Stereophile Series on MQA Technology

[Galileo365]

On 12/7/2017 at 9:15 AM, Rt66indierock said:

Problem is MQA Ltd. hasn't developed the perfect filter. Charles Hansen was a friend and we recently had phone conversations about digital filtering. The last one seemed to be a brain dump on his filtering ideas shortly before he passed away. 

 

Isn't it the case that all "filters" cause temporal smearing and that, whether MQA actually does or not do this, interpolation of one kind or another (sinusoidal, B-spline, etc.) would work just like a filter does without the time smearing? I believe that Shannon indicated that interpolation is necessary to reconstruct an analog waveform and then someone else used a mathematical model to show that filtering was equal to interpolation in the frequency domain at a lower relative cost but did not address the issues of temporal smearing that such low-pass or brick wall filters cause?

[Galileo365]

Just now, mansr said:

No.

 

Why not? Please explain in more detail if you don't mind.

[Galileo365]

Just now, Audiophile Neuroscience said:

 

Does anyone actually understand what "temporal smearing" means, in contrast for example to "jitter", and what MQA purports to do to fix it ?

 

In so far as I understand it, jitter has to do with the rate at which a stream of bits travel from point A to point B or match an ideally perfect master clock governing a system from end-to-end (i.e, from the original ADC to the final DAC). It's metaphorically like making an analog recording at a given speed then playing it back at a slightly different speed or varying speeds creating a type of distortion in the sound.

 

Temporal smearing is when different analog frequency bands in a signal get pushed out of phase or synchronization with each other for whatever reason. In the same sense that white light is made up of red, orange, yellow, green, blue, indigo and violet light, an analog waveform can be composed of many different frequencies that stay aligned with respect to each other in nature, but fall out of sync somewhere along the line in digital sound reproduction (supposedly as a side of any brick wall filter that cuts the upper frequencies off abruptly assuming that the highest frequency in the signal or waveform is above the cut-off frequency of the brick wall filter) creating a type of distortion as well that is similar to, but not the same thing as,  jitter.

 

Observational evidence suggests that human hearing is far more sensitive to all types of temporal distortion than previously thought.

 

What MQA claims to do, independent of the compression or folding and and unfolding that occurs in the frequency domain, is to preserve the integrity of the time domain by avoiding excessive filtering, using interpolation instead to reconstruct the analog waveform, which has no practical effect in the time domain.

 

I have no way to prove or disprove what MQA claims to do or not do. It seems fairly clear, though, that we, as human beings, are far more sensitive to both jitter and frequency-dependent time or temporal "smearing" as described than was first thought.

[Galileo365]

33 minutes ago, mansr said:

Interpolation is a form of filtering.

 

No, actually, it's not at all. In mathematics, interpolation is defined as, "the process of determining the value of a function [or signal] between two points at which [the signal] has [specific] values", and filtration is defined as, "[the act of] pass[ing] (a liquid, gas, light, or sound) through a device to remove unwanted material." In short, interpolation restores missing information whereas filtration removes, as an example, hypothetically unwanted information. Shannon's proof rested specifically on the notion of interpolation, "Shannon's proof of the theorem is complete [when] he goes on to discuss reconstruction via sinc functions, what we now call the Whittaker–Shannon interpolation formula [...]"; it was only later that someone (I honestly don't have a name) found that, "The interpolation formula [as] derived in the Nyquist–Shannon sampling theorem [...] can also be expressed as the convolution of an infinite impulse train with a sinc function [...]. This is equivalent to filtering the [signal] with an ideal (brick-wall) low-pass filter."  However, a brick-wall filter is defined as, "A certain type of low-pass filter exhibiting a steep cutoff slope which resembles a 'brick wall.' These filters are often found in A/D [and D/A] converters to prevent aliasing; while they are acceptable for this purpose their steep slope introduces unwanted side-effects on the audio signal, such as [frequency-selective] phase shift."--https://musicterms.artopium.com/b/Brickwallfilter.htm. I'm not sure if you're being a troll or just haven't done any real research/experimentation of your own, but all of what I quoted is a matter of fact, not opinion.

[Galileo365]

7 minutes ago, Audiophile Neuroscience said:

Mansr is def not a troll and is highly knowledgeable. He can be a little dismissive at times but due to his young age...19??....we make allowances in deference to prodigious talent ?

 

OK, I see, Audiophile Neuroscience. Thank you. I was once 19, too, and naturally gifted (or so I have been told) in the areas of predicate logic, binary math, and software design and development, relying primarily on my intuition until I more formally educated myself both at college and afterwards. I just learned early on not to dismiss any idea out of hand simply because it contradicted my own thoughts or ideas; but that's more an issue of personal style and humility than anything else. I will say, just because I'm in the mood to do so, that whatever prodigious talent Mansr has, if he or she wants to grow and learn and hone his or her own skills, that listening before you talk or a least explaining why you believe something to be true will likely get you farther than being dismissive. Although, that is his or her burden to bear alone.

[Galileo365]

9 minutes ago, Dr Tone said:

Grabs some ?

 

Hey, in my first full-time job as a software developer and designer, I heard the expression, "Find someone who gives a damn and tell them", just enough times to phrase things more as questions than blanket statements ... (true story).

[Galileo365]

5 minutes ago, kumakuma said:

 

Before you embarrass yourself any further by drawing conclusions based on erroneous information, you should know that Audiophile Neuroscience's comment about Mansr being 19 was a reference to his youthful appearance, not his actual age. 

 

Oh, well, what can I say? I'm new here. <grin>

[Galileo365]

23 minutes ago, kumakuma said:

 

Before you embarrass yourself any further by drawing conclusions based on erroneous information, you should know that Audiophile Neuroscience's comment about Mansr being 19 was a reference to his youthful appearance, not his actual age. 

 

Plus, I have to say that, regardless of Mansr's age or level of experience, I still feel ethically-bound to offer him the same unsolicited advice.

[Galileo365]

7 minutes ago, Don Hills said:

He was right, by the way. In the current context, interpolation is performed with a filter. He knows this because he's disassembled and analysed the software in several MQA decoders.

 

I specifically said that I wasn't referring to MQA but to interpolation versus filtering in the broader context. If he had said, "... in the context of MQA", that would have made more sense. Plus, if you mean that he's actually extracted the firmware from a non-FPGA-based DAC chip, disassembled it, found the part of the code related specifically to MQA, and has proved with certainty that filtering is used in place of interpolation, that would--indeed--be very impressive and convincing. With regard to, "You would do well to take your own advice," I did ask politely several times for him to elaborate, but he choose not to do so; so I did perform due diligence. Regardless of all that, Bob Stuart has openly said that MQA uses B-spline interpolation, a form of interpolation often applied to reconstructing complex curves--not just pure sine waves--from sparse data points. Maybe Bob lied, I don't know; but I honestly don't care how much authority or street-credibility someone has. I want to see the evidence myself or at least have it explained. I will not put any faith into a comment or conclusion made by someone who's only defense appears to be, "Because I said so".

[Galileo365]

1 minute ago, MikeyFresh said:

 

Right, for instance someone like Bob Stuart?

 

Please read exactly what I just said, "Maybe Bob lied, I don't know [...]". I wasn't being sarcastic, I meant I really just don't know.

[Galileo365]

1 hour ago, kumakuma said:

Welcome to CA!

 

Thanks, I'm enjoying the ride.

[Galileo365]

18 minutes ago, PeterSt said:

I don't think so.

 

Yes, precisely. Interpolation is never performed with a filter. A filter can be used in place of a interpolation in some contexts but not without consequences in the real world (i.e., equating a filter to interpolation requires a "perfect" low-pass or "brick wall" filter, something that does not actually exist at this point in time). That comment was originally made in reference to MQA, which claims to use B-spline interpolation but indirect evidence suggests is not true, i.e., several instances of the MQA firmware got reverse engineered and the person doing so came to the conclusion that MQA actually does use filters as opposed to interpolation. I can neither prove nor disprove that. It remans a question in my mind answered only partly by fragments of information and indirect evidence such as the reverse engineering of machine code and ultimately subjective interpretation of the result of doing so.

[Galileo365]

1 minute ago, firedog said:

Correction: on a bagel, the correct term is "schmear". Let's not insult bagels by associating them with MQA.

 

Yes, a "time schmear" is actually a misspelling of "thyme schmear" meaning cream cheese with herbs, totally different thing altogether (ha ha).

[Galileo365]

2 minutes ago, PeterSt said:

That's because the interpretation is merely interpolated.

 

You are, of course, 100% correct. I didn't to say it myself because I thought it would undermine my credibility even more than it already is.

[Galileo365]

13 minutes ago, PeterSt said:

Can someone make an interpolating roll off filter with coils and caps and resistors etc., for a speaker ?

 

I think part of the problem is that "filter" is a kind of a misnomer for an active process such as interpolation, which can lead to confusion. I know the meaning of words change over time and and are often context-dependent. Still, if I was in charge of the world, I'd probably refer to interpolation as more of process than a filter, not trying to be pedantic at all. It's kind of like how most people interpret the word, "calculus", to be the calculus that Newton among others invented to make some calculations involving curves fit into Euclidean geometry, based on points and lines alone. However, "calculus" really means, "a particular method or system of calculation or reasoning". "The Calculus" is simply the most noteworthy current example of a calculus.

[Galileo365]

7 minutes ago, PeterSt said:

Thank you for that introduction. You will recognize me by my tandem bike, double dutch being my partner.

 

 

Cool. Can you arrange for a tour of the Phase One headquarters in Copenhagen and/or get me a reservation at Noma? You must know Peter Qvortrup ... <joke>

[Galileo365]

9 minutes ago, PeterSt said:

But seriously, does someone have a link to that Youtube or text where Bob Stuart tells about the B-spline interpolation ? At least I like to have some fun in laying out how it could be interpreted.

Thanks !

 

I read it first here: https://www.stereophile.com/content/mqa-aliasing-b-splines-centers-gravity.

[Galileo365]

4 hours ago, mansr said:

In signal processing, a filter is a complex multiplication in the frequency domain. This is equivalent, through the Fourier transform, to a convolution in the time domain. This is true for both continuous-time and discrete-time systems. Conversely, any convolution can be regarded as a filter operation.

 

Yes, I know; although, the mathematics behind that presume a “perfect” brick-wall filter, as one example, which is something that at least currently cannot be built. Plus, Fourier transforms work well at expressing attributes of the frequency domain but not so much the time domain, convolution notwithstanding. So using this particular point to say that interpolation and filtering equivocate particularly in reference to “temporal smearing” has no mathematical support and little basis in reality.

[Galileo365]

26 minutes ago, mansr said:

I suggest you spend a few more years studying these topics before trying to argue about them.

I suggest that you stop assuming I haven’t already done so and consider the possibility that you are in error.

[Galileo365]

1 hour ago, mansr said:

 

I studied logic, math, computer science, philosophy, literature, and a field called semiotics at Stanford. The dictionary definition was meant to disambiguate the use of the term filter in the mainstream from how it applies to signal processing such as interpolation, although I did not make that exact point very clear. Fourier analysis is a family of mathematical techniques, all based on decomposing signals into sinusoids. The discrete Fourier transform (DFT) is usually  used with digitized signals. The shift theorem says that a delay in the time domain corresponds to linear phase, a property of a filter. The result is that all frequency components of the input signal are shifted in time by the same amount, referred to as the group delay. There is no phase distortion due to the time delay of frequencies relative to one another. The predominant type of digital filter used in audio is called an FIR (Finite Impulse Response) filter. High-quality sampling ensures that no aliasing occurs by lowpass filtering the signal. A example digital filter would bring a 24 kHz signal down to about -100db. That's a really stiff slope and, in analog electronics, the filter would have a phase shift or non-uniform delay from frequency to frequency. However, digital “filters” are different from analog “filters” because digital “filters” interpolate, an active process in the digital domain, not “filter” passively in the analog domain. So FIR digital “filters” have no group delay. That brings me around to my original point, that the word “filter” in the digital domain is a misnomer given the common meaning of “filter”. So insisting, out of context, using self-accolades and derogatory comments aboit me rather than elucidation (a rhetorical strategy as opposed to a logically deductive proof) that interpolation and filtering are the same thing in general will only confuse the layperson. The “time smearing” or phase shift I refer to occurs as the result of passive filtering in the analog domain, often but not always employed in DAC’s for the consumer market. So when you say that MQA uses “filters” as opposed to interpolation without explication, you are making an inherently ambiguous and self-contradictory statement. Trying to dazzle anyone with long lists of marhemstivsl terms to appear to have superior knowledge without actually disambiguating terms and saying precisely what you mean will not work on me. It’s the equivalent of robbing a bank and offering, as your defense, not if you did or did not rob the bank, but saying simply, “I’m not the kind of person who would rob a bank”. So either say what you mean or stop commenting in dismissive and derogatory ways, because you’re making an ass of yourself as opposed to actually informing anyone.

[Galileo365]

2 minutes ago, mansr said:

That makes no sense.

 

It you read my post carefully, you will find the explanation. I don’t know if you have a background in analog electronics or just signal theory,, but certain terms overlap between the two areas with very different meanings, That's  the last word I have to say on the matter.

[Galileo365]

21 minutes ago, MikeyFresh said:

21 total posts, exactly all of them in the thread, what is his industry/MQA affiliation, if any?

 

Quite sincerely and honestly, none whstsoever. I’m just trying to figure what it’s supposed to do, how it might work, and how to confirm that it does so. I would guess that most of those 21 posts involve defending myself for “asking stupid questions”.

[Galileo365]

27 minutes ago, Ralf11 said:

I'd like to hear more about the shift theorem and what it has to do with MQA

 

Without going into a lot of detail right now because I’m tired of getting flamed, reversing engineering MQA DAC’s is fine if you know how an MQA ADC works, because it is not supposed to be based on Shannon signal theory. Assuming that only the Shannon sampling model is valid and assessing MQA DAC’s from that perspective is like only evaluating quantum mechanics from a Newtonian perspective. 

[Galileo365]

Myek is coming out with a consumer-level MQA ADC and, of course, they already make an MQA DAC. Ayre used to make a QA-9 and a QB-9 ADC and DAC, both of which you can find used. Take an analog recording, preferably from two-track R2R tape. Run that through an oscilloscope. Then connect the MQA ADC directly to the MQA DAC and the QA-9 to the QB-9. Run the signal from the tape through both parallel chains and analyze the output of both DAC’s with the oscilloscope. Then see if the trace on the MQA chain differs at all from the non-MQA chain and, if so, which trace more closely matches the pre-sampled trace. That will tell you more than reversing engineering an MQA DAC out of context. There may be no apparent difference at all or there might be but that’s hard empirical evidence. If all three traes look the same, perform some additional analysis using other measurement techniques to see if any difference ever shows up.

[Galileo365]

5 hours ago, MikeyFresh said:

Or, it's like saying "MQA is many things", and "you don't truly understand MQA" , or it's all hidden behind a very clever NDA so good luck in getting any firm understanding.

 

With all due respect, do you really think BS, Stereophile, TAS, et al. have provided any real response or rebuttal to the findings of @Archimago, or @mansr, or @FrederickV among others? (They haven't).

 

Or are you attempting to do that in their conspicuous absence?

 

 

I’m new to the forum and I don’t know those people at all, only know mansr by virtue of his abusive behavior. Maybe he forgets to take his Thorazine when they let him out of the nuthouse for the weekend. I said MQA is “supposed to do X”, something and that you can only prove it does or doesn’t do by looking both an MQA DAC and an MQA ADC, something that no one here has appparently thought to do. If you had read my very next post, you would have seen a controlled experiment designed to asses that, but maybe the person who reads for you given your apparent illiteracy had to leave for the day.