Tidal adds hi-res audio streams with 'Tidal Masters' (with MQA)

Discussion in 'General Audio Discussion' started by weldp, Jan 5, 2017.

Tags:
  1. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    MQA data folding/unfolding is hierarchical. This means that you get different output quality depending on how you replay it. Simplifying things as far as possible, and going based on my understanding of what's been published so far, it works out like this:
    • In all cases you get the (alleged) benefits of the temporal resolution enhancements that MQA promises (these are an encode-time problem not a decode-side one).
    • With no MQA replay hardware or software you effectively get 16/44.1 output (with the caveat that it's really dithered 13/44.1 ... which is really only a relevant "issue" if the source content is using more than 13-bits of dynamic range in the first place). Technically it's possible for the file to be higher resolution that this, so you could see 24/96 output even without an MQA-enabled player/DAC in the mix, though for practical purposes what's actually present in that stream is going to be at an effective rate of 13/96.
    • With a software MQA player (like Audirvana or TIDAL) you will (should) get all 16-bits of dynamic range at whatever relevant bit-rate (or approximation thereof). So, in general, 16/96.
    • With an MQA-enabled DAC you get 24-bits of dynamic range and all of the above, along with some filter-level benefits. This is an artificial restriction, driven by the usual studio/label-paranoia bullshit about having actual "hi res" content in an non-DRM'd form. So, much like DVD-A, SACD and TA, they don't allow digital output at full bit-depth and resolution.
    Still wading through things to find out how accurate this is (i.e. where I'm missing something, or where the model is deliberately subverted) ... but that's the general idea ... that they can progressively unleash "better playback" from a single file depending on your replay chain.
     
  2. Merrick

    Merrick A lidless ear

    Pyrate
    Joined:
    Jan 6, 2016
    Likes Received:
    12,568
    Trophy Points:
    113
    Location:
    Portland, OR
    That feels like shenanigans to me, and I cannot see MQA being pervasive if they play those kind of games.
     
  3. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Try to make a 16/44.1 file quieter by 18db and then normalize it to the original peak level. Maybe add some dithering too. I think I even noticed a bit-depth reduction when going from 24 to 20bit. (In theory the file didn't have more than 10bits of dynamic range, but there's more to it than just SNR. Noise isn't just noise.)*
    As for the sample-rate, I tried a few sample rate converters (@trung225 let me experiment with files he generated with Saracon and Pyramix) and felt that none were transparent with my Gungnir DS when going from 176.4 to 44.1 (or even 88.2) kHz. Maybe a Gungnir Multibit with its superior filter would change the results a bit here.

    *I think it's a good idea if the quantization noise is below the absolute threshold of hearing. This of course depends on your listening level and the dynamic range of the file. 20bits don't seem so unrealistic when you realize that peaks can be 15db+ higher than average with good files and that the threshold of hearing is actually less than 0db in the upper mids. Coincidentally 20bits is what HDCD had in a best-case scenario.
    I also don't think that it's impossible to perceive a difference when the quantization noise is below the absolute threshold of hearing.


    EDIT: My comment was not directed so much at MQA; it was more about digital audio in general.
     
    Last edited: Jan 9, 2017
  4. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    This is only relevant in the event that you're not using any kind of MQA decoder. Even with just a software decoder you will recover all 16-bits of dynamic range*.

    Unless it's the same dithering-with-encoding used by MQA its going to have a different effect and, again, isn't applicable unless you don't have at least a software MQA decoder*.

    Completely irrelevant to MQA. It is not implemented as a higher sample-rate file that is selectively down sampled. The raw PCM stream contains native 44.1 KHz samples (albeit at an effective 13-bit depth + encoded data in the three LSBs). Higher bit rate playback is achieved by using additional sample data that is folded/packed/compressed into the encoded LSB data. This is represented and encoded hierarchically, so the decoders can take advantage of as much of the additional sample data as permitted by the licensing model for that type of output and to the extent that the available hardware can handle it.

    Right now I'm listening to a file that plays on a standard 16-bit PCM DAC without any MQA processing anywhere in the chain. With MQA software decoding it's showing 24/96 (and appears to be a padded 16-bit/96 KHz raw stream). With MQA hardware decoding it's showing as 24/192.

    --

    There's a lot off jiggery-pokery going on, but this is not up/downsampling in any conventional sense, nor is it traditional compression. It's handling things very differently from existing encoding schemes, regardless of the net end result.

    --

    *I think it's extremely likely that all relevant commercial software players will implement MQA decode support. Free/open-source players will be constrained here depending on the precise licensing terms for software MQA decoders. And, for the moment, the only real source for MQA content already has software-decode support.
     
  5. weldp

    weldp New

    Joined:
    Jan 30, 2016
    Likes Received:
    5
    Trophy Points:
    3
    Location:
    Massachusetts
  6. weldp

    weldp New

    Joined:
    Jan 30, 2016
    Likes Received:
    5
    Trophy Points:
    3
    Location:
    Massachusetts
    Wouldn't MQA receive a license fee from the software only players too? And from the streaming providers/players like Tidal they might be getting nothing in order to open up the content market. From a startup sense I understand letting early references and market makers get a deal.

    -Chris
     
  7. trung225

    trung225 Facebook Friend

    Joined:
    Oct 22, 2015
    Likes Received:
    123
    Trophy Points:
    43
    I think Meridian still takes license fee from software player makers, but possibly a small amount
     
  8. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    While this has nothing to do with MQA just yet ...

    Initial listening to the Explorer 2 places it a bit ahead of the Dragonfly Red. The DFR is a little quieter (less hiss) with sensitive IEMs and seems to play louder for a similar volume setting (software controlled for both, even if implemented in hardware on the DAC/amp itself).

    Now the Explorer 2 won't work from my phone (iPhone 7) and the DFR will. So for minimalist travel I'll stick with the DFR, but for laptop duties I think the Meridian is going to be sticking around. It has a richer, more detailed and persuasive presentation. Extension seems similar. Articulation more nuance with the Meridian.

    But, if you had to pick one to do duty between laptop and phone, then only the AudioQuest DFR actually handles that.

    With, or without, MQA, the Meridian is really quite good and, at least with usefully efficient full-size cans (e.g. TH-X00) it's a nice pairing.

    Much more to come ... including MQA-related thoughts.
     
  9. weldp

    weldp New

    Joined:
    Jan 30, 2016
    Likes Received:
    5
    Trophy Points:
    3
    Location:
    Massachusetts
    Explanation of the relationship between software (Tidal) and hardware decoding. Explains the 4 decoding possibilities: no decoding, software decoding, hardware decoding, and a combined software/hardware decode.

    MQA Decoding Explained
    http://www.audiostream.com/content/mqa-decoding-explained

    One thing to note: if the original MQA file is 24/48, 24/96, or 24/88.2, it will pass through the software decoder and be 'unfolded' to its original resolution. This brings up the interesting fact that people who love their DAC and its proprietary digital filter may very well be able to have their cake and eat it too, especially when streaming MQA content from Tidal HiFi.



    -weldp
     
    Last edited: Jan 12, 2017
  10. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    That might be the way it's supposed to work, but it's not what's actually happening.

    If I play back a 24/192 MQA file with no decoding on a regular DAC it's showing 16/48 as the actual bit-rate. Maybe this varies with different files, but I've not seen a case of that so far. In fact I've not seen anything flag as 24-bit unless at least a software decoder is involved.

    When I get back to this tonight, and after I've finished with the Vega, I'll build a table with a selection of specific tracks that show what the DAC sees for undecoded, software decoded and hardware decoded playback.
     
    Last edited: Jan 12, 2017
  11. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    More thoughts ...

    Making any kind of informed comparative evaluation of MQA, regardless of decode path, is proving to be a pain in the arse. More than that, it's pretty much impossible right now, excepting via a very limited selection of tracks from the 2L hi-res test bench.

    Determining the provenance/identity of any specific master has always been a problem, outside a few niche "audiophile" labels. This hasn't gotten better with MQA. While MQA DACs and software decoders can indicate the provenance of the file as an "MQA master", that tells you absolutely nothing about which mix or master was actually used to prepare the MQA file - only that the MQA file is unmolested from when it was processed into MQA.

    It's not even possible to determine, properly, the cause if there is an audible difference between software and hardware MQA decoding when you DO have an MQA-enabled DAC. If you have an MQA DAC it will ALWAYS do part of the unfolding work regardless of whether a software player does a partial unfold first or just passes the data on still fully folded!

    You can see this readily in the TIDAL player. Turning on "Passthrough MQA" should disable the software decoder. Let's assume it does. Play an MQA file and it'll come through to your DAC and all the right lights will come on (MQA authentication indicator ... i.e. the "blue light" and the proper sample rate indicators). Turn off "Passthrough MQA" and now the software decoder should be doing it's thing. Play the same file, you'll get all the same indicators on the DAC.

    Now, this is clearly by design - and for actual usage one would want to get the maximum possible quality out of the chain, so it's not problematic except for trying to evaluate what is going on where.

    So the kinds of comparisons we can currently perform are enormously limited:
    • Using a non-MQA-enabled DAC you can compare un-decoded vs. software decoded output for MQA source files.
    • Using an MQA-enabled DAC you can compared mixed-mode decoding vs. pure DAC decoding.
    • You could compare any MQA DAC with any non-MQA DAC and use regular files to evaluate the RAW DAC's performance.
    But:
    • You cannot compare un-decoded and decoded MQA source files on an MQA enabled DAC.
    • You cannot reliably compare MQA and non-MQA versions on any DAC unless you can find reliable information about the master used and be sure they're identical.
    You either wind up in a situation where your comparisons have to be with a different DAC which is not very a useful comparative evaluation when it comes to determining why you hear a difference and where it originates, since the DAC is obviously a major source of such potential differences. Or you have to find encoded and non-encoded sources, which is fraught with issues.

    We can, of course, see if MQA files sound better than the masters we have available for the same albums/tracks. We just cannot make any reliable statements about why that might be:
    • It could be a different raw source master or mix.
    • It could be the same master and the difference is just in the higher resolution.
    • It could be the same master and the difference is due to MQA's temporal de-blurring.
    • It could be a combination of any of these.
    Unfortunately all of these factors can affect a non-MQA DAC as well as an MQA-enabled DAC. Which, fundamentally, means that right now, even if the whole world suddenly agrees that "MQA files sound universally better", we can't reliably know why.

    I find that quite irritating and a bit suspicious.
     
    Last edited: Jan 12, 2017
  12. yotacowboy

    yotacowboy McRibs Kind of Guy

    Pyrate Contributor
    Joined:
    Feb 23, 2016
    Likes Received:
    10,682
    Trophy Points:
    113
    Location:
    NOVA
    Home Page:
    I think the audiostream link might have a typo - as I understand it MQA takes a 24 bit FLAC and "hides" the top 8 bits under the 16/48 - so, MQA file with no software decoding through a non-MQA DAC will be 16/48:

    http://www.stereophile.com/content/mqa-questions-and-answers-udio-origami-or-folding-questions
     
  13. lm4der

    lm4der A very good sport - Friend

    Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    1,461
    Trophy Points:
    113
    Location:
    Seattle, WA
    Big thanks to everyone contributing info about MQA.

    So, it strikes me that the one thing, maybe the only thing, that is interesting about MQA is that it acknowledges the 'master' as part of the encoding chain. Well, maybe this is two things, as there is some emphasis placed on the mastering and also on improving the AD/DA process from the recording to the playback device.

    So this makes me wonder, perhaps naively, if that means we can hope that the industry will start paying attention to mastering with actual sound quality in mind? I'm not ignoring the potential for licensing and DRM abuse, but wondering if this particular aspect could give the industry a push towards better masters.
     
  14. Clemmaster

    Clemmaster Friend

    Pyrate Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    3,268
    Trophy Points:
    113
    I tested with my Geek Out v2. The MQA albums show "MASTER" in lieu of "HIFI".
    The DAC must be set to Exclusive Mode (in the device's settings: TIDAL > Settings > Streaming) to ensure bit perfectness (windows).

    I confirm that MQA titles play @ 96kHz on the GOv2. Switching to regular HIFI titles goes back to 44.1kHz.

    Those with MQA compatible devices must select "Passthrough MQA" in the device's settings.
     
    Last edited: Jan 12, 2017
  15. Clemmaster

    Clemmaster Friend

    Pyrate Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    3,268
    Trophy Points:
    113
    The software has an annoying bug where the volume will go down to 0 when switching tracks.

    The slider also moves down a bit after you set the volume. Weird.

    Not a problem if you max out the volume (or disable it). But with the GOv2, everything that touches the volumes makes my heart beat very strong...
     
  16. Clemmaster

    Clemmaster Friend

    Pyrate Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    3,268
    Trophy Points:
    113
    The file, non-decoded, is sent as a 24/48 stream. The bottom 8 bits (maybe less) are still by the DAC (if resolution allows) in the form of noise.

    Auro 3D uses a similar encoding scheme to embed their "Height" channels in the regular 5.1 or 7.1 PCM stream.
    The metadata (embedded in the form of noise in the lowest bits) is used to reconstruct the missing part (top channels) from the "base PCM" stream.
     
  17. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    This isn't actually necessary - though it's advisable.

    If you leave the software MQA decoder enabled the player will do the first level of unfolding and the MQA-enabled DAC will do the subsequent level (and any MQA-specific filtering). Even with Passthrough disabled, the MQA-DAC will show the correct sample rates and authentication status (blue light).
     
  18. yotacowboy

    yotacowboy McRibs Kind of Guy

    Pyrate Contributor
    Joined:
    Feb 23, 2016
    Likes Received:
    10,682
    Trophy Points:
    113
    Location:
    NOVA
    Home Page:
    Yeah - I meant that running an MQA file sans software and hardware reconfiguratory monkeymotion business nets a 16bit 48kHz file, not a 24bit 48kHz file per the audiostream chartographic thing.
     
  19. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Some initial thoughts on how "MQA" sounds ...

    I have "MQA" enquoted here since it is, as I commented above, not deterministically possible to know if it is actually anything about "MQA" that is having an effect or if it's all just down to the master that was used and things would sound the same way if we had a raw PCM copy of that.

    With that out of the way ...

    There is a very readily audible, and essentially consistent, difference between MQA tracks and any version I have available of them in non-MQA format. This is true regardless of whether those files are sent directly to a non-MQA DAC with a no MQA software decoding, or if they're sent to a fully MQA-enabled DAC. Additionally I find a specific character or signature to MQA-encode tracks that is, again, quite consistent even just assessing the result on it's own (i.e. without comparison to a non-MQA version).

    That signature is of increased brightness, a sense of increased detail, an apparent increase in the audibility of low-level sounds and the perception of faster transient response. On top of this, I am finding a generally faster onset of listening fatigue than I'm used to with the music in question.

    This signature is observed with these files regardless of whether they're sent directly to Yggdrasil with no MQA-decoding in the chain, or to a Meridian Explorer 2 operating in full-blown hardware MQA decode mode.

    Now, when I italicized "sense" of increased detail it's because, without access to a known non-MQA master, I don't know if I'm just hearing the additional brightness and am perceiving it as "increased detail" - which is a common thing and something I find often with lesser ESS implementations. I am immediately aware of this brightness with any MQA track I've tried, including a randomized mix of MQA and non-MQA tracks listened to blind. If I had to liken it to something else, non-audio-related, I'd say it's a bit like a slightly over-sharpened digital image.

    The "apparent" ease with which low-level sounds are discernible could simply be the result of dynamic range compression. It's more apparent with a non-MQA DAC and no-decoding than it is otherwise, which adds to my suspicion that DRC is the cause (or partial cause) of that effect. And of course DRC is something that increases the perception of quality vs. actually improving it. Again, no way to be clear if this is better reproduction, a different mix, or just an artifact of the way MQA works, deliberate or otherwise.

    The "perception" of faster transient response is, for now, just that, a perception. This could also simply be an artifact of the increased brightness. Or maybe the brightness is simply the net product of actual changes in the other areas I mentioned.

    Superficially these changes result in an immediately "impressively clear"* sound. Those that are used to listening to things critically will rapidly dissect this into other factors that can/do and/or are resulting in a perception of "clarity" that may, or may not, also be real. But it is easy to understand why the uninitiated would find MQA to sound very good indeed. That's not to imply that it doesn't sound good ... but as with many things first-appearances rarely tell the whole story. Also, there's enough immediate difference for it to be A/B'able to an inexperienced listener.

    Better than what we have? You tell me.

    And even if I thought it was, I could NOT make a statement there that would stand up as to whether it was due to MQA.

    As it happens I'm actually finding it a bit harsh and more rapidly fatiguing than my normal sources. That's a problem from an enjoyment and immersion perspective. It could be that there's enough additional information being retrieved and replayed, or that the natural shift in a better reproduction just happens to be increased brightness, that it's causing some level of sensory overload. CD did this to me when it was a new format. Of course, back then it was down to things like broken pre-emphasis flag handling and excessively bright and grainy implementation, which might be the case here.

    I don't know which is "closer to the Master" at this point, but so far the non-MQA tracks are easier to listen to and enjoy - once beyond any superficial "fireworks".

    ...

    Next step will be to sit with the 2NL reference tracks and try and do some proper back-to-back, like-for-like comparisons with as many potential variables removed as possible.

    --

    *Random fact: I find that any non-audio-geek that has ever heard one of my systems makes an entirely predictable first statement: "It sounds really clear!" Which I take to mean "I don't know what to say, but I'm supposed to say something!"
     
  20. Logan Ross

    Logan Ross New

    Joined:
    Dec 27, 2016
    Likes Received:
    3
    Trophy Points:
    3
    Location:
    Seattle Area
    I am having a serious "I think I like it" / "I am not sure I like it" moment with MQA so far. At first blush, it sounds different, which could be better (or not). I can turn DAC MQA (4) off or on, and honestly I am having a hard time distinguishing the Tidal Masters tracks with DAC MQA turned on (1-4) from the Tidal Masters track with the DAC's MQA decoder turned off (1-3).

    After volume leveling, and not knowing which DAC I was auditioning while AB'ing (for an extensive listen), I ended up (blindly) selecting the non-MQA DAC as my preferred listening chain.

    QUERY: If a lot of the change in the sound we are hearing is really in the base file, and we decide that we prefer non-MQA, does that mean that we are forced into MQA (Master) instead of being able to select plain old Hifi? I am not sure.

    One thing is for sure. I think it will take a lot more access to music I am familiar with to help decide how much I enjoy it.
     

Share This Page