MQA is Vaporware

[opus101]

Just now, manueljenkin said:

There's absolutely nothing wrong in the content I posted (the link is just a reference and the content that I agree with I have quoted).

 

So you claim. But others disagree with that claim. Isn't that what forums are for - to explore disagreements and learn stuff from each other? Not to shut down discussion with comments such as 'Can we keep a check on unsolicited comments?' If you don't like unsolicited comments on what you post, simply don't post.

[FredericV]

1 hour ago, manueljenkin said:

I'm sorry, I wasn't asking your opinion on MQA or some random video. Can we keep a check on unsolicited comments?

You quoted from an MQA article:
 

Quote

"Of particular interest are new techniques which essentially discard brick-wall anti-alias filtering as we currently know it, and employ new forms of sampling and reconstruction kernels that can resolve transient signal timing with extraordinary resolution, even conveying positional time differences that are shorter than the periods between successive samples!"

while regular PCM already offers positional time differences that are shorter than the periods between successive samples

The video is not some random video!

 

1 hour ago, manueljenkin said:

I was asking for books/resources on wavelet theory and time weighted kernels. Not sure how hard that is to understand. I have done enough math to know that the particular paragraph I have quoted has no mistake (not sure of the rest of the article and I couldn't care less).

The paragraph you quoted insinuates MQA invented the wheel, which it did not.

Why are you interesting in "more research"?

 

[Currawong]

The original MQA article has a false premise, in that the idea that the ringing surrounding an impulse response (an illegal, out-of-bandwidth signal) is related to anything audible, which it is not. It's the same false premise surrounding those graphs showing no ringing from an impulse response put through a non-oversampling DAC as being better since, when you put actual music through a NOS DAC, you get very audible and measurable distortion.

[manueljenkin]

21 minutes ago, Currawong said:

The original MQA article has a false premise, in that the idea that the ringing surrounding an impulse response (an illegal, out-of-bandwidth signal) is related to anything audible, which it is not. It's the same false premise surrounding those graphs showing no ringing from an impulse response put through a non-oversampling DAC as being better since, when you put actual music through a NOS DAC, you get very audible and measurable distortion.

I am not really thinking in terms of audibility and I certainly am not convinced by the rest of the article. NOS is a different thing and I don't subscribe to it either for reasons you have mentioned. It's pretty much stepped tones with lots of aliasing (not sure if the aliasing is audible, but these quick steps can cause analog system to be very unstable and hence cause distortion).

 

Though pretty sure it's time domain properties at the sampling instants if done well (not sure if that happens though) are better than when convolved through sinc. Schiit seems to have a pretty interesting way around it where they both try to interpolate with something similar to sinc but not exactly sinc while keeping the sampling instants unaltered.

 

That's besides the point anyway. I'm mainly interested in reading about wavelets and splines to learn and possibly see what they are truly optimising for (of course it's relating to temporal structure) instead of just cocooning myself with a basic short time Fourier transform analysis. The exact sentences I've quoted don't have any content relating to audibility, just math, albeit something of a higher level.

[KeenObserver]

3 hours ago, jparvio said:

Not sure what to make out of this. What exactly has been changed in MQA:

 

https://positive-feedback.com/industry-news/simaudio-announces-tidal-master-mqa-problem-with-the-moon-mind-2-fix-coming/

 

Are they trying to hide the fact that its MQA?

[KeenObserver]

2 hours ago, manueljenkin said:

There's absolutely nothing wrong in the content I posted (the link is just a reference and the content that I agree with I have quoted). @The Computer Audiophile I wasn't expecting to face random condescending comments for a gentle post requesting resources on a particular topic relevant to this thread.

 

You came to a MQA thread and posted a quote from an MQA article. You started with the basic assumption that the quote was fact because you "knew enough about math" to know it was fact. Others challenged the basic assumption that you were basing your whole post on. And you became upset that someone questioned your statement of fact as being fact. As you state. forums are open to discussion and examination of statements given as fact.

It's kind of like the " Do you still beat your wife" kind of statement.

Since its inception MQA has been misleading and outright lying ( it is not lossless). People have responded by carefully analyzing and refuting false claims.

[KeenObserver]

This resurgence of MQA BS is very concerning. It makes me think that MQA and its backers are making new efforts to ram MQA down the throats of the music consumer.

[Rexp]

5 hours ago, manueljenkin said:

Not here to defend or accuse MQA but to learn something about signals. One major thing people tend to forget is most recordings done using delta Sigma adc don't have the high frequency content at all due to the noise shaper structure (and the brickwall that follows). Pulling through a windowed fft based spectrum analyser software like spek I seldom found anything to have ANY content above 20khz regardless of recordings. Very few recordings had such content preserved.

 

 

This is 24/192, shows content above 20k

[manueljenkin]

1 hour ago, Rexp said:

This is 24/192, shows content above 20k

Some (very few) do which is why I used the word most instead of all. Thanks for the recommendation btw, I'll try to get the same music. Amber rubarth was fairly fine with hf content but ottmar liebert and the ones I got from 2l no hires didn't have high frequency content in the spectrogram.

 

Any idea what gear the above song was recorded with btw?

[manueljenkin]

1 hour ago, KeenObserver said:

 

You came to a MQA thread and posted a quote from an MQA article. You started with the basic assumption that the quote was fact because you "knew enough about math" to know it was fact. Others challenged the basic assumption that you were basing your whole post on. And you became upset that someone questioned your statement of fact as being fact. As you state. forums are open to discussion and examination of statements given as fact.

It's kind of like the " Do you still beat your wife" kind of statement.

Since its inception MQA has been misleading and outright lying ( it is not lossless). People have responded by carefully analyzing and refuting false claims.

The entire paragraph I have quoted is fully relating to math (specifically to wavelets and time-frequency resolution), and as I said I linked the webpage just as reference. Some people seem so insecure (of being shown wrong) that the only way they find an escape is to mute any further analysis of the subject. Idk about your sentiments on MQA, I personally don't care for it since I don't stream my music. As simple as that. I am just interested in learning signals and all I did was ask for resources to learn wavelets. There's been over a dozen comments since then none answering the actual question. Pretty much sums it up.

[KeenObserver]

"Look! There's Bigfoot".

[John Dyson]

3 hours ago, Currawong said:

The original MQA article has a false premise, in that the idea that the ringing surrounding an impulse response (an illegal, out-of-bandwidth signal) is related to anything audible, which it is not. It's the same false premise surrounding those graphs showing no ringing from an impulse response put through a non-oversampling DAC as being better since, when you put actual music through a NOS DAC, you get very audible and measurable distortion.

Man the non-ringing ringing is a major hotspot with me...   We all use the term 'ringing' when talking about 'Gibbs', and sly sales people  will use the 'ringing' (not ringing) to scare good, normal audiophiles into thinking that the Gibbs effect is bad or has ANYTHING to do with true ringing.  (Gibbs can cause clipping or extra nonlinearity, but those are the real attributes that can be audible -- the ringing just isn't there.)   That is, if you ADD the correct things to the signal, then the 'ringing' (Gibbs) can be made to go away (odd, huh?

 

I know that you @Currawongunderstand, because this has been explained over and over again around here, but Gibbs is a manifestation of MISSING sine wave harmonics of a signal, therefore do not cancel the signal ad-infinitum in frequency, therefore leave the residue that we all mistakenly call 'ringing'.   This is sometimes even misunderstood by non-DSP technical people from time to time even in places like ASR.

 

The ONLY way that Gibbs can be heard is about the loss in HF, not anything about ringing or sustaining certain aspects of a signal.

The reason why 'Gibbs' appears to move around based upon constant-delay filters (linear phase), or not-constant-delay filters (often minimum phase) is because the frequency components are delayed differently when using filters that are NOT constant delay (linear phase.)

 

Any organization that uses 'Gibbs' as an attempt to show superiority, inferiority or show any kind of attribute WRT audibility (other than the associated rolloff) is not being honest, or just does not know.   When a psuedo-technical claim is made that suggest the 'badness' of Gibbs (even suggests it), then they lose a lot of credibility with me.  If you really want to suppress 'Gibbs', a nonlinear scheme can be used, or don't brickwall the bandwidth.  (I have a filter method that I use for control signals that mitigates the Gibbs ringing, but the BW is also not brickwall.)   In control signals, the peaks and changes in transient behavior CAN have serious negative effects, just like Gibbs can cause clipping in certain cases.   So, Gibbs isn't benign, and in the sense of potentially causing clipping or encourage nonlinearity, I guess that Gibbs can be audible, but that is about it.

 

The reason for my vehemence is that I don't want my non-technical audiophile friends to be misled by silly or dishonest claims based on most claims about Gibbs (other than the clipping or encouraging more nonlinearity.)   Even technical people can initially misunderstand the Gibbs thing until the really think about it or work through the math.

 

[manueljenkin]

20 minutes ago, John Dyson said:

Man the non-ringing ringing is a major hotspot with me...   We all use the term 'ringing' when talking about 'Gibbs', and sly sales people  will use the 'ringing' (not ringing) to scare good, normal audiophiles into thinking that the Gibbs effect is bad or has ANYTHING to do with true ringing.  (Gibbs can cause clipping or extra nonlinearity, but those are the real attributes that can be audible -- the ringing just isn't there.)   That is, if you ADD the correct things to the signal, then the 'ringing' (Gibbs) can be made to go away (odd, huh?

 

I know that you @Currawongunderstand, because this has been explained over and over again around here, but Gibbs is a manifestation of MISSING sine wave harmonics of a signal, therefore do not cancel the signal ad-infinitum in frequency, therefore leave the residue that we all mistakenly call 'ringing'.   This is sometimes even misunderstood by non-DSP technical people from time to time even in places like ASR.

 

The ONLY way that Gibbs can be heard is about the loss in HF, not anything about ringing or sustaining certain aspects of a signal.

The reason why 'Gibbs' appears to move around based upon constant-delay filters (linear phase), or not-constant-delay filters (often minimum phase) is because the frequency components are delayed differently when using filters that are NOT constant delay (linear phase.)

 

Any organization that uses 'Gibbs' as an attempt to show superiority, inferiority or show any kind of attribute WRT audibility (other than the associated rolloff) is not being honest, or just does not know.   When a psuedo-technical claim is made that suggest the 'badness' of Gibbs (even suggests it), then they lose a lot of credibility with me.  If you really want to suppress 'Gibbs', a nonlinear scheme can be used, or don't brickwall the bandwidth.  (I have a filter method that I use for control signals that mitigates the Gibbs ringing, but the BW is also not brickwall.)   In control signals, the peaks and changes in transient behavior CAN have serious negative effects, just like Gibbs can cause clipping in certain cases.   So, Gibbs isn't benign, and in the sense of potentially causing clipping or encourage nonlinearity, I guess that Gibbs can be audible, but that is about it.

 

The reason for my vehemence is that I don't want my non-technical audiophile friends to be misled by silly or dishonest claims based on most claims about Gibbs (other than the clipping or encouraging more nonlinearity.)   Even technical people can initially misunderstand the Gibbs thing until the really think about it or work through the math.

 

I understand Gibbs phenomenon which is why I didn't take the rest of the article into quotations, and yet the thread got diverged to that way. My questions and the content I quoted was fully relating to wavelets and splines. Maybe this isn't the right place to ask for suggestions on books/resources to go through a "specific topic", relating to the content of the thread, if it has any deviation from established cocooned thinking.

[DuckToller]

12 minutes ago, manueljenkin said:

I understand Gibbs phenomenon

here are some behind the scene informations about Gibbs optimization:

 

[manueljenkin]

Why am I being constantly given replies with content totally irrelevant to the actual question I was asking. This is like asking for directions to a nearby hotel and getting responses to go to the beach at the next edge of the continent. I am not here to have arguments on manufacturers wierd naming schemes for their products. One can go ahead and ask is the schiit wyrd really wierd. Or is it named so because it is wired? Or is it a white yellow Red denim color scheme?

 

I asked for resources on wavelet theory and splines, and the exact sentences I quoted relate to that alone.

[jparvio]

3 hours ago, KeenObserver said:

 

Are they trying to hide the fact that its MQA?

 

Your guess is as good as mine. 

[DuckToller]

27 minutes ago, manueljenkin said:

Why am I being constantly given replies with content totally irrelevant to the actual question I was asking

Sorry about making you complaining again ....
Could you - perhaps in a less scientific manner - explain to non-mathematicians like me why & at which point of the discussion in this thread you started thinking about "wavelet theory and splines" having great importance for the "MQA is Vaporware" thread??

[manueljenkin]

8 hours ago, manueljenkin said:

Not here to defend or accuse MQA but to learn something about signals. One major thing people tend to forget is most recordings done using delta Sigma adc don't have the high frequency content at all due to the noise shaper structure (and the brickwall that follows). Pulling through a windowed fft based spectrum analyser software like spek I seldom found anything to have ANY content above 20khz regardless of recordings. Very few recordings had such content preserved.

 

https://www.soundonsound.com/techniques/mqa-time-domain-accuracy-digital-audio-quality?amp

 

Paragraph below sampling evolution - 

"There are other approaches that don’t suffer the same uncertainty between time and frequency information that is inherent in the Fourier model... and developments in advanced mathematics over the last 15 years or so (particularly in relation to Wavelet Theory and complex image processing) have led to many momentous advancements in both the fundamental concepts and the practical techniques of sampling theory.

Of particular interest are new techniques which essentially discard brick-wall anti-alias filtering as we currently know it, and employ new forms of sampling and reconstruction kernels that can resolve transient signal timing with extraordinary resolution, even conveying positional time differences that are shorter than the periods between successive samples! This seems counter-intuitive but it is possible if, instead of using traditional adjacent rectangular sampling periods, a series of overlapping and time-weighted triangular sampling kernels are used (see Figure 6). Even better results are possible using higher-order ‘B-spline’ kernels, which allow both the position and intensity to be identified of two or more separate pulses occurring within the same sampling period!"

 

Any idea where I can find the math about such time weighted kernels and whole of wavelets in general?

Here is your answer. This is where it actually began and you've tried to poke something random in the middle without understanding the context of the discussion. So it clearly shows that most people here in this thread (a few exceptions) are not interested in a discussion, just find ways to deviate discussion into irrelevant oblivion.

 

What has ifi's usage of Gibbs name in a specific product has to do with what I wrote. Elon Musk has a car company named Tesla. Doesn't mean Nikola Tesla autographed on the units being sold.

 

Why would you assume someone who is researching resources in wavelets hasn't finished basic stuff like Gibbs phenomenon, and windowing functions?

[John Dyson]

1 hour ago, manueljenkin said:

I understand Gibbs phenomenon which is why I didn't take the rest of the article into quotations, and yet the thread got diverged to that way. My questions and the content I quoted was fully relating to wavelets and splines. Maybe this isn't the right place to ask for suggestions on books/resources to go through a "specific topic", relating to the content of the thread, if it has any deviation from established cocooned thinking.

One should be aware of the whole picture and wasteful worrying about interpolation schemes based on gratuitious clipping off a couple of bits to then re-insert the data later.   MQA  is just silly for the application.  Gimme the 16bits with 20kHz bandwidth, no need to take something away, then pay more (more licensing) to essentially create an inferior version.   There is NO audible benefit to provide low level details above 20kHz, esp when the original material is usually mostly noise above 20kHz anyway.  (practically all of the time on anything created with analog tape, when you see 'signal' above 20kHz, you are mostly really seeing noise modulation from an NR system.)   There are *some* boutique outlets that give you clean recordings that have >20kHz content, but they would stay the heck away from MQA.

 

MQA: Take away data/detail from audible frequency ranges, substituting what is effectively HISS, then when 'decoding',  try to approximate essentially inaudible frequency ranges, only for the purposes of specsmanship to sell an engineering wise corrupt scheme.   Once one 'pays the license fee' in one way or another, then they get an APPROXIMATION of the original 16bit signal.   MQA might be scientifically interesting, but is a poor engineering solution -- we, as a society get 'science' and 'research' mixed up with good engineering -- they are different matters, where engineering is usually entails the integration of many things.  MQA is ludicrious engineering, but for mental exercise (the substituted word starts with 'm') and research wise -- no argument.

 

Sure is an 'emperors' new clothes' scheme.

SO -- MQA is a great science project -- just keep the science project away from people who want to hear music.

Talking about MQA about it being a science project -- it makes sense as an intellectual exercise -- so does chess, just keep any marketeering out of it.   We don't need to use the term 'MQA' to talk about the associated math&science project.   As a package, for the corruption of useful audio data that it really is, MQA is an anathema to TRUE signal integrity*.   It is great for license holders and further obscuring of IP.

* Don't look at the 'man' (distortion) behind the curtain -- the distortion is really good for 'you' (I mean, me -- the license holder.)

 

 

 

[firedog]

46 minutes ago, manueljenkin said:

clearly shows that most people here in this thread

Nonsense. You have at most a handful of people who have interacted with you. On an almost thousand page thread.

[Confused]

On 3/24/2021 at 6:29 AM, Currawong said:

I've had it in mind, for quite some time, to make a video about MQA.  If anyone is interested in helping with that, please PM me. 

 

17 hours ago, botrytis said:

Both sides do it, calling names. It is not one side or the other. It really doesn't matter but it kind of does. I mean Chris did not have to put up with the crap he did at RMAF. That shows the Pro-MQA people are more interested in not having the truth out.

 

This gives me an idea.  Maybe Chris could finish his RMAF presentation as a YouTube clip?  A simple voice over for the original presentation would do it.  This could then be billed as the "RMAF presentation that the MQA executives would not allow to be completed."

 

Just a thought ......

[SoundAndMotion]

1 hour ago, manueljenkin said:

Why would you assume someone who is researching resources in wavelets hasn't finished basic stuff like Gibbs phenomenon, and windowing functions?

I can't speak for others, but for me, the clues are:

-you don't know how to google. (presumably)

-you can't distinguish between "sub-Nyquist artifacts" and "sub-sample time resolution"

-your entire "performance" in this thread: 

 

[DuckToller]

1 hour ago, manueljenkin said:

Here is your answer.

Sorry, you lost me with that reply.
I asked for something else you perhaps won't like to write about.  Again: What was your intention to enrich this thread with advanced mathematics and wavelet theory and splines ?

Re: Gibbs, mansr's exposé about the GTO filter, developed by MQA Labs together with iFi Audio in order to counter attack the problems arriving from the Gibbs phenomen and - as of the researcher - very close of MQA filter, supports imho Mr. Dyson's contribution. A solution in search of ... and a obfuscation in support of the MQA scheme.

In my logic, there is no existential need to discuss the mathematical theory en detail in this Vaporware-thread, you even couldn't (or did not want to?) name a non-existential (important vs, trivial) reason for starting this discussion. (my comprehension, ymmv). I may want to think you found the article you'd posted interesting and searched a discussion about it, with request to special topics occuring from it. In my view it looks as well kind of obscuring the real world problems deriving from MQA. (again: personal opinion, ymmv)

I remember you were asking once about a science sub forum - is the objective forum perhaps a better place to discuss these kind of subject in the manner of your interest? Less views - okay - but you might not have to bother with the kind of responses you are unhappy to receive?

[manueljenkin]

11 minutes ago, SoundAndMotion said:

I can't speak for others, but for me, the clues are:

-you don't know how to google. (presumably)

-you can't distinguish between "sub-Nyquist artifacts" and "sub-sample time resolution"

-your entire "performance" in this thread: 

 

I got your logic. If you cannot comprehend it the other person must be wrong!

[SoundAndMotion]

3 minutes ago, manueljenkin said:

I got your logic. If you cannot comprehend it the other person must be wrong!

What can’t I comprehend?