm not suggesting such talk is audible, euphonic, pleasant, annoying or what have you, more that there are many things outside of the signal chain MQA can control that affect these characteristics, let's keep this to what MQA says it can do. 

 

On the first point I'm not convinced that 'rarely goes above 48k before its basically all noise' is the end of the story. Put simply - that's not how spectral methods work. If you have a waveform that's a perfect sine at 48k and a sampling process that captures that frequency as a centre frequency perfectly, then anything else you see spectrally is legitimately all noise. This can be through windowing or other mathematical phenomena prescient to converting between time and frequency domains. If you don't - and you very much almost always don't - that 'basically all noise' are spectral components used to recreate a waveform that might be of a centre frequency that you can psychoacoustically hear, but simply isn't a perfect waveform. The amplitude of these components is very small and not easily discerned from noise, particularly at higher frequencies. I get what you're suggesting but it's not a correct read of spectral content with respect to audible content, and to be fair we get a lot of published crap attempting to supporting the case for higher frequencies that simply doesn't understand sampling theory (Hans on Youtube, some American academic not understanding phase delay potential in stereo Redbook, etc). 

 

IMHO we sometimes confuse the two because a prevalent early commercial work in digital - Redbook as we know and love it - has cutoffs and filters based on pure psychoacoustics, and it was the 80's - so unless you had (1) a LP collection of Jarre, cymbals or  the dulcet sounds of paper being scrunched with (2) a vinyl rig that was out of this world and (3) epic recall of what these sounded like live, Redbook kicked a** over the status quo on so many fronts that we didn't really have much time or resource to realise that pure psycoacoustics weren't the actual limit of spectral content we'd care about. And if we had, the world was flat out putting ~650MB on an optical medium for the first time - there wasn't much room or time to do much else.

 

This is an important point because just to appreciate the sensitivity of the spectral content at hand is a significant body of work, and here Meridian deserves praise. Though TBH even now you'd need a rig that's not sufficiently filtered elsewhere to make use of it. 

 

The second point is one of provenance. 'We guarantee that it leaves the DAC as close to the studio master as feasible for the bandwidth and content inherent'. This is important as DAC IC manufacturers and DAC designers are traditionally on one hand - some offering users the choice to change filter properties (hi Clay and others) - and studios and mastering people are way over on the other, and rarely do the two talk about ultimate fidelity to the master. If anything our paradigm has been traditionally quite different - 'I'll do the best I can given the limitations of the playback format I'm mastering to, best of luck'. 

 

The s**t with MQA here is that it'd be very easy to make an Amanero-type board that had some postprocessing on it (hell you could even just implement it as an audio driver), or at least open that much up, so that the I2S content headed to your DAC IC was at the maximum content possible. Yes, you could still pay for a certified DAC, though anyone having bought MQA content wouldn't feel stiffed either. 

 

BTW there is nothing BS about a brickwall filter if it's implemented properly. If. It's just a filter, as are the rest of them just filters. Whether simple or sophisticated they're just tools, and each about as good as they're used. 

 

My 2c.

[legend]

8 hours ago, Sir Sanders Zingmore said:

Thanks @bhobba for your very detailed explanation. No disrespect intended but my question was actually for @legend

 

The diaphragms of microphones and ears are both obviously reacting to impinging sound waves in much the same way to create electrical signals but how these are then processed and interpreted are obviously different – and this can lead to different conclusions as Bhobba rightly says.

 

Which is ‘more sensitive’ is therefore a difficult question to answer.  It depends on the situation one is investigating – and in the case of MQA I really don’t know.

[legend]

I am not an electronics engineer (just a physicist) and have only a broad understanding of the theoretical underpinnings of MQA (having read the patents as an ex-patent examiner) but I think part of its problem is it does a number of things at once and it is easy to attack one part without appreciating it as a whole – and this is whole is necessary for its technical and more importantly commercial success (as the article I linked earlier says).  Also a number (but not all) of the attacks seem to be from those with skin in the game (those with vested interests) and these include both producers and consumers of music, neither of whom obviously likes pay more than they have to for anything.  But unfortunately very few things in life are actually free – someone somewhere carries the cost.  Finally marketing BS is regrettably sometimes (often?) necessary for success in the real commercial world – as a scientist I am not very good at it but sometimes grudgingly admire those who are!

[Guest rmpfyf]

30 minutes ago, legend said:

I am not an electronics engineer (just a physicist) and have only a broad understanding of the theoretical underpinnings of MQA (having read the patents as an ex-patent examiner) but I think part of its problem is it does a number of things at once and it is easy to attack one part without appreciating it as a whole – and this is whole is necessary for its technical and more importantly commercial success (as the article I linked earlier says).  Also a number (but not all) of the attacks seem to be from those with skin in the game (those with vested interests) and these include both producers and consumers of music, neither of whom obviously likes pay more than they have to for anything.  But unfortunately very few things in life are actually free – someone somewhere carries the cost.  Finally marketing BS is regrettably sometimes (often?) necessary for success in the real commercial world – as a scientist I am not very good at it but sometimes grudgingly admire those who are!

 

I'd go slightly differently. 

 

That MQA does a number of things at one is not a critique, it's an advantage. Some of the things it does are quite novel, and beyond as much a good thing.

 

I don't think anyone is having a fundamentally significant gripe for new mastering tools either, that's happened throughout many times. And TBH it's a stream compressor - nothing exorbitant there, it's not as though Meridian's asking a studio to move to DSD mastering or similar. 

 

At a deeply fundamental level MQA is the only audio distribution format that requires you to pay a second license to play content you already own. This is, IMHO, an extremely poor business decision. We understand how Meridian is trying to milk the cow and it's a very short-sighted approach (not least as their volume revenues on distributed material likely beat DACs shipped by a fair few orders of magnitude). And not least as their core IP isn't so novel that another organisation can't come up with something similar, stuff that into a FLAC container, open source or public license it and then we all move on with life. And sure, some systems and some ears aren't going to hear much difference between fully-unfolded content on a MQA-certified DAC with a little blue light saying 'Bob Stuart approves these waveforms' or not, but you'd at least have that choice having bought the material that contains the content.

 

When I buy 2L audio at 192kHz it doesn't play back at 96kHz until I send them more money for a new DAC I didn't want. 

 

Similarly marketing BS isn't successful. However smart he may be for many Bob Stuart comes across as a smug git. It's not hard to get up and say 'sometimes hires just sounds better, we've found a way of storing hires in not-so-hires files'. And if you want more detail than that - here's something hard for MQA freaks to stomach - signal processing, digital filtering and the like are taught to millions of people from all walks of life every year. There is nothing new here. There is zero need to invent a new language in communicating these concept, it does more for ambiguity than it does progress. (Whilst hating to admit it) I have a degree that covers signal processing in great detail, I use it regularly and when I first heard the term 'temporal deblurring' I didn't get it the first few times... it's really poorly communicated.

 

At the very least Meridian needs better marketing, and a business plan that reflects as much.

Edited December 2, 2017 by rmpfyf

[legend]

I only pay once because mainly stream MQA music on Tidal Masters (which does some decoding) and the occasional freebie MQA file from 2L and elsewhere.  In both cases I use a non-MQA DAC (DEQX or Redgum) and still think they sound better than the non-MQA files, even higher resolution ones.  Presumably if I were using a full MQA encoding/decoding system they would be better again!

 

To summarise one of your problems is you pay once for the MQA-encoded file and again for the new and MQA-licensed DAC?  If you were to pay only once which would you prefer? Or not pay at all so Bob & others are not recompensed for their intellectual effort?

 

As an ex-patent examiner I would probably disagree that the MQA system is not novel or inventive.  Even if the many of the parts were already known (sometimes from Stuart et als previous work) I think the sum total of them is non-obvious and so inventive.  And though I am not sure (because I am not an electronic engineer or signal processing trained) it seems to me that the breaking of the Shannon-Nyquist rule together with very short filters for minimum time effects may also be inventive by itself??

 

Re Bob’s smugness, I have not seen him since I met him while working for Linn in the late 1980s when he seemed a genuinely nice guy (and incredibly smart).  Perhaps he has changed – learnt that ‘nice guys come last’ as they used to tell us during my MBA at the (greed-is-good) AGSM where I also did some marketing subjects!

Edited December 3, 2017 by legend
clarification

[Guest Eggcup The Daft]

19 hours ago, bhobba said:

 

See:

https://www.computeraudiophile.com/ca/ca-academy/A-Comprehensive-Q-A-With-MQA-s-Bob-Stuart/

 

Have a look at some of the figures that show extensive measurements eg the noise floor of decoded MQA is below the thermal limit except for about the first 1k, so is not an issue - unless of course you think you can hear below the thermal limit of electrical devices?   Actually I have read that, strangely, some people can - but they would seem to be rare - and certainly I don't think if you are below that it's really a worry.  How to they do it with just 17 bits and not 24 - they are tight lipped - but use tricks that usually are not used like subtractive dither that is much better than normal dither.

 

Regarding impulse response the claim is made about a Dirac pulse .  Dirac pulses are a mathematical abstraction and do not exist - the plots they give are theoretical calculations based on what they do.   Is that a valid way to do such things - decide for yourself - but really its all that can be done - material containing Dirac pulses do not exist.

 

Look, the issue with MQA is the same as all such things in audio - some go gaga over it - some go - blah.   Its very individual.  I think it sounds a bit thin and too clean personally - but still rather like it - others have a different view.

 

So simply go listen for yourself.   An explorer 2 DAC is dirt cheap (a little under $350.00):

http://www.noisymotel.com/product.asp?ProductID=866

 

If you like MQA - do a post - if you don't - do a post and we will soon get to the bottom of it.

 

I have said what I think - ie a little thin and a bit too clean - but still like it.  I have sat with others that don't.  Only by finding the consensus of a lot of people can a real appraisal be made.   Others have also made that comment - a little thin and too clean:

https://www.computeraudiophile.com/forums/topic/33766-upsampling-mqa-files-to-original-resolution-with-sox-will-sound-like-the-original-resolution/

'Full MQA decoding is no longer master quality but a polished version. Brian Lucey has already stated that MQA sounds thinner than the original, and with more distortion.
https://www.gearslutz.com/board/12548751-post460.html
https://www.gearslutz.com/board/12551196-post482.html

This is logical: the post-ringing energy in the tail that MQA tries to get rid of, needs to be distributed to some other place in the waveform, causing distortion such as aliasing and HF noise. MQA created a solution for a non-existing problem.'
 

Me, the above article, and a few of my friends is not really much to go on.

 

BTW I do not agree with a lot pf the things Lucy says - but wont go into it in detail.  Suffice to say -  he/she is both right and wrong.

 

Thanks

Bill

Sorry. I see two graphs there that would seem to be measurements of the noise floor, but the rest is all claim and statement. To be clear, I searched that document  for "distortion" (appears once, in a question) and "THD" (no appearance). I can find no measurements of distortion with reference to playback of MQA files anywhere. Similarly, there are still no clear instances of a controlled DBT with guaranteed identical source files through the supposed end-to-end process to indicate that differences are present.

 

Distortion figures would give the game away regarding what the various cuts and restorations are leaving us with. Is distortion from MQA equivalent to 14 bit, 16 bit, 17 bit, or 24 bit? Because dithering is used, the noise floor can be very low, but distortion may only be changed in nature.

 

I am using Tidal  with a Dragonfly Red and Audeze EL-8 headphones to listen to Tidal Masters and hifi streams. To be honest, the experience has left me more confused than thinking that MQA is good or bad in any particular way. I hear bigger differences between alternate 16 bit files than I normally get between hifi and Master files and I suspect different masters and other conversions are getting in the way of hearing the effect of MQA itself. But I get the feeling that MQA is doing just as I suggested - changing the nature of low level distortion with a lower resolution than true 24 bit.

[Guest rmpfyf]

3 minutes ago, legend said:

I only pay once because mainly stream MQA music on Tidal Masters (which does some decoding) and the occasional freebie MQA file from 2L and elsewhere.  In both cases I use a non-MQA DAC (DEQX or Redgum) and still think they sound better than the non-MQA files, even higher resolution ones.  Presumably if I were using a full MQA encoding/decoding system they would be better again!

 

To summarise one of your problems is you pay once for the MQA-encoded file and again for the new and MQA-licensed DAC?  If you were to pay only once which would you prefer? Or not pay at all so Bob & others are not recompensed for their intellectual effort?

 

As an ex-patent examiner I would probably disagree that the MQA system is not novel or inventive.  Even if the many of the parts were already known (sometimes from Stuart et als previous work) I think the sum total of them is non-obvious and so inventive.  And though I am not sure (because I am not an electronic engineer or signal processing trained) it seems to me that the breaking of the Shannon-Nyquist rule together with very short filters for minimum time effects may also be inventive by itself??

 

Re Bob’s smugness, I have not seen him since I met him while working for Linn in the late 1980s when he seemed a genuinely nice guy (and incredibly smart).  Perhaps he has changed – learnt that ‘nice guys come last’ as they used to tell us during my MBA at the (greed-is-good) AGSM where I also did some marketing subjects!

 

I'd go easy on the pronouns - it's far from just 'my' problem. An easy solution would be to provide means to unfold fully on any device per paid content. Meridian gets their royalty, customers get to use their purchased content. And you could still sell MQA-certified DACs this way. 

 

I don't think anyone wishes for Meridian not to be recompensed for intellectual effort (I certainly don't wish this). 

 

As I'd written - some of the things MQA does are indeed novel. The underlying theories aren't new, their integration is novel.

 

There is no breaking of Shannon-Nyquist.

 

Can't vouch for Bob personally, just his presentation. Being familiar with eng/sci and marketing/corporate bits of professional life, I don't suscribe to nice guys having to finish last, more that well-marketed efforts do well, and MQA could be improved here. 

 

Not least as right now there's space in market for something that does the same thing insofar as a codec with less fuss over licensing. It's a gap that could easily be closed with a little forethought at no loss of revenue or profit - maybe even with gains.

[legend]

Apologies rmpfyf just scientist short hand - I meant the royal you (plural).

 

Ivor Tiefenbrun of Linn provides a good example of non-nice guys not coming last!  And marketing BS paying (hierarchy of components etc).

 

If there is a gap in the market that can be easily closed then why not do it?

Edited December 3, 2017 by legend
typo

[Guest Eggcup The Daft]

12 hours ago, bhobba said:

 

With the current state of the art its utter BS.   Here is why.  If you place a linear phase brick-wall filter above the highest frequency of music what does it do?   A big fat zero.  Even their published diagrams of typical recordings show it rarely goes above 48k before its basically all noise.   That being the case normal 96/24 does nothing as far as time sear goes and MQA is useless.  I have looked at a lot of currently available HD material - mostly its below 30k - very few material is much above 50k.   But that's precisely the material we need to judge MQA claims.  The 2l recordings are basically a crock:

http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html

 

The maximum that material goes to is 60k - mostly it's lower than that.   We need material that goes to 176K to judge it.   I have one recording like that - is it available in MQA - no.  Until such material is we cant really judge their claims sound-wise.

 

Thanks

Bill

 

 

That article makes the same mistake that others have - it confuses the details in the patent with the final MQA product.

[bhobba]

2 hours ago, legend said:

I am not an electronics engineer (just a physicist) and have only a broad understanding of the theoretical underpinnings of MQA (having read the patents as an ex-patent examiner) but I think part of its problem is it does a number of things at once and it is easy to attack one part without appreciating it as a whole – and this is whole is necessary for its technical and more importantly commercial success (as the article I linked earlier says).  Also a number (but not all) of the attacks seem to be from those with skin in the game (those with vested interests) and these include both producers and consumers of music, neither of whom obviously likes pay more than they have to for anything.  But unfortunately very few things in life are actually free – someone somewhere carries the cost.  Finally marketing BS is regrettably sometimes (often?) necessary for success in the real commercial world – as a scientist I am not very good at it but sometimes grudgingly admire those who are!

 

Bingo.

 

I am not a recording engineer either, just an applied mathematician.

 

I too have read the patents and other technical stuff.  Yes the claims they make are true - the issue is if it sounds better.

 

Only time will tell on that one.

 

Also I beieve only when better recordings are available will that be apparent - that will take time.

 

I am just watching the following right now:

 

Interesting

 

Thanks

Bill

[bhobba]

27 minutes ago, Eggcup The Daft said:

Distortion figures would give the game away regarding what the various cuts and restorations are leaving us with. Is distortion from MQA equivalent to 14 bit, 16 bit, 17 bit, or 24 bit? Because dithering is used, the noise floor can be very low, but distortion may only be changed in nature.

 

Why exactly do you think it introduces 'distortions' in the usual sense?

 

The issue is aliasing, which I suppose can be described as distortion, but is different to THD.

 

It's a trade-off between time smear distortion and alias distortion - conventional measurements are not really relevant here - at least in that area.

 

The issue people usually get concerned about is noise - that's why they gave a lot of graphs on that.   If you get effective 20 bits either by it being 20 bits or via dither then its below the thermal limit.   They use 17bits and subtractive dither (which cant be explained easily - but you can look it up) to get effectively at least 20 bits from 17 - but in fact a lot better - closer to actual 24.  BTW quantisation noise is distortion as well - it just shows distortion is not an easy issue to discuss.

 

Thanks

Bill

Edited December 3, 2017 by bhobba

[sir sanders zingmore]

Meridian have been very successful at two things I think:

 

- getting people to call the decompression of a lossy codec "unfolding"

 

- inventing a new problem (time smear) and coming up with a solution 

[Guest Eggcup The Daft]

1 hour ago, bhobba said:

 

Why exactly do you think it introduces 'distortions' in the usual sense?

 

The issue is aliasing, which I suppose can be described as distortion, but is different to THD.

 

It's a trade-off between time smear distortion and alias distortion - conventional measurements are not really relevant here - at least in that area.

 

The issue people usually get concerned about is noise - that's why they gave a lot of graphs on that.   If you get effective 20 bits either by it being 20 bits or via dither then its below the thermal limit.   They use 17bits and subtractive dither (which cant be explained easily - but you can look it up) to get effectively at least 20 bits from 17 - but in fact a lot better - closer to actual 24.  BTW quantisation noise is distortion as well - it just shows distortion is not an easy issue to discuss.

 

Thanks

Bill

Ah, but I am concerned here with boring old THD. Practical, traditional playback of 16 bit has effectively 0.2% distortion at best at -60dB (I believe the lowest measured is now 0.192%). 24 bit playback is close to zero across the full range. So THD is a practical issue, it sits just on the threshold of audibility for 14-16 bit audio while, for example, jitter and noise are well below audibility and to all intents and purposes irrelevant with modern technology. (Note - using the Ethan Winer files referred to in an earlier thread and some others, I claim that we can hear -50dB or a bit more in some circumstances and distortion is present there with 16 bit files: -60dB may be, well, pushing it).

 

And you can't just upsample 16 bit to 24 bit audio and get around this distortion issue like you can with noise, for example.

 

This is relevant to MQA because the effective real world resolution of MQA is that of the equivalent level of THD in bit depth terms. MQA doesn't itself produce this THD figure. It's beyond MQA's role, in the final conversion to sound.

 

 

 

[Ittaku]

I'm late on the scene with regards to any discussion regarding MQA but I still have trouble coming to grips with any concern regarding expansion of file size by increasing bitrates and sample frequencies as a blunderbuss approach to fixing any underlying limitations of the medium. It may be a braindead approach to fixing the problem, but it also means you don't need smart engineers to do the encoding side of the audio transfer in some magical way to get it right. One of the godsends for me having moved from redbook CD to highres digital downloads provided by virtually all labels (at least for classical music) is that even the big labels that traditionally made some pretty meh sounding CDs are now very decent simply by virtue of the fact they are destroying less in the process with more data end to end. Yes of course anyone can destroy things if they try hard enough but really bad sounding recordings are now extremely rare - mostly (in classical at least) it's more a matter of too distant mic'ing, unbalanced mixing between instrument groups, bad capturing of hall acoustics and so on, but there is virtually none of that 1980s digital sound today.

 

Now irrespective of that, what fascinates me in Bob's interview are the claims that even 768/24 barely approaches what MQA can achieve. I'm really struggling with this concept on so many levels because a lot of the issues that he claims MQA fixes, it's hard to believe are actually issues. I'd like to believe we may eventually be able to define what it is that actually does sound better to us humans as measurements have failed us to date to correlate with what sounds good to the human ear-brain combination, but I see another measurement fix here that I cannot put much weight on in isolation. Bob's graphs mostly look of marketing hype rather than meaningful measurements anyway - especially the "sound quality versus sample rate" one. I haven't heard any MQA decoder myself so I'm only going on internet hearsay, but the blind testing internet experiment was telling.

 

At the very least, to me this MQA push reminds me of what happened when HDCD came out (and I had a HDCD DAC in the 90s), except that perhaps the world is riper for a higher resolution lower bitrate alternative... but will that matter long term? The difference in redbook cd with and without HDCD, decoder on or off was dramatic, but basically it was another form of compression to get more out of 44.1/16. Am I really missing something that MQA fixes that cannot be fixed by bitrate? I can't hear the difference between anything above 88.2/24 no matter how hard I try or how much money is spent on a system already.

Edited December 3, 2017 by Ittaku

[Guest Eggcup The Daft]

43 minutes ago, Ittaku said:

I'd like to believe we may eventually be able to define what it is that actually does sound better to us humans as measurements have failed us to date to correlate with what sounds good to the human ear-brain combination

We do know that when tested blind people prefer the sound that is most accurate according to boring, old fashioned, traditional measurements (from the work of Toole and others). I had to have this repeated to me again and again before I got it, but it's true.

 

Measurements do correlate, unless we let things other than the soundwaves get in the way.

 

Boring old fashioned measurements and tests would tell us a lot about what MQA is actually doing and whether it's worth pursuing.

 

55 minutes ago, Ittaku said:

 Am I really missing something that MQA fixes that cannot be fixed by bitrate? I can't hear the difference between anything above 88.2 no matter how hard I try or how much money is spent on a system already.

There are things that an MQA type system might fix, and most apply to existing digital files. The algorithms used can/do improve temporal resolution, and the filters to fix ADC problems with older digital recordings may well improve the sound (of course, the early digital release is the master in this case and was released sounding like that, leaving a philosophical question if you are interested in fidelity). For new recordings, bitrate may "fix" a lot, at least for recording and editing. A high enough bit rate should be equivalent to MQA at a lower bitrate for any of the claims that are made for it, and at some high enough rate even the finest claims for human temporal resolution would be met.

 

One of the problems with hi-res is the sheer amount of wasted data, which is where any algorithm that discards more unwanted noise data can be useful, especially in streaming.

 

On balance, we can assume for the moment that 44.1/16 does contain all the data we theoretically need to play back the performance with maximum fidelity to the human ear - that is what the science still says, until it is disproved. The questions then are about the performance of recording/editing/playback technology and whether we need higher data rates in the real world. I suspect the answer is yes, for now - but 88.2/24 probably ticks all the boxes for playback, at least. Your last comment shouldn't surprise anyone.

 

Let's see if any new evidence comes to light.

[Guest rmpfyf]

7 hours ago, bhobba said:

It's a trade-off between time smear distortion and alias distortion

 

Huh? Aliasing isn't distortion, it's a f****p caused by not sampling at sufficiently high frequencies. Conventional measurements apply, and have since the earliest days of signal processing. Spectral leakage is not aliasing, it's a simple function of digital Fourier methods being discrete transforms, and the waveforms we seek to characterise in audio almost never being perfect sinusoids. There's no tradeoff here. Got Nyquist? You're good. Way above it? Better. 

 

Time smear distortion is a BS term. It's just group delay - can we call it that? The term that's been used for decades, what Meridian actually references in their patents?

 

Filters have compromises. MQA reckons full-rate they're able to hit a time-domain accuracy of 3-4us if their techniques are used in original recording through mastering, i.e. a full MQA workflow. Most of this concerns better-than-brickwall filters that have come into prominence since Rebook. Again, nothing new here, less ringing, more temporal accuracy etc.

 

Not to flog a dead horse but that last part still doesn't necessitate the licensing model they've employed.

 

These two phenomena don't trade off (you can certainly have low aliasing and low group delay), and FWIW it's possible to get the same quality is other formats we'd usually associate with original recordings. Mind these effects are accentuated at higher frequencies, so not all music is going to benefit - if your Redbook rig sounds pretty damn live as is (some do), MQA or hires isn't going to do much for you. 

 

4 hours ago, Ittaku said:

Now irrespective of that, what fascinates me in Bob's interview are the claims that even 768/24 barely approaches what MQA can achieve. I'm really struggling with this concept on so many levels because a lot of the issues that he claims MQA fixes, it's hard to believe are actually issues. I'd like to believe we may eventually be able to define what it is that actually does sound better to us humans as measurements have failed us to date to correlate with what sounds good to the human ear-brain combination, but I see another measurement fix here that I cannot put much weight on in isolation. Bob's graphs mostly look of marketing hype rather than meaningful measurements anyway - especially the "sound quality versus sample rate" one. I haven't heard any MQA decoder myself so I'm only going on internet hearsay, but the blind testing internet experiment was telling.

 

Bob is being disingenuous, even though the claim is technically correct insofar as time domain accuracy for samples captured with poor filtering. Another way of suggesting what's possible here is that MQA 768/24 allows for some content within the stream to be rendered with filters employed at higher frequencies. Which is great, though to be fair you'd need your 768/24 content to be pretty bloody marginal in order to have a whinge, either that or it's a reference-grade comparison of small cymbals having the s**t kicked out of them (this is a small market segment - most metal drummers don't listen to cymbals on an audiophile rig ).

 

Bob would do better to suggest (not to be his PR department):

• Better filtering techniques (i.e. apodizing filters) can be employed to either reduce group delay compared to brickwall filters for a given sampling frequency, or can be employed (as a convolution) to correct for filter effects in already-mastered material, 
• These filters need to be employed carefully, or group delay can actually be increased (multiples of them are bad) - but done right (probably once only, so probably at mastering) they make for a more accurate waveform than anything mastered with a brickwall filter
• Employing such filters typically requires sampling headroom above the original source material, 
• Double sample frequency is typically impractical unless some means can be found to compress the required additional content to effect the above filtering technique into existing or comparable bandwidth,

AND FINALLY

• MQA is a means of realising all the above.

 

A word of caution here - different filters sound different simply because they manipulate content in different ways. Some people I'd imagine are going to find upsampled audio preferable to MQA as it allows a tighter time-accurate take on filtering paradigms they already know. Apodizing sounds different, you're changing the distribution of spectral energy around a given waveform in the time domain. You might like it, you might not. 

 

No doubts MQA can result in something more accurate though.  

 

2 hours ago, Eggcup The Daft said:

On balance, we can assume for the moment that 44.1/16 does contain all the data we theoretically need to play back the performance with maximum fidelity to the human ear - that is what the science still says, until it is disproved.

 

Not exactly - we can assume Redbook covers the entire range of psychoacoustic frequency sensitivity as we know it. 

 

We can similarly embrace, scientifically proven, that we can hear group delay effects beyond what's possible with Redbook, and - whether or not such content exists in what music we listen to (it varies) - there are formats that cover this too.

Edited December 3, 2017 by rmpfyf

[Ittaku]

1 hour ago, rmpfyf said:

A word of caution here - different filters sound different simply because they manipulate content in different ways. Some people I'd imagine are going to find upsampled audio preferable to MQA as it allows a tighter time-accurate take on filtering paradigms they already know. Apodizing sounds different, you're changing the distribution of spectral energy around a given waveform in the time domain. You might like it, you might not. 

 

No doubts MQA can result in something more accurate though. 

Well in that case, according to:

3 hours ago, Eggcup The Daft said:

We do know that when tested blind people prefer the sound that is most accurate according to boring, old fashioned, traditional measurements (from the work of Toole and others). I had to have this repeated to me again and again before I got it, but it's true.

Then if it's more accurate, blinded people will prefer the sound.

 

But...

1 hour ago, rmpfyf said:

Bob is being disingenuous, even though the claim is technically correct insofar as time domain accuracy for samples captured with poor filtering. Another way of suggesting what's possible here is that MQA 768/24 allows for some content within the stream to be rendered with filters employed at higher frequencies. Which is great, though to be fair you'd need your 768/24 content to be pretty bloody marginal in order to have a whinge, either that or it's a reference-grade comparison of small cymbals having the s**t kicked out of them (this is a small market segment - most metal drummers don't listen to cymbals on an audiophile rig ).

Correct me if I'm wrong for I may have mistaken you? Are you saying that 768/24 (cymbal clashing) content with MQA would be audibly different to simple 768/24 content when anything above 96/24 is already indistinguishable? With modern recording techniques and high bitrates I cannot accept that. Restoring old content maybe MQA has something to offer.

[Guest rmpfyf]

@Ittaku I don't know about preferring accuracy over any form of distortion; I have a DAC tends on the warm side and it's just lovely. I'd love others for comparisons though realistically very few of us (certainly not me) are going to be in a position practically to afford as much, so audiophilia becomes a game of reliance if you will; of understanding that certain components will perform predictably and then then bending that chain to whatever presentation you like.

 

Certainly one implementation of MQA goes a ways towards ensuring end waveforms are as masters intent; whilst MQA can be used to correct old recordings of known filtering provenance it should be noted, for various reasons, that such corrections are simply not absolute, and reflect neither an end-to-end process or the intended mastering work of a sound engineer. 

 

As to what's indistinguishable above 96/24, again we're confusing the following:

 

• Psychoacoustics (what frequencies you can hear)
• Frequency domain discretisation effects (anything that's not a perfect sine at a sampled centre frequency will have frequency components well above their centre frequency when viewed as spectra). This more broadly feeds into pulse-code modulation limitations for abrupt waveforms: unlike pulse-density modulation formate e.g. DSD, a square wave in PCM in any discretised frequency domain has nearly infinite components in spectra, even if they're very low energy they're not any less important).
• Group delay effects (time delay in various amplitude envelopes as a function of frequency)
• Ringing artefacts (inherent to the brickwall/sinc filter)

 

All we get from looking at spectra is a moderate-fidelity assessment of psychoacoustic content. None of the rest. And to be fair whilst we know most about psychoacoustics, human audible limits for the rest are areas of ongoing research... furthermore you'd need a pretty kickass system to make the most of any of this. We can all rag on ringing (and pre-ringing is decently suboptimal) though no speakers exist without some natural tendency towards post-ringing. 

 

I'd think cymbal clashing in MQA would sound pretty damn good, though I'd stress that the argument at 768 gets pretty limited (hence the arcane example). Someone mentioned earlier that it all looks great on impulse response tests which is true, and that life and music aren't a series of them - this is correct. When a friend did an MQA demo I joked with him 'what'd they play, paper getting scrunched?' and he said yes! Sounded awesome. And not music. 

 

That's not to say that music can't sound better with (in a spectral sense) entropy dedicated towards the original intended waveform more than ever before. Just that I wouldn't be judging it's potential with an impulse response plot for a worst-case brickwall against a best-case MQA response. 

[Ittaku]

@rmpfyf I appreciate your thoughtful response and I understand the differences you're pointing out that go beyond just sample rate. I guess my point is that with modern recording equipment, high bitrates should be enough to go beyond the audible limits of humans if I'm finding it impossible to distinguish any difference between 24/96 and 24/384 for example. Given Bob's fancy marketing bullsh... graphs show that MQA outreaches what normal recordings can do until the bitrate is prohibitively high (768 or higher) then I'm not really sure what target it's chasing if 96 is already beyond the audible limit of difference except a set of numbers they claim is important. By the MQA logic, it's audibly better till 768+ whereas I say it's not. I've not seen any scientific test ever where someone has reliably told any higher sample rate apart; it's not just me (and I do have a pretty high resolution system but this extends to others.) I say cymbal clashing on a good 96/24 recording sounds pretty damn good. I'd say generally 44.1/16 doesn't sound as good but extremely carefully recorded encoded modern versions can. Again this is not to say MQA has no place, just that I don't see how it can offer an audible advantage over true high bitrates. Now that's not to say that high bitrates guarantee a good recording, it's just that it's likely we're reaching a point where just about any modern high bitrate recording ADC process is likely enough without the addition of more proprietary technology and remastering every damn thing from scratch. I see it as a tool for rescuing older digital recordings.

Edited December 3, 2017 by Ittaku