I whine about the impulse thing some pages back(or was it another topic?) do not just accept everything they say because it feels ok with the example they show. music is not just how fast an impulse response can stop. that's what they focus on so they show the objective data that makes only that part clear. marketing being marketing.
not only many people would argue that what they propose is improved music, but making impulse responses look good is nothing new under the sun:
https://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf
care for the "white paper" name because while it's a simple example of a few possibilities, the interpretations and claims about improvements are still only the Ayre guy's opinion. not facts.
let's look at page 1 vs page 2. the impulse is clearly better on 2 "wow much time accuracy, such impress, woof woof!" the maximum frequency in the signal basically just goes higher. by the time the attenuation for proper band limiting is reached, we're factually at a higher frequency. if the signal had just been filtered with the first filter but at higher frequency, it would have looked better too.
in the end it's a very basic and simple logic. to reconstruct an ideal impulse, you need to "stack" an infinite number of frequencies, just like with square waves. we see one signal but it's composed of so many sine waves of the right amplitude at the right phase. when you low pass the impulse you remove the highest frequencies, the ones that go up and down the fastest, so the ones that can "draw" vertical lines the best. so you can look at the impulse and think "OMG it's ringing the timing is ruined". or you can think "ok so it's the same sound without some ultrasounds that I can't hear anyway". both are saying the same thing. if the record is a guitar and a dog whistle, filtering the ultrasounds out will make us lose most of the whistle, all the sounds we weren't hearing anyway. but the guitar is just fine. when we filter out some ultrasounds that's what we're doing, removing some signals that never really concerned us humans in the first place.
one argument against my views on this, is about transient response. if we want to draw something with an instantaneous rise, we need crazy high frequencies. that is true. but here is my tiny little problem. do we notice the difference if some ultrasounds are missing and the signal doesn't climb as fast as it should? well I can think of a simple enough testing method. CD vs highres. more frequencies, much improved transient response. again it's a fact. but if it's so beneficial how come we have such a hard time telling both apart in a blind test? to me failing a blind test disproves the need for perfect transient reproduction.
and it's about the same with filters,
when they are clearly out of the audible range, we tend to fail blind tests. and it's obvious why, the ringing or whatever we call it, is happening outside our hearing range. so why do we notice a change when we try different filters on some DACs? why is a MQA dac really changing the sound? why are Ayre bothering with their stuff? well the answer is on those graphs from the previous link. on page 2 the more gentle filter that improves time accuracy, it starts in the audible range and clearly rolls off the trebles in a perfectly audible way. time domain=wins frequency domain=loses. is that better sound? from an objective point of view it's obviously not, even if we were to disregard the extra aliasing. we have a 2 axis signal and we mess one up to make the second look good. then we show marketing stuff about only the one looking good. and this concept is used over and over by everybody including MQA. in fact even the compression process of MQA is losing bit depth for more samples. they can't pull resolution out of a hat, so they move it around from one variable to the next.
back to the Ayre paper again, the third page is a well known, well liked kind of filter for the audiophile. because the ringing is delayed and many people think pre ringing is unnatural(fair enough). here there is phase shift so it's a bad filter from a time perspective, ringing plus shift. yet many people say that they like it better (I have a DAC working that way and one working like the first page, I can't tell the difference. but hey I'm not golden ear and 16khz is about as far as I go ^_^). so already we have many people who do not agree that time should be the only focus right here in the audiophile world.
the last page on the PDF isn't too far off from the meridian apodizing filter concept (included in MQA DACs). the PDF forgets to mention how it still has most of the elements deemed negative in page 2. like upper frequency roll off and extra aliasing(=more distortions), and some phase shift from the page 3 example. because like MQA, Ayre focus on what is important to them, not to the signal fidelity. they subjectively find that it's good. and why not, subjectivity is taste, some like justin bieber, some like celine dion. IMO those choices are just that, choices and cannot be called higher fidelity.
/!\ warning etc.
[COLOR=B22222]Now this is pure speculation[/COLOR], but the ideas I can think about for the ADC/recording bit:
we can take note of where the band limiting is done based on the ADC, then reencode the signal in MQA applying a more gentle filter starting at lower frequency(so cut the end again to change the shape of the cut). something along the idea of page 4. that way the so called "bad" filter that was picked by professionals because they're idiots and only MQA knows better(I assume that's how they think), is mostly cut out of the signal.
I don't know if that's what they wish to do, but it would seem to fit the "make impulses great looking again". the upper freqs that were fine are now rolled off, the distortions levels have increased, there is probably a little phase shift, but hey, the impulse looks great! the consumer can dream of better sound if we only ever discuss time domain stuff.
that was my wild guess about how MQA planned to "improve" the sound from ADCs. the rest of the perceived improved audio I guess goes back to mastering and never needed MQA to be done right(or wrong).