Saturday, May 21, 2016

MQA heard and liked but

I heard MQA and it sounded good to me.  The audio society sponsored a demonstration with alternating cuts of standard PCM (the first was actually an 88.2kHz 24 bit) vs MQA.  The first person in the audience to describe it used the exact word I was thinking: "natural" for the sound of MQA.  And the presenter echoed that.

I hear clearer attacks and decays, and ambient echoes following notes.  Though the presenter said he heard more highs, I wouldn't necessarily agree.  It was not obviously different in any sort of macroscopic way such as big dynamics or frequency response.  It was in the tiniest details related to focus and I think ultimately timing that things improved.

This confirmed my suspicion that MQA is a worthwhile format, and I will fit it in when convenient.

Of course neither you nor I should believe the results of a sighted test with 3 comparisons.  Done this way the testing was both invalid and insufficient, or it would have been insufficient if there had been validity, but since it was a fully sighted test there was no validity period, the effects could be entirely based on selective memory and preconceptions.

I heard...exactly what I hoped to hear, actually.  I've long understood MQA as cleaning up the timing details.  The culmination of the "apodizing" filter ideas pioneered by Bob Stuart.  Intially he was able to clean up the pre-ringing, but making the post-ringing worse (or maybe I have that backward?).  Such special filters have become commonly available in last decade or so.  The idea of MQA is that if original encoding is accounted for, the pre and post errors can be eliminated.  One can re-create the closest possible replica in time to the original, closer than single ended high resolution systems with apodizing filters, because you can simultaneously account for the phase errors in both encoding and decoding.

That's the idea, anyway.  It sounds good.  It sounds even better than the ideas which went into HDCD, which I've also been a fan of (not without some mixed feelings, but the best HDCD's are better than SACD in every way IMO, whereas CD is only better than SACD in some ways...and yes that's what I meant to say...).  In fact, this is the best sounding idea since...I don't know,  I can't say digital recording itself since that has had issues...perhaps it's the best idea since magnetic recording, though Doug Sax showed that even bypassing that could be helpful.

It's of course just fantasy that my barely tutored ideas about what little I know about digital technologies has any bearing on how they would actually sound.

But anyway, Archimago has done some excellent research on MQA, he's not at all a fan of it, but in one of his tests, pecularliary involving the DSP impulse correction part of MQA which could, in theory, be applied without MQA, he quickly passed his first Double Blind Test (DBT...he knows how to do DBT and does it correctly and sufficiently).

Well Meridian themselves published (not as a journal article, but an "update") in JAES a brief description of their own positive DBT results.  This article was favorably discussed at websites like What's Best Forum, but trashed at sites like Hydrogen audio.  The Hydrogen Audio critique was this: the "control" side of the test used old reconstruction filter technology, whereas the test included apodizing filter.  Well then the test could prove nothing about MQA itself, only the use of apodizing filters, which was become almost standard among the technically literate nowadays.

Wait, that's similar to Archimago's tests.  Not at all actually, but in both cases they were testing a subset of the techniques used in MQA vs old-fashioned-digital.

Anyway it does appear, contrary to the strict interpretation of Meyer-Moran (an AES published test in the 2000's) which some apply (essentially, all digital conversion is transparent, up to 10 layers thick, regardless of digital technology) that some digital is different than others.  This is blasphemy in some objectivist quarters...but can we say that the numbers are now coming showing some variations in digital above 44.1/16/1983 can be audibly different?  And these are involving sets of technologies used in MQA.

OK, this sounds like more than just my fantasy now.

But I have 3 questions:

1) How technically successful is MQA.  How much more accuracy does it provide to impulse response?  How much resolution is lost to make room for the packed information?

2) Is MQA the best possible impulse cleaning technology, and if not, will it prematurely lock us into an inferior standard?

3) Can MQA be used in conjunction with room equalization DSP?

#3 is crucial for me.  Actually I don't know how deeply I can get into MQA without abandoning my use of DSP for crossovers, time alignment, level adjustment, and room EQ.  And I simply won't do that.

For sure I cannot simply tack on an MQA DAC at the end of my system.  First the encoded signal cannot be DSP'd like ordinary PCM.  And once it has been DSP'd, the MQA information would be corrupted.

Quite possibly I can do what I currently do for SACD and HDCD.  I take the analog output of a device which plays SACD or HDCD and re-encode it to PCM digital at 24/96.  (My DSP is limited to 24/96.)  Little resolution is lost, of course the "magic" of DSD is lost, but as I've previously argued, the magic of DSD is hype--there's really nothing there even at best than in high resolution PCM.  A bit of the magic of HDCD might be lost too (the final reconstruction filtering selection...I get that in the intermediate conversion to analog but not the final conversion to analog).

It seems like this can in principle preserve the same part of MQA that would be preserved by people not using MQA DAC's for MQA but instead "MQA converters."  In fact, I could use an "MQA converter" as well, something that takes MQA and converts it to digital at 24/88 (or 24/96, but in this case I think 24/88 would clearly be better).

Working in this way, MQA cannot possibly deliver more than the format it is converted to.  I remember that MQA used end-to-end would in principle provide better impulse response that 24/192 or even much higher, but if I'm converting MQA to 24/96 I'm getting the impulse response which my converter can provide for it, nothing more.

This still makes MQA an excellent (or at least as well as it actually works) way to stream high resolution quality with much lower bandwidth.  Also a way to store that higher resolution quality on my harddrive taking fewer bits.  It's just not giving me something better than high resolution quality I already have for high resolution content.

If there is something more than this that MQA can do, if you actually have MQA end-to-end, but you don't get without having MQA end-to-end, then MQA could be limited by being a closed proprietary system.  It would be better if we had the pieces of MQA available to us and could plug them in appropriate places ourselves.  So I could route the "magic" around my DSP and plug it back into the final DAC (which would have to be a special MQA DAC permitting PCM and bypass "magic" inputs).

A better approach might be to use your already existing CD library, but obtain the MQA supplementary information, with information about your DAC, to give it the best possible correction.

Then you could just buy the MQA supplementary information, which would no doubt be much cheaper than re-buying everything you own.

So I support the approach of the Schiit people, actually.  And for me, end system DAC's are not going to do MQA without some sort of allowance for DSP, which actually I can't imagine.  So for most of my DAC's, non-MQA DAC's would be just fine (I've been looking at Schitt Gungnir with balanced outputs for my subs, also the Emotiva DC-1, but there's some guy who homebuilds DAC's with TOTL Burr Brown for $199,  and the Lite On 1704's with balanced outputs for less than $600).

Still I might get a small MQA dac to convert streaming audio to high res analog, then back to digital, with higher resolution than standard.

No comments:

Post a Comment