DimitryZ
pfm Member
I think he will do what he believes to be honest.
So stop your innuendo and second mouthing mansr's gross accusations.
MQA pt II
- Thread starter Jim Audiomisc
- Start date
So stop your innuendo and second mouthing mansr's gross accusations.
flatpopely
Prog Rock/Moderator
<moderating>
I've deleted some noise. Can we please move away from cyclic arguments.
Those who continue can look forward to a break.
<end>
I've deleted some noise. Can we please move away from cyclic arguments.
Those who continue can look forward to a break.
<end>
DimitryZ
pfm Member
Thank you for removing the gross stuff. Much appreciated.
flatpopely
Prog Rock/Moderator
Can everyone read the AUP please.
Werner
pfm Member
Yes. We know that. Have known that since end of 2014.
The question is: what for?
You don't need MQA for that. If you level the field by selecting the same passband and the same resolution, then FLAC gives similar reductions in the amount of data. Perhaps even better, for all I know.
So I ask again: why proprietary algorithms and proprietary hardware, if not for getting a grasp on the entire distribution chain and extract money from it?
Which is valid. But it is equally valid to resist.
Now where did you get that from???
A CD channel is 44.1kHz, 16 bits. An MQA file is 44.1kHz (or 48kHz), 24 bit. You can force MQA through a CD channel, but only by further taking out data, ending with something that, when left undecoded, is decidely less-than-CD-quality, and when decoded is, well, what exactly?
But let's leave MQA-CD alone. That is just an afterthought, added to the family in order to ... extract money from more places.
You were probably referring to the fact that an MQA high-res stream (i.e 2x, 4x the base sampling rates of 44.1k or 48k) can pass through a 1x channel.
That would be a nice thing if it had any use.
Maybe it was faintly relevant in 2010. Today no more. Look at what is racing around the internet today. Do you (or MQA) really think that adding 2x or even 4x audio streams is going to be a challenge????
I agree. I've never used its lossy nature as an argument against MQA in a context of audibility and when fully decoded. But many others got hung up on this. Perhaps understandably, as an audiophile knee-jerk reaction.
But what I am firmly objecting to is:
1) what it does to the undecoded sound: there is real potential for harm there, that is a simple fact that anyone who knows how the MQA encoder works will see immediately. This potential for harm is getting more accute these days now that Tidal apparently is replacing its formerly CD-sourced files with MQA files with the bottom bits stripped out (allegedly: I am not actively following this because I do not give a sh*t about Tidal, MQA, and whatever in my daily music life). In other words: it plays, but it sure no longer is what you thought you were paying for. Even if this story is untrue, it is perfectly possible to start doing so tomorrow. Digital Rights????
2) as stated above, the totally unnecessary and wasteful nature of all of this.
3) the never-ending barrage, since 2015, of utterly deceptive Good News to convince the gullible, with the established audio press on front row, how this is going to start a new golden era with supreme audio quality, guaranteed and authenticated, no less. Analysis of the decoder (see the work of Mans and others, which BTW is perfectly legal work) has blown some serious holes in that.
4) the deceptive, obfuscating, and in some cases uncouth, nature of MQA's handling of all enquiries and critique. Either by MQA themselves, or by their horde of useful idiots. The resulting smokescreens are so effective than in any attempt to debate both sides, proponents and adversaries, are throwing around lies and half-truths more often than not. (Which is why I left these debates already years ago.)
As for technical knowledge ...
Given the inclination and time (plenty of time: I have no practical experience in coding for modern platforms and in modern languages) I could build something like MQA. I am sure Mans can do this (much better and in a lot less time), and Jim probably too.
Can you? Can Amir?
Don't bother to reply. MIT degree or not.
Jim Audiomisc
pfm Member
OK, here is a starter dealing with what I'm doing:
http://www.audiomisc.co.uk/MQA/investigated/MostlyQuiteHarmless.html
Note that it is just the starter, and focusses on a specific area. More to come later taking a wider scope. There will be typos, etc, that I'll eventually fix. (I have 'Liz Dexia' to some extent so can only see some typos, etc, once my brain has time to 'forget what I wrote' and lets me read what I *actually* typed. )
I have to finish something else I'm writing, and then will get on with the next stage of investigating MQA.
What statement from 'Amir' was being referred to earlier? I've not been reading ASR as I've been busy. But will have a look.
I noted the comment in one posting about pre-ringing being removed. That might tie in with the dispersion filtering found on GO's output file.
BTW 'lossy' for a codec means that we can expect information to be discarded, but if the codec is a good one, and you use a high rate, the effect of the loss may not matter (much) or even pass un-noticed. As R3 listeners may know, their 320k aac is pretty good. FWIW when they also ran a flac stream of the proms I ended up agreeing with their engineers that the two were essentially indistinguishable. Although one of them after repeated tries could spot some differences in a particular case.
However with MQA the issue isn't simply 'loss' it is the *addition* of things not in the source. Waving hands over a 'magic triangle' where 'no music can possibly be' actually in IT terms implies an *inefficiency* in MQA as well as adopting the same stance as 'lossy' encodings of assuming some things can be discarded. This may not matter in practice if no-one can hear the losses or additions, but it makes it shaky in IT terms to say 'lossless' to mean more than a personal opinion on the basis of your individual hearing and experience of the content *when you can judge by comparision* with the same source material conveyed by formally lossless, etc, means. One of our problems here is that last bit.
http://www.audiomisc.co.uk/MQA/investigated/MostlyQuiteHarmless.html
Note that it is just the starter, and focusses on a specific area. More to come later taking a wider scope. There will be typos, etc, that I'll eventually fix. (I have 'Liz Dexia' to some extent so can only see some typos, etc, once my brain has time to 'forget what I wrote' and lets me read what I *actually* typed. )
I have to finish something else I'm writing, and then will get on with the next stage of investigating MQA.
What statement from 'Amir' was being referred to earlier? I've not been reading ASR as I've been busy. But will have a look.
I noted the comment in one posting about pre-ringing being removed. That might tie in with the dispersion filtering found on GO's output file.
BTW 'lossy' for a codec means that we can expect information to be discarded, but if the codec is a good one, and you use a high rate, the effect of the loss may not matter (much) or even pass un-noticed. As R3 listeners may know, their 320k aac is pretty good. FWIW when they also ran a flac stream of the proms I ended up agreeing with their engineers that the two were essentially indistinguishable. Although one of them after repeated tries could spot some differences in a particular case.
However with MQA the issue isn't simply 'loss' it is the *addition* of things not in the source. Waving hands over a 'magic triangle' where 'no music can possibly be' actually in IT terms implies an *inefficiency* in MQA as well as adopting the same stance as 'lossy' encodings of assuming some things can be discarded. This may not matter in practice if no-one can hear the losses or additions, but it makes it shaky in IT terms to say 'lossless' to mean more than a personal opinion on the basis of your individual hearing and experience of the content *when you can judge by comparision* with the same source material conveyed by formally lossless, etc, means. One of our problems here is that last bit.
DimitryZ
pfm Member
Well, though I only have an MIT degree, I will endeavor against odds.
1) I have noted MQA flaws as a universal format for several years, though audibility of its flaws for the general population is debatable (greater than 16KHz).
2) Necessaty is entirely up to market forces.
3) Mostly agree. Their PR strategy has been poor.
4) Complicated. "Enquries" were often essentially hacks, and businesses don't deal well with it. Tied into 3)
mansr
Objectionist
Yes, I probably could do it. I'm not going to, however, because I think it is something that should not be done.
davidsrsb
pfm Member
If I thought that there was a commercial need for bandwidth reduction and a real need for the 24-48 kHz band, I would think about a hybrid structure, FLAC for 48 ksps lpcm and something like 8 bit NICAM like for the ultrasonics to 48 kHz. At least this could leave the base band unmolested
Jim Audiomisc
pfm Member
I stopped going to conventions decades ago. So can't present a convention paper like MQA have. BTW That can bypass peer review and they can then pay to make it openly available.
You are probably unaware that I've written for various printed magazines. Inc New Scientist and Nature (commissioned as well as body). And have written often for HFN for well over a decade, for example.
You may have missed my Liz Dexia comment.
More to come, but I'm currently writing something for a magazine...
But the point of the webpages is to make things openly available. Up to others what they make of it.
Jim Audiomisc
pfm Member
The challenge would hinge on if it was possible to write a 'clean room' MQA decoder that did the job to get the same (or better) results as the 'official' decoder and make it openly available *without* this being scuppered by IPR. I may be able to comment on the first part at some point. But have no idea about the qualifier at the end!
I have also speculated earlier about if it might be possible to remove some of the 'cruft' that encoding-decoding may currently leave in its wake. But as yet, again, I'm not clear on this possibility.
mansr
Objectionist
It is obviously possible to reverse engineer MQA and create an independent implementation. This would be free of any copyright claims, but patents would still apply. The legal situation would be similar to many other open source implementations of proprietary codecs like Dolby wherein distribution of the source code is legal whereas using it, especially commercially, requires a patent licence. Something tells me MQA would not let you have one.
Regardless of the legal status, I don't think a freely available MQA decoder is desirable at this time. Making the format more accessible would only encourage uptake, and we don't want that. If MQA becomes unavoidable, this will obviously change.
Jim Audiomisc
pfm Member
Overall, I'm happy to accept the above. My interest is in this is to understand and explain. And a good way to check you understand a process is to find you can replicate it. That also is a good basis for explaining it to others. But I have no intention of trying to make a decoder and release it for all to use.... even if I was a good enough programmer - which I'm not.
That said, I'm not sure about one point. If someone buys an MQA file or CD then that should mean they are entitiled to listen to it because some of their money has already paid for it. The patents don't cover the details of how the scheme is implimented by MQA. So I'm not sure it would be legally forbidden. However IANAL and I gave up on assuming law made sense a long time ago. I agree with Dickens on that.
mansr
Objectionist
The patents cover the process, not the data. This is no different from MP3 (before the patents expired) or AAC. However, since MP3 and AAC are ISO standards, the patent holders are required to offer licences under fair, reasonable, and non-discriminatory (FRAND) terms. MQA is not a published standard, so they are under no such obligations. Until the patents expire, you are entirely at their mercy, of which they have very little.
Jim Audiomisc
pfm Member
Overall, accepted. But still doesn't clarify a point.
The patents don't specify the filter *coefficients*, for example. Just give a wavy 'example'. So the coefficient *values* are not patented. *A* method of using them is patented. What if someone finds out the coeffients and makes them available.
Can someone else use a process to then decode MQA using them, albeit by working out an alternative approach to the patent?
I'm asking this as a legal point, not a technical one as I'm not sure of the law on this and wonder which might be 'legal'. Regardless of if it is actually possible in technical terms.
Reason being that I object to the way people play the "bun AND the sixpence" with such matters by patenting some parts to hold control for mumble years and also keep critical details trade secrets so they can go on blocking competition and development later when the patent lapses. Makes a mockery of the reasons patents were devised and made law in the first place.
mansr
Objectionist
The "rendering" part of MQA is just a trivial FIR filter with specific coefficients which I have published. The process really ought not be patentable at all, but obviousness and prior art are rarely much of an impediment to getting a patent these days.
The "core" decoder is much more complicated. I expect it to be sufficiently covered by patents that they can't all be worked around. The building blocks (linear prediction filters, band splitting/joining, etc) may not be individually patentable, but the specific combination of them that makes up MQA is (at least in practice). If you don't follow at least some of the patented steps, you end up with an incompatible decoder, and that's of no use to anyone. At least this is how codec patents are usually done.
George J
Herefordshire member
Dear Jim,
Have exactly the same problem. Once I having typed something I read what I intended to write. I have to forget exactly the words and read the words actually written to see where it diverges from what I obviously meant to write!
Thus I usually edit my posts after a half hour, and sometimes several times, though with no wish to alter the intended original meaning ...
When I wrote English essays against a time constraint at school, obviously this meant that there would be mistakes I could not see see immediately. This was before Lis Dexia was a thing of course, and my especially cruel English master, between the age of eleven and thirteen, delighted in humiliating me by reading out what I had written in error to the whole class.
I hated that master. And the feeling was mutual in reality.
Halcyon days! Best wishes from George
Last edited:
DimitryZ
pfm Member
Obviously, if MQA is ever a commercial success, "free" decoders will flood the market from China, not from you.
DimitryZ
pfm Member
GO also claimed that their work was not peer reviewed, but had to make a public retraction in his second video.
My CV is nowhere as illustrious as your but I also have many symposium papers...always peer reviewed.