Advertisement



  1. Things you need to know about the new ‘Conversations’ PM system:

    a) DO NOT REPLY TO THE NOTIFICATION EMAIL! I get them, not the intended recipient. I get a lot of them and I do not want them! It is just a notification, log into the site and reply from there.

    b) To delete old conversations use the ‘Leave conversation’ option. This is just delete by another name.
    Dismiss Notice

MQA pt II

Discussion in 'audio' started by Jim Audiomisc, May 16, 2021.

  1. Jim Audiomisc

    Jim Audiomisc pfm Member

    Yes.It seems likely that an analog filter to remove the dispersion would be somewhere between 'impossible' and 'nightmare'. Hence the curiosity about being able to undo it in the digital stream of info.
     
  2. Jim Audiomisc

    Jim Audiomisc pfm Member

    Having a switch to enable/disable MQA decoding 'mode' isn't the same as having the decoder operate to recover the encoded HF *without* it also producing the dispersion which has been added by the encode-decode by default.

    Can you reference any DACs that can decode MQA and give the HF *without* the added dispersion? I'd be interested to find out about them.

    The question wrt to software decoding is a parallel with the above. Can they be set to remove the dispersion? if not, the stream sent to the final DAC is dispersed. That means the DAC then convolves *its* filter with the already-dispersed info. Can any DAC provide an "undo MQA dispersion" filter? if you know of some, again, references to where they explicity say this would be welcome.

    Otherwise using the software to decode and then having a 'choice' of filters with the eventual DAC simply convolves what the DAC does *in addition* to what the MQA filter has done. i.e. it doesn't remove the 'salt', just adds some 'pepper' according to the taste of the consumer...

    BTW as I understand it, the claim by MQA is that you get 'master quality *assured*' - which implies that when you use it *all* MQA DACs will sound indistinguishable in MQA mode. If they differ, then at least one of them can't be 'master quality assured'. Or their claim is nonsense... 8-]
     
    leroyd likes this.
  3. Jim Audiomisc

    Jim Audiomisc pfm Member

    In principle, if something like roon is running on Linux and *if* it then outputs via ALSA I suspect an ALSA plugin or divert could then run the stream though a 'MQA dispersion removal' filter before it goes to the DAC. Beyond my 'pay grade' as a programmer to do that, but wonder if it is practical and anyone has had a shot at it. ?

    No idea if something similar can be done with Doze/Macs as I've not used them for decades.
     
  4. Werner

    Werner pfm Member

    Which is to be expected, given that the ‘rendering’ amounts to next to nothing, and especially not some magic DAC-individual tailoring.

    Curious then that, IIRC, quite recently BS claimed that rendering was the most important step ;-)
     
  5. mansr

    mansr Objectionist

    Maybe that's where he expects to collect the most licence fees.
     
  6. adamdea

    adamdea You are not a sound quality evaluation device

    It was always going to be tricky having a single inventory item which could be used by everyone and at the same time saying that it was really important to have a special dac. i come back to the early diagram which appeared to show that there was an additional hidden store (IIRC zone c) of information which appeared on the second unfold (or was it rendering, can't remember when that term appeared). And yet all subsequent investigation has suggested that there was no zone c ie it is at most 24/96.

    How popular would MQA have been even with the hifi journalists if it had been presented as a sort of lossy compressed 24/96 with restricted filter choice? Especially when many of them had already pronounced 24/96 as not being enough. I seem to remember john Atkinson saying that vinyl rips only came alive at 24/192. Anyone want to take a bet on how likely it would then have been that they discovered through the unimpeachable scientific method of long term sighted listening that the MQA files sounded better than anything else?
     
  7. DimitryZ

    DimitryZ pfm Member

    Having had four MQA DACs, I can attest that they sound very, very close.

    I think there are several points in the process, where low slope filters will introduce some artefacts, albeit at low levels. Since one has to unfold using matching protocol (I think), it would not be possible to use a different filter in this process. @mansr may wish to comment here.

    Since one doesn't have access to the encode/fold and the unfold is locked by design, the only user access is through the choice of the final "rendering" filter.

    Is it possible to restate the objective of the tests you are running?
     
  8. Jim Audiomisc

    Jim Audiomisc pfm Member

    OK, I've just put up a new page on this at http://www.audiomisc.co.uk/MQA/questions/QuriousAssumptions.html

    Note that it is just a section of what I'm doing at present, so a 'page 1' based on measurements I've been making. I didn't want the page to get too long, or keep delaying making the first parts of the results, etc, becoming available. I'll be working on other aspects of this and will put up another page about them when ready.

    My 'objective' is quite simple. It is to apply the usual methods of science and engineering to probe the claims made, and the actual behaviour of, the MQA system. The aim being to provide evidence based on measurement and scientific analysis which people can then see and consider to help them come to their own conclusions, etc. i.e. SOP for scientists and engineers.

    More to come... :)
     
    adamdea likes this.
  9. Jim Audiomisc

    Jim Audiomisc pfm Member

    So when using 'software decoding' you presumable get the dispersion and any DAC recndering filter is added onto that.

    That means someone would have to add in a digital process to remove the dispersion. I can see this may be possible with ALSA on Linux, but no idea if it could be done with Doze/Macs. Someone who is a better programmer than me would be needed, though!

    What I may be able to do, though. is build a filter to remove the dispersion from the files I captured to see what that does. Useful for analysis purposes. But not much use for normal users who may want dispersion removal.

    I wonder if DAC makers can have a 'dispersion removed' mode or if MQA would refuse them being able to deploy that and keep MQA. Either way, seems curious given the significance they assign to 'deblurring'.
     
  10. mansr

    mansr Objectionist

    Rosewind likes this.
  11. DimitryZ

    DimitryZ pfm Member


    I have written many test plans and reviewed many more, and I would flag you 'objective' as far too broad.

    Test objectives have to be specific and precise and tied to individual measurements. What 'claims' are you studying and what tests/measurements are you using to prove or disprove them? What 'behavior' are you investigating?

    This is of course 'academic' as you are your own customer and the value proposition is entirely yours to approve :)

    Look forward to your results, of course.
     
  12. Jim Audiomisc

    Jim Audiomisc pfm Member

    Thanks. Yes, useful as a reference to for audibility. WRT the dispersion and the magic of 'deblurring': What I have in mind to do is add real-world context via consideration of the temporal behaviour, etc, of mics, speakers, old analogue recordings, etc. I'm also thinking about things like reversing the blanket-applied dispersion to see what it does to the crest factor for example. Might matter given nonlineariities in the chain.
     
  13. Jim Audiomisc

    Jim Audiomisc pfm Member

    You're welcome to do your own analysis and put it on the web.

    However your comments above seem to show you've not actually read what I wrote. That specifies what claims I was considering, etc.
     
  14. DimitryZ

    DimitryZ pfm Member

    I did now.

    Specific claims inquiry should be tied to a test methodology, and, optimally, to specific measurements and values. But this is your wheelhouse and you obviously will do as you see fit.

    I am not educated in this field and can't technically do the required measurements nor interpreted the results adequately. However, I ran many tests that combined mechanical environment with electrical performance (together with EEs and OEs) - like wavefront distortion in heated IR windows, or angle measurement noise in a spacecraft IMU under high ultrasonic vibration. So I feel (perhaps erroneously) that I can make general comments.

    I think if you want to have tight control on MQA/NON-MQA replay of identical inputs, a DAC that has "MQA Enable/Disable" switch will be beneficial. Additionally, some DACs come in MQA and non-MQA flavors - IFI Zen, for example.

    And you may know this already, but some MQA DACs have been faulted for keeping MQA filter on for non-MQA material ( especially if there is one following the other). I would expect Meridian's own Explorer to do this (some of your results maybe suggest this?) - which may be helpful for some of your tests, but unhelpful in others.

    Good luck.
     
    Last edited: Jun 29, 2021
  15. Jim Audiomisc

    Jim Audiomisc pfm Member

    The difficulty here wrt your first point is that the MQA claims tend to be 'vague and sweeping' but implictly 'absolute' ones. That makes the test process different to, say, work I've done in the past for the old NPL as they were able to properly define what needed to be tested. With MQA we have to deduce what MQA's claims *actually* mean when it comes to measured behaviour. So this becomes more of a 'exploration' process.

    And, yes, as reported on the webpages, the Meridian DAC *does* seem to leave the decoder running for a time after MQA ceases to be fed into it. This explains why the non-MQA 'tails' got the MQA dispersion. Useful as it read out the impulse function for the dispersion. In effect, let me get single-non-zero impulses traced out of their filter, showing the function applied.

    At some point I may try to find out how long that 'freewheeling' of the decoder runs from a longer tail and the other test patterns (which I didn't really discuss or use for the current page, but probably will later on. ). Have various other tests in mind.

    By comparing MQA with the 'soxed-to-88k' versions I was able to 'common mode' many aspects of the test comparison. That is better than a change of DAC which could add in other uncontrolled variables. (In a sense it acts *as* an 'MQA on/off switch'.) As it is, the results imply the upper HF approaching/around 40kHz look like 'distortion' products, but seem to fit nicely with the MQA-decoded 1/f slope. So may be deliberate. Too soon to decide.
     
  16. DimitryZ

    DimitryZ pfm Member

    With respect, if you start with vague claims under investigation, you are likely to end up with vague conclusions. I recommend that you distill the claims into a testable hypothesis that then can be unambiguously proved or disproved. My personal interest would be to measure the level of artefacts introduced by bandsplitting and determine if they are at a level likely to be inaudible. And in non-MQA DAC playback, determining if the extra HF noise is audible (I think I have heard this).

    As far as the MQA filter issue is concerned, it's not "freewheeling" phenomena that you captured (this is a new thing that you should be credited with first discovering ;)). What I am describing is that some MQA DACs, when receiving LPCM music after having processed an MQA piece, retain the MQA-specific reconstruction filter (perhaps until they are turned off), instead of switching to the filter the user has picked for non-MQA content. I have read that DAC makers were being encouraged by MQA to do this as it aids in gapless playback of playlists that include MQA and non-MQA material.

    If Meridian Explorer does this, you will get a different read on the tail impulse response depending on the test order, for example. LPCM followed by MQA will show results you posted, while MQA followed by LPCM will show an assymetric results for both. Perhaps you have already checked it...or perhaps power cycling the DAC in between each test clears its' memory.
     
  17. Jim Audiomisc

    Jim Audiomisc pfm Member

    The specific tests I do then give specific results. By doing a variety I can start to 'box in' the claims and find any contradictions, etc. Given a set of vague and sweeping claims, internal contradictions become useful as way to find that at least *some* of the claims must be shakey. :) You can't have your kayak and heat it...

    The 'bumps' show some of the artifacts. This alters in the output, but remains, and seems *very* unlikely to be a common feature of the source material prior to MQA encoding.

    One of the reasons I'm looking at making (linear) corrections to the added dispersion, etc, is to help untangle linear changes from non-linear or added-subracted artifacts that are harder to detect because the dispersion, etc, make a simple 'sample-by-sample-diff' fail to show the nonlinear differences. (I'm also looking at other ways of doing that.) However like steganography, that can be a long game.

    Your pointing out what some DACs do, is useful. I will try to look at that. However when I play the 'tailed' files the MQA led does go out during the tail being played. So it does detect - after a delay - that the material isn't MQA. But as you indicate that doesn't mean the decoder isn't still applying at at least some MQA changes. So thanks for pointing out the other DACs. I'll have a look at this for the Explorer 2. It would be very handy for investigation if it will leave the decoder on for long periods.

    MQA wanting this for gapless reasons seems weird to me. It impies non-MQA files will get 'faked' alterations applied. If I were a user I'd not be delighted by that. Can you list some of the DACs that do this and point me at any details?

    I'll be doing more tests on the Explorer 2 anyway. e.g. want to check what happens when the internal gain is wound down 6dB. As it stands it can't cope with 'overs' very well so want to see if that fixes the limitation.

    And - as always - I encourage you and others to do your own tests to investigate. e.g. if someone is using roon to decode MQA see if you can capture the output or apply changes to it to undo the dispersion, etc. Given what I've seen thus far I suspect that using even a fairly 'budget' ADC like the Scarlett 2i2 3rd Gen would yeild useful data. (I have one of these and it works fairly well. Just not in the ADC1 class - but of course, only a fraction of the price!)
     
  18. mikechadwick

    mikechadwick pfm Member

    Read this and was really surprised that MQA was used in the mastering to capture the “Original Sound”:

    Montreux Jazz Festival and BMG have announced the forthcoming releases of ‘Nina Simone: The Montreux Years’ and ‘Etta James: The Montreux Years’ on Friday, 25 June 2021. The live albums, which will be available in multiple-formatconfigurations, including double LP and two-disc CD editions, feature sublime collections of the musical icons’ finest Montreux Jazz Festival performances, including previously unreleased material, all restored to their full glory and more. The audio will also be available on digital download and streaming services, and in full master quality on TIDAL.
    Simone’s and James’ albums are the first releases of Montreux Jazz Festival and BMG’s brand-new collection series “The Montreux Years”. The collections will uncover legendary performances by the world’s most iconic artists alongside rare and never-before-released recordings from the festival’s rich 55-year history, with mastering performed by Tony Cousins at London’s iconic Metropolis Studios, incorporating MQA to capture the original sound of these special live performances. Each collection will be accompanied by exclusive liner notes and previously unseen photography.
     
  19. r0dd3r5

    r0dd3r5 Active Member

    Perhaps that's why they say "...all restored to their full glory and more." [Emphasis mine]
     
    mikechadwick likes this.
  20. Jim Audiomisc

    Jim Audiomisc pfm Member

    At times the behaviour of the music biz reminds me of the line sometimes attributed to Ronald Regan: "What matters is sincerity. Once you can fake that, you've got it made!" :)

    Anyway, I'm currently diverted by more important things for a while. e.g. just been ordering some Darjeeling FF teas. :)
     

Share This Page





Advertisement


  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice