Article: MQA: A Review of controversies, concerns, and cautions

[tmtomh]

Excellent piece, and I think sections 2 and 3 in particular are key: The distortions and nonlinearity MQA creates; and the "crown jewels" argument, which makes absolutely no sense unless MQA is inferior to the 24/192 PCM masters the labels have.

[tmtomh]

On 3/3/2018 at 2:50 PM, John_Atkinson said:

 

The 2014 article in which I used that phrase can be found at https://www.stereophile.com/content/ive-heard-future-streaming-meridians-mqa

 

I request that CA posters read the full text of what I wrote, in order to comprehend the context.

 

John Atkinson

Editor, Stereophile

 

 

Like others, I read the full article to comprehend the context, and by any rational accounting, @Archimago's use of the partial quotation is totally consistent with the context. In fact, Arch omitted some language in your piece, in adjacent sentences, that is even more over the top.

 

You can argue that your language was justified, but the notion that Arch quoted you in a way that made your comments sound more superlative than they really were, or that took your words out of context in some way, is mistaken.

[tmtomh]

1 hour ago, james45974 said:

can you translate for non math geeks?  

 

I'm not a math geek and can't begin to explain it clearly using the actual math. But FWIW, here is my understanding of the core logic of what this rebuttal is saying. I am of course more than happy to be corrected if I've got it wrong.

 

1. Fourier Uncertainty Principle dictates that there's a limit to how accurate digital sampling can be when it comes to timing and frequency. In other words, there's always going to be some small, irreducible level of uncertainty, and therefore potential variation or inaccuracy in the digital sampling. Just for the moment, let's call that level of inaccuracy/uncertainty, which digital sampling cannot get beyond, "X." 
2. The original study tested human subjects, apparently by playing them three pulses that varied slightly in frequency and/or timing. It found that the humans could detect variations that were smaller than X.
3. On this basis, the original study claimed that humans can discern timing differences beyond what digital sampling is able to control for - in other words, very high sample rates are necessary in order to better compete with how good human timing hearing is.
4. This rebuttal article says the original article mis-used that X figure. They say that for the type of test the original researchers ran, the limits of Fourier Uncertainty are far smaller than X. Therefore, humans' ability to do better than X in that type of test does not in fact demonstrate that human timing accuracy is better than digital sampling can provide.

[tmtomh]

1 hour ago, botrytis said:

 

Thanks for translating for us non-math geeks. Even better, you made it understandable. 

 

cheers!

 

No problem! But I made it understandable only if my summary actually is accurate! Hopefully one of our more math expert colleagues will weigh in to say if I've gotten the basic idea right, or if I've completely mangled it. 

[tmtomh]

I agree it would be great of @John_Atkinson remained here - but only because it is clear that the discussion here has helped shape Stereophile's coverage of MQA. He won't admit that of course, and as @crenca has noted, he's explicitly stated that he will not actually engage here, only weigh in to correct what he sees as factual misstatements about the magazine and/or him and other authors.

 

And more recently he's made a very polite but nevertheless transparent argument suggesting @Archimago's data should be dismissed - or at least not actually engaged with - because of the pseudonym issue.

 

But even with that, there's no need to antagonize Atkinson. Just because someone enters a community and behaves uncivilly by refusing to participate in the accepted, collegial manner, does not mean that we have to be uncivil or rude in response. Nor does it matter whether or not he ever will admit that CA's discussions might have an impact on Stereophile's coverage. No point in trying to fight that battle when the tone and content of Stereophile's actual coverage is the more important issue.

[tmtomh]

34 minutes ago, The Computer Audiophile said:

One other note about not engaging or not refuting data from an "anonymous" person, this is one of the first things Bob mentioned to me when he called. I'm not saying that Bob is setting JA's agenda or giving him talking points, but I just don't see why JA is sticking to that argument so hard. 

 

I think this is very common. People who oversee or are part of an enterprise, whether it be MQA or Stereophile or whatever, have to stay focused on their main goal and not get distracted or drawn into situations where there is no upside for them.

 

Trying to put myself in their shoes, I don't see any upside to Stuart or Atkinson engaging with Archimago's data, or with CA in general (beyond Atkinson's self-appointed fact-checker role). Better for Stuart to say on message, and better for Atkinson to incorporate any valuable insights here into Stereophile's articles as he sees fit. Why elevate CA's stature if they don't feel it will help them, and why participate in someone else's format and narrative when you can instead tell  your story with your own narrative structure?

 

I of course agree with you Chris, and with @mcgillroy above, and with many others here, that it's a shame (or worse) that Stuart and Atkinson see Arch's info and CA's discussions as not having any upside for them to engage with. That tells you a lot about their agendas and attitudes, and in particular it speaks volumes about @John_Atkinson's level of professional and collegial respect for @The Computer Audiophile, unfortunately.

[tmtomh]

1 hour ago, james45974 said:

Yes, but a no upside calculation for MQA or Stereophile is not necessarily neutral, there could be a downside, like to your reputation!

 

No disagreement there - especially given how big CA's readership/membership base is.

[tmtomh]

3 hours ago, sullis02 said:

Speaking of  'time domain performance' correction--  remember Plangent processing?  It "basically uses recovery of the bias tone off analog recordings to "realign" the audio to the state it was in while being tracked or mixed."   So I'd guess the time domain aberrations being corrected with PP (tape wow/flutter) are likely orders of magnitude worse than the ADC-based (non?)issues MQA is aiming at...but then again I don't really know what MQA is aiming at!  

 

http://audiophilereview.com/analog/plangent---a-better-way-to-transfer-analog-tape.html

 

 

 

For years I've been wishing there'd be a computer audio editor plugin for the Plangent Process.

[tmtomh]

Our old friend Lee Scoggins is at it again over at the Hoffman forums, including some passive-aggressive character assassination of both @Archimago and @The Computer Audiophile:

 

http://forums.stevehoffman.tv/threads/mqa-a-clever-stealth-drm-trojan-ccc-talk.735825/page-6#post-18296864

[tmtomh]

4 hours ago, T.S. Gnu said:

Mister Stuart compared himself to Copernicus a while ago (Last para of article: http://www.theabsolutesound.com/articles/meridians-master-quality-authenticated-the-interview/). Sadly his following comment ("I guess I’m luckier, because the worst we’ll get is the wrath of audiophiles and scientists, not excommunication.") fearing excommunication further indicates fluidity with facts; Copernicus wasn't, and couldn't have been, excommunicated because his work was published after his death.

 

Having read this post, if I may be so bold, I would suggest you walk down the beach and think, "Now I know what Martin Luther felt." After all, this is addressing indulgences in a way. One could also say you could guess you will be luckier, because the worst you’ll get is the wrath of audiophiles and non-scientists, not excommunication. And you might be right because he, on the other hand, was excommunicated.

 

Irrespective of how this shakes out, history will view favourably the fact that issues were allowed to be discussed on this site free of dogma and preconception with intelligence not viewed as some hideous affliction. 

 

As to the question of anonymity that Mister Atkinson, Mister Stuart and their ilk bring up, it is perhaps respectfully fitting to quote a true follower in the steps of Copernicus:

"In questions of science, the authority of a thousand is not worth the humble reasoning of an individual." Galileo

 

Thanks for the link to that TAS interview with Stuart.

 

What knocks me out is that he's extolling the use of undersampling (via Nyquist Shannon) and using that notion to dismiss pretty much the entire audio/audiophile community as old-fashioned and ignorant (like flat-earthers compared to Copernicus).

 

To the best of my knowledge, undersampling makes sense only when the range of frequencies you need to sample is significantly narrower than the range of frequencies between DC (zero Hz) and the upper frequency you need to capture.

 

For example, a common cell phone frequency band is 1900MHz (not kHz), with a 100MHz bandwidth, meaning the phones operate from 1850MHz to 1950MHz. By using simple Nyquist theory, to digitize those waves, you'd have to sample them at 3900MHz, which is 2x the max frequency of 1950MHz.

 

Undersampling theory says it's a waste of data space and energy to use such a high sample rate, because the frequency band from zero Hz to 1850MHz contains no info, and so all the sampling of those frequencies is wasted. Instead, you can undersample and use the aliased frequencies to reconstruct the original signal, because while the aliased frequencies will be wrong, you will know exactly by how much they are wrong and you can use math to reconstruct the signals.

 

But in audio, the range is 20Hz to 20kHz. The only "wasted: sampling occurs between zero Hz and 20Hz, which is virtually nothing.

 

Now, it is true (as far as I know) that MQA uses undersampling theory to encode, enfold, and reconstruct some of the ultrasonics in MQA files. But as @mansr explained somewhere in these forums about a year ago, the result of doing this via MQA's chosen filters is that the price you pay for the reconstruction of ultrasonics is aliasing in the audible band. It's just silly.

[tmtomh]

4 hours ago, mansr said:

Right. The only restriction is that no multiple of half the sampling frequency may fall in the band of interest as these frequencies are impossible to capture unambiguously. If this is observed, reconstruction is as simple as applying a bandpass filter. As a special case, for sampling a modulated radio signal, it is ok for the carried frequency to be such a multiple since we only care about the deviations anyway. Needless to say, none of this is relevant for audio.

 

Thanks for the confirmation and further explanation!

 

I've bolded the last part of your comment, because in the TAS interview Stuart precisely makes the claim that this is relevant for audio, and that he's bringing the Good News to the us Philistines of the audiophile world.