Jump to content
Computer Audiophile
  • The Computer Audiophile

    MQA (for civilians)

    At CES 2017 Tidal announced it was streaming MQA masters and MQA Ltd announced software decoding of the MQA signal. Two big items for all of us who enjoy music. Immediately the questions and conjecture started flowing. It's human nature. We ask questions and make guesses about what's happening, when we don't have all the information.

     

    Shortly after the announcements I setup a meeting with MQA's Bob Stuart to get more details about decoding MQA signals. I wanted to know the differences between software and hardware decoding and where rendering comes into play, in addition to many other items.

     

    A PhD isn't required to enjoy MQA. This article is my attempt at explaining how decoding and rendering work, from a civilian perspective. Most of us have seen the music origami graphs and deep technical explanations, but have no idea what any of the information actually means for us, enjoying music at home or on the go. I want to help members of the CA community understand how to get the best sound quality out of MQA.

     

     

     

    From The Distribution File Forward

     

    Currently MQA music is offered through online stores for purchase and download, and through Tidal for streaming. I'm willing to bet more music will be available through both channels and both channels will have more outlets in the coming months.

     

    Consumers purchasing or streaming MQA music will see either 24 bit / 44.1 kHz or 24 bit / 48 kHz files without playing the audio (16 bit MQA files are outside the scope of this discussion). These are what's called the distribution files. They have been through the MQA process that deblurs and folds them into a smaller package, readying them for transport and playback on almost any device.

     

    The MQA distribution file, the file that's actually purchased or streamed, is like a chameleon. In its packaged state the files are 44.1 or 48 kHz, but decoded and rendered the files can expand into the highest supported sample rate of the digital to analog converter inside the DAC..

     

     

    Real world example:

     

    1. The studio creates a track at 24 bit / 352.8 kHz DXD.

    2. The studio uses the MQA process on the track, packaging it as 24 bit / 44.1 kHz.

    3. The consumer purchases or streams the 24 bit / 44.1 kHz track.

    4. The consumer's playback system decodes and renders the track at 24 bit / 352.8 kHz DXD.

     

     

     

    Squeezing The Best Quality From MQA Music

     

    With the aforementioned real world example in mind, let's look at how to play MQA music and how to get the best sound quality possible. There are four "ways" to play MQA music. I use the word "ways" for lack of a better, more specific term.

     

     

    A. No decoder

    B. Software / Core Decoding

    C. Software / Core Decoding with Hardware Rendering

    D. Hardware Full Decoding

     

     

     

    No Decoder

     

    Similar to a dual layer SACD that plays the CD layer in a standard CD player and the Super Audio layer in an SACD player, MQA music is playable through almost any playback system, but the highest quality is only possible with the appropriate solution.

     

    Playing MQA on a system without a decoder, will enable the consumer to hear the 24 bit / 44.1 kHz (or 24 bit / 48 kHz) version of the music in the example above. According to MQA Ltd, playing the un-decoded version still enables the consumer to benefit from the deblurring processes used in the creation or folding of the track.

     

    Examples of systems without decoders are plentiful in this early phase of record labels rolling out MQA music. JRiver Media Center, Amarra, HQPlayer and many others are applications that don't decode MQA. In addition, most hardware on Earth doesn't decode MQA at this time.

     

    One scenario that may confuse consumers, is when an MQA renderer is present without a software or hardware decoder. This will result in an un-decoded signal exactly as it would without the MQA renderer. The 44.1 or 48 kHz version of the file will play, undecoded. One example of this is the upcoming AudioQuest DragonFly (updated Red and Black versions). Without a decoder in the playback chain, an MQA renderer has no effect on the audio.

     

    No-decoder.png

     

     

     

    Software / Core Decoding

     

     

    MQA is a whole host of processes and technologies, but for purposes of this civilian discussion, let's look at it as three processes. MQA files can be 1. Fully decoded, 2. Software / core decoded, and 3. Rendered. Software decoding is capable of exactly what its name suggests, decoding MQA. Rendering must be done in hardware because it is custom matched to the DAC system.

     

    Software decoding, what MQA Ltd calls core decoding, provides what I consider to be about 90% of the MQA benefits. Decoding in software unfolds / unpacks the music to a maximum of twice the base sample rate, 88.2 or 96, for either analog or digital output.

     

    EQ, bass management, and other non-MQA DSP can take place after core decoding.

     

    Using the real world example above, the Tidal desktop application, Audirvana, and soon Roon would decode the MQA 24/44.1 distribution file and unpack it to 24/88.2. This can be output digitally to any DAC, digitally to an MQA DAC for rendering, or output as analog audio.

     

    Another example can be seen when streaming Beyonce's album Lemonade. The MQA distribution file is packed to 24/44.1 and the decoded file is also 24/44.1. The album must have been recorded at 24/44.1 and the studio is being honest with us, rather than upsampling it to 88.2 or higher.

     

    When a master is 44.1 or 48 kHz, the core decoder Authenticates, decodes full dynamic range and matches to the current PC playback settings. (Depending on the soundcard and audio configuration, the Tidal App may decode this example to 44.1k or a provide a compatible 88.2k output for smoother playlisting). If you select Passthough, the raw 44.1/24b MQA file is passed downstream to a decoder. For music where the original sample rate is 88.2k or higher, the core output is always either 88.2 or 96kHz.

     

    Note that other Apps and products implementing Tidal may be subtly different.

     

    core-decoder.png

     

     

     

     

     

    Software / Core Decoding with Hardware Rendering

     

     

    The third way to play MQA music is through a software decoder and a hardware renderer. As you read above, MQA has three process required for the full MQA experience, 1. Full Decoding, 2. Software / core decoding, and 3. Rendering. In this method of playback, a combination of software and hardware is used to deliver all that MQA has to offer. Don't ever use this as the answer to an MQA exam question, but you can think of it this way - software / core decoding serves up the file and hardware rendering hits it out of the park.

     

    Everyone looking to get the best sound from MQA music will want to use this method or the all hardware method discussed last. In this method, the core decoded MQA file is passed from a software application to the MQA hardware renderer.

     

    Using the real world example above, the Tidal desktop application, Audirvana, and soon Roon would decode the MQA 24/44.1 distribution file and unpack it to 24/88.2. This file is output from a computer via USB or S/PDIF or even a phone via Lightning or USB on-the-go, to the hardware renderer. For this example, we'll output via USB to an AudioQuest DragonFly. The core decoded file enters the DragonFly at 24/88.2, then expands to the full 24/352.8 kHz resolution of the original studio master file.

     

    Readers familiar with the DragonFly will know that the DragonFly supports audio up through 24/96. However, that's only on its USB interface. Internally the DAC goes up through 768 kHz. MQA enables the audio to duck its head to get under the door frame, before standing straight up once again. Kind of like a balloon as well. Squeeze the middle of a long balloon and the two ends will get larger while the middle shrinks. The two ends are the studio master file and the fully decoded MQA file, while the middle is the packed undecoded MQA file.

     

    The above method is a really good way to work around the lack of USB Audio Class 2 driver support in many Windows operating systems and to get around interface sample rate limitations. It's possible to play 24/352.8 on a class 1 device and without custom drivers.

     

    What happens when using software / core decoding and hardware that's capable of full decoding like the Meridian Explorer2? If desired, it's possible to use an app like Tidal to do the core decoding and send the MQA signal to the DAC for rendering only. If the Explorer2 is fed with an MQA core (decoded) signal, it only does the rendering.

     

    Note about renderers: There are no generic MQA renderers, as each one is custom designed for each piece of hardware. According to MQA Ltd, the analog output is custom tuned for each device to most closely recreate the sound heard in the studio. As always, you'll have to be the judge to see if the marketing matches the end result.

     

    One additional piece of information that fits somewhere between this section and the next, systems like Meridian that run digital to the loudspeakers, send a core decoded stream to the speakers before final rendering separately for each drive unit. This core decoding takes place in hardware / software loaded on Meridian hardware.

     

    core-decoder.png

    render.png

     

     

     

     

    Full Decoding (Hardware Only)

     

     

    There's not much more to say about this one. Full decoding is only possible in hardware and it's considered the full monty. Both aspects of core decoding and rendering are controlled by a single manufacturer and the requirements for third party software are gone. The final analog output however, is theoretically identical to a software / core decode and hardware render. We'll have to see once more opinions come in from people testing both methods.

     

    A DAC or home theater processor capable of full decoding can receive an untouched MQA file (distribution file or stream) or a core decoded file from a software decoder, for rendering only. I'm not sure why anyone would prefer to use a software decoder when s/he has a full decoder in hardware, but it's entirely possible.

     

    All details about the renderer hold true for a full decoder. Very custom and tuned to each hardware device.

     

    A note about this tuning for each DAC. I've talked to many manufacturers who have products in the process of MQA certification. Every one of them says the process is thorough and a bit demanding, often requiring many updates to hardware and firmware until everything is as close to the target as possible. This piece isn't marketing, it's all based on engineering.

     

    Using the above real world example, any software or hardware capable of sending a bit perfect audio stream to an MQA DAC with a full decoder, will work just fine. The DAC must see either a core decoded stream or undecoded stream without alteration.

     

    full-decoder.png

     

     

    Let There Be Light

     

     

    Whether playing MQA content fully decoded, core decoded and rendered, or just core decoded, an MQA DAC or application will always signal the file is true to what the record label released. If it's MQA it will illuminate a blue/green light.

     

    When I first investigated MQA I knew the blue authentication light on MQA DACs was helpful in identifying bit perfect output to the DAC, but I thought the whole authentication piece was unneeded. However, I've since learned that the music supply chain is full of challenges and less than stellar versions of our favorite albums (some related to up/down sampling). Much of this isn't malicious, it's just a matter of large companies with many people involved who may not know exactly what's going on. MQA has the ability to provide record labels with a single deliverable file and the consumer can verify this is the file s/he plays back at home. It's a cool concept.

     

    Conclusion

     

     

    Enjoying MQA music isn't rocket science, but it takes a little education to make the right choices. Obtaining the best quality MQA playback requires either a combination of software decoding and hardware rendering or a full decoding DAC. Fortunately, I believe 90% of MQA's benefits can be realized by only using a software decoder, called core decoding. Now that some of our favorite music is available in MQA, it's time we listen for a while rather than talk over the music. Set your systems up right and press play.

     

     

     

     

     

    1-Pixel.png

    Edited by The Computer Audiophile

    Sign in to follow this  
    Sign in to follow this  

    User Feedback




    When I think about lossy, I ask myself what is lost. With MP3 and AAC real music is lost. With MQA I don't believe real music is lost. MQA changes digital, making existing terminology require more discussion than in the past.

    Lossy means you lost information... :)

     

    Does it sound better than the original file? In many cases it does seem to be the case - could be MQA itself or a careful remastering (I bet you it is the latter). Regadless, I enjoy some albums more in their MeQA version than otherwise. And this is with software decoding alone.

    Share this comment


    Link to comment
    Share on other sites
    True. The same is true with upsampling: most of the time we are listening to an upsampled version of the original file and we like it. My point is MQA per-se is not a better solution than delivering the high res file itself. Whether the difference can be heard, or whether the result of hardware decoding compensates for shortcoming in cheaper DAC chips are different questions.

     

     

    Many MQA titles I've listened to sound better than other versions I have, including the ones billed as "high resolution" (eg Joni Mitchell's 'Blue' or 'Both Sides Now'). I frankly don't care what shenanigans MQA is coming up, I like the result, and as such I will be a customer.

     

    I look at it similarly. I'm a customer. If a product is offered that I like, I'll buy it. If it's something I must have but I can't have exactly what I want, I'll select the best option and get on with my life.

    Share this comment


    Link to comment
    Share on other sites
    Lossy means you lost information... :)

     

    Does it sound better than the original file? In many cases it does seem to be the case - could be MQA itself or a careful remastering (I bet you it is the latter). Regadless, I enjoy some albums more in their MeQA version than otherwise. And this is with software decoding alone.

     

    I can't argue with your definition :~)

     

    It may be helpful for me to start a different discussion about "old" lossy versus MQA lossy.

    Share this comment


    Link to comment
    Share on other sites
    When I think about lossy, I ask myself what is lost. With MP3 and AAC real music is lost. With MQA I don't believe real music is lost. MQA changes digital, making existing terminology require more discussion than in the past.

     

    This seems to be a real stumbling block for many (i.e., lossy=bad therefore ...). As you say, it's not the same as lossy MP3's etc.

    Share this comment


    Link to comment
    Share on other sites
    I can't argue with your definition :~)

     

    It may be helpful for me to start a different discussion about "old" lossy versus MQA lossy.

    That's like walking into a Louisiana town with a Bernie Sanders banner. You're on your own, man!

    Share this comment


    Link to comment
    Share on other sites
    This seems to be a real stumbling block for many (i.e., lossy=bad therefore ...). As you say, it's not the same as lossy MP3's etc.

     

    Lossy is by definition not as good as lossless. Sometimes the loss is audible, sometimes it isn't. Even mp3 can be very hard to distinguish from lossless in many cases. The trouble is with those cases where there is an audible difference. No matter how many tracks you listen to without noticing the loss, the next one you play may be the one that sounds terrible. Is that a risk you're willing to take?

    Share this comment


    Link to comment
    Share on other sites
    Lossy is by definition not as good as lossless. Sometimes the loss is audible, sometimes it isn't. Even mp3 can be very hard to distinguish from lossless in many cases. The trouble is with those cases where there is an audible difference. No matter how many tracks you listen to without noticing the loss, the next one you play may be the one that sounds terrible. Is that a risk you're willing to take?

     

    Well, I'm willing to actually listen to it for myself, hopefully without prejudice, just as I did with other lossy compression schemes. IMO there is still enough regular old PCM out there to last me a lifetime if MQA is indeed BS.

    Share this comment


    Link to comment
    Share on other sites

    I don't stream and probably won't. I also don't think audiophiles are gonna drive the market. Pono has failed. DSD isn't popular. All the publications and web sites rave about the latest technology and then it seems to go nowhere. We'll see with Tidal but the company seems to need cash infusions all the time.

     

    I'll spend more money on discs this year then streaming and Roon etc. I'm not paying for a subscription for liner notes that should come with those expensive downloads. I got the Yes remix albums on BluRay for cheap money and they have tons of material 5.1 mixes etc.

    Share this comment


    Link to comment
    Share on other sites
    This was the most suspicious piece of the whole writeup in my opinion. In my testing, defeating the software unfolding in the Tidal desktop app and playing the "direct pass through" on my Musical Fidelity M6si's USB input (which has never heard of MQA) sounded noticeably worse than the Red Book versions of the same tracks. Duller, flatter, less dynamic and with a narrower stereo image. Software unfolding is a must to make these files sound passable.

     

    It would also be great to get some clarity around Auralic's plans concerning software unfolding on its Aries line - their Facebook page made some mention of a forthcoming "proprietary upsampling" algorithm, but it sounds like MQA unfolding is more than a simple up sample.

     

     

     

     

     

    Sent from my iPad using Computer Audiophile

    Share this comment


    Link to comment
    Share on other sites
    Lossy is by definition not as good as lossless.

     

    Absolutely incorrect. You're taking an objective standard, lossy, and applying subjective commentary and suggesting it's the definition.

     

    Lossy can be equally as good. I'm not saying please give me the lossy versions of music, but I'm saying you shouldn't promote your view as fact and "by definition."

     

    When you say not as good, do you mean subjective sound quality? Do you mean not as good because the goal is to make the largest file sizes? Please help.

    Share this comment


    Link to comment
    Share on other sites
    Absolutely incorrect. You're taking an objective standard, lossy, and applying subjective commentary and suggesting it's the definition.

    I think we would all agree that lossy compression "loses" something and as such something is missing. I think we would all agree that if MQA was lossless, we would not be arguing about this. Does MQA's lossy nature make for better sound? No, and there's a chance that we will find cases where the lossy feature can be heard.

     

    Now to the practical side of reality: MQA is here. Strictly lossless MQA is not. Would I like MQA to be lossless? Yes. Would I like MQA decoding to be open source? Yes. Would I like to have a different president? Yes. But that's not what we have so we will have to move forward.

     

    Do I like the sound of MQA? Yes.

    Share this comment


    Link to comment
    Share on other sites

    In my opinion, MQA is marketed as a package of de-blurring and compression because it appears to be a greater value than just the de-blurring alone would be. I think every DAC and audio sofware could probably come up with a de-blurring correction thingy. But the compression part of it seems too complex. However it's not really needed all the time. I'm guessing they won't sell the deblurring a-la-carte for those reasons.

    Share this comment


    Link to comment
    Share on other sites
    I think we would all agree that lossy compression "loses" something and as such something is missing. I think we would all agree that if MQA was lossless, we would not be arguing about this. Does MQA's lossy nature make for better sound? No, and there's a chance that we will find cases where the lossy feature can be heard.

     

    Now to the practical side of reality: MQA is here. Strictly lossless MQA is not. Would I like MQA to be lossless? Yes. Would I like MQA decoding to be open source? Yes. Would I like to have a different president? Yes. But that's not what we have so we will have to move forward.

     

    Do I like the sound of MQA? Yes.

     

    The thing is, it's new! If we could compare lossy-lossless to de-blurred without the lossy-lossless compression, there probably are differences but they'll never allow that. The demos have been carefully crafted thus far.

    Share this comment


    Link to comment
    Share on other sites
    The thing is, it's new! If we could compare lossy-lossless to de-blurred without the lossy-lossless compression, there probably are differences but they'll never allow that. The demos have been carefully crafted thus far.

    I listened to MQA at Meridian in NYC in March 2015. The MQA files in the demo sounded way better than the "standard" files. The people giving the demos were either completely ignorant or deceptive as they would not answer any of my questions. And I do not believe the MQA versions came from the same master. So really very deceitful if you ask me, frankly.

     

    But... Do I like the sound of many of the MQA albums on TIDAL? Yes. Are they better than even my so-called hi res versions of these albums? Almost across the board the answer is: Yes. Is it because of MQA, deblurring, origami, or some of Bob Stuart's hair interspersed in the file? I don't know and I frankly don't quite care. Call me a cynical pragmatist.

    Share this comment


    Link to comment
    Share on other sites
    I think we would all agree that lossy compression "loses" something and as such something is missing. I think we would all agree that if MQA was lossless, we would not be arguing about this. Does MQA's lossy nature make for better sound? No, and there's a chance that we will find cases where the lossy feature can be heard.

     

    Now to the practical side of reality: MQA is here. Strictly lossless MQA is not. Would I like MQA to be lossless? Yes. Would I like MQA decoding to be open source? Yes. Would I like to have a different president? Yes. But that's not what we have so we will have to move forward.

     

    Do I like the sound of MQA? Yes.

     

    Your comment made me think of an additional unrelated item. Many people with objective points of view only, have frequently pointed to Monty's claim that high resolution is actually detrimental to audio. Then some of these same people are angry that the ultra high resolution content of MQA files has been removed. Ahhhh. The madness.

     

    Let's just listen. As you say, MQA is here.

    Share this comment


    Link to comment
    Share on other sites

    Like it or not, MQA is as close as most people are ever going to get to mainstream musician music released by the big studios and it will likely be thru streaming.

    Share this comment


    Link to comment
    Share on other sites

    There are a couple of points that still seem unclear to me.

     

    1. With the music that has been released so far with MQA being described as "remastered" I wonder if people are reacting to the remaster itself or to the MQA process? I guess we will probably never know, I doubt if the record companies will release the remasters in any other format without MQA.

     

    2. As far as I can tell you never know what you are getting. If you had the software and hardware to do a full MQA decode of a file from Tidal how do you know what should be playing? Will it be 32/384, 24/96, etc.? It seems like this is being kept secret for some reason. If you buy a download from anyone like Apple, HD Tracks, Acoustic Sounds, you can order what you want and you know exactly what you are getting.

     

    I still am on the skeptic side of the fence on MQA and the more I read about it the more I think of P.T. Barnum.

    Share this comment


    Link to comment
    Share on other sites
    True. The same is true with upsampling: most of the time we are listening to an upsampled version of the original file and we like it. My point is MQA per-se is not a better solution than delivering the high res file itself. Whether the difference can be heard, or whether the result of hardware decoding compensates for shortcoming in cheaper DAC chips are different questions.

     

     

    Many MQA titles I've listened to sound better than other versions I have, including the ones billed as "high resolution" (eg Joni Mitchell's 'Blue' or 'Both Sides Now'). I frankly don't care what shenanigans MQA is coming up, I like the result, and as such I will be a customer.

     

    Hi Miguel -

     

    Blue sounded better to me in the 24/192 HDTracks hi res version than in the MQA version "unfolded" to 24/192 by software, both upsampled to DSD256.

     

    How were you listening to it?

     

     

    Sent from my iPhone using Computer Audiophile

    Share this comment


    Link to comment
    Share on other sites
    Blue sounded better to me in the 24/192 HDTracks hi res version than in the MQA version "unfolded" to 24/192 by software, both upsampled to DSD256.

     

     

    How were you listening to it?

    I have the HDTracks 24/192. That was playing through Roon+HQP (possibly some 1:1 filtering in HQP). For comparison purposes only, I captured the TIDAL software decoded MQA version (to a 24/96 file) and played it in Roon+HQP (HQP now upsampling to 24/192).

     

    Mitchell's voice had a presence and naturalness that was better than the 24/192 HDTracks file. My system is particularly adept at voice reproduction, I would say.

    Share this comment


    Link to comment
    Share on other sites
    I have the HDTracks 24/192. That was playing through Roon+HQP (possibly some 1:1 filtering in HQP). For comparison purposes only, I captured the TIDAL software decoded MQA version (to a 24/96 file) and played it in Roon+HQP (HQP now upsampling to 24/192).

     

    Mitchell's voice had a presence and naturalness that was better than the 24/192 HDTracks file. My system is particularly adept at voice reproduction, I would say.

     

    So definitely HQP filtering with the MQA version plus whatever filtering accompanied the downsampling to 24/96, versus possible 1:1 filtering with the non-MQA file.

     

     

    Sent from my iPhone using Computer Audiophile

    Share this comment


    Link to comment
    Share on other sites
    So definitely HQP filtering with the MQA version plus whatever filtering accompanied the downsampling to 24/96, versus possible 1:1 filtering with the non-MQA file.

    Correct.

    Share this comment


    Link to comment
    Share on other sites



    Create an account or sign in to comment

    You need to be a member in order to leave a comment

    Create an account

    Sign up for a new account in our community. It's easy!

    Register a new account

    Sign in

    Already have an account? Sign in here.

    Sign In Now

×