In the patent application for MQA from 2013 there is a diagram that shows the compression, noise shaping, and evidently about 17 bits of real resolution.
This is also referenced in this article by Benchmark Media.
Thanks - I have seen them both before.
I guess the only thing you can really compare it to it a single rate 24 bit FLAC - ie the equivalent size container (lossless 24 bit @ 44.1 or 48k) vs an MQA encode of 96k/24bit to the same container and subsequent full decode (or full decode and render).
It may be more interesting to see a technical measurement of such a comparison, but what should be measured?
In the end, Tidal have some music that Quboz doesn’t have and visa versa, so I live with both fully rendered MQA from tidal and h-res from Qobuz and I don’t obviously notice consistent differences that favour one or the other. Production quality seems to trump the CD vs Hi-res vs MQA debates.
While that is fine for streaming, when buying, then naturally I want to highest resolution I can get just because our knowledge tends to improve over time. Next year maybe another great idea comes along that is worth paying some attension to.
This is same reason I made do with 320K mp3 on the go in the past, but only bought digital as CD quality if I could, so later I could use those formats directly as and when mobile technology allowed for it. These days I can stream CD quality all over the place and if I really need to, my phone has enough storage for a load of CDs as FLACs. So mp3 is finally dying a death.
TBH - I think MQA would better serve the world if it were a format that users could encode themselves as with mp3 which of cournce means it would have to scale back on its ambitions a bit and maybe be a bit more focussed in the use cases it targets. But, then it would be harder to monetize the technology at a time when network bandwidth while mobile may soon be taking another leap forward making any kind of compression irrelevent for many places.
That seems like a very fine idea indeed! Let users encode MQA themselves, like Opus or MP3, or FLAC. It takes all the problems out of it. Then it becomes a format just like any other format, with pros and cons, that you are free to encode or transcode as you see fit.
The problem is that MQA wants the end to end licensing and the money generated from it. Free to transcode cuts off one of the $ sources.
Just like DSD in SACD, the way MQA is going, I would guess, they will eventually offer a paid or free software MQA encoder/decoder to the public, if they are desperately want to keep the format alive😄
First MQA always exist in 24/44.1/48k in any lossless containers. The decoded part is always 88.2/96k with a reduced resolution of 17 bit. Of course the output ‘data’ can be 24 bit but don’t confuse it is ‘24 bit resolution’.
All MQA DACs will always display the authenticated sample rates, Not the actual sample rates that goes to the DAC chip. So many are tricked into thinking that the actual samples are indeed higher than 88.2/96k (i.e, 192k or even 384k), which is of course not true. There’s no further details extraction after MQA get decoded. It is basically doing ‘rendering’ or so call conditioning with final MQA filters. Of course, the selection of MQA filters, 32 of them, required the decoder output to preserve the ‘render info’ before applying any DSP, and put back so the renderer will know which correct MQA filters (32 of them) need to turn on. That’s it so simple😁!
The MQA has little to do with higher and higher sample rates. It’s about an analog sound free from pre ringing and timing errors.
High resolution in the analog domain, ie what you can hear and discern with your ears and mind.
2 Likes
Actually DSD in SACD already solved this problem 2 decades ago. The impulse response is anything but truly analog in nature, truly analog sounding.
Odd - those all look like plots of linear phase DSP filters, even the ‘analog’ rather then a true illustration of the capabilities of the various mediums.
The DSD plot is a little odd with more post ring than pre-ring - I guess it was trying to be an apodizing(-ish) filter.
Anyone know their way around analog and DSP filters enough to comment?
Maybe the measuring instrument is adding some pre-ring as there does appear to be some correction between ‘analog’ pre-ring and DSD pre-ring.
Either way, I guess its decent enough illustration of the higher temporal resolution of DSD, but what you hear is still going to depend on the chosen filter.
Recently, I moved my Core to a new machine. I restored the Roon library, but forgot to point to a new local drive.
Later, I wanted to listen to Joan Osbourne’s “Relish”. It was unbearable. Joan’s wailing, cat-in-heat voice was now un-listenable.
After some investigation, I realized that instead of my local copy I was getting the MQA version from Tidal.
Because I didn’t realize I had omitted my libraries, I had no expectation bias as to whether MQA was going to be different.
So MQA doesn’t improve everything, if anything.
DSD does not use any form of sharp cut off digital filter like PCM do. That’s why the transient response and peak amplitude is virtually identical to analog ‘click’. The response time is also fast, we talking 3 micro seconds!
MQA is trying to improve the transient response by using weak and leaky filters. Unlike DSD which uses high sample and noise shaping, there is no imagining in the audio range, i.e, you don’t get distortion. If you do this in PCM, this is going to be problem, weak and leaky filters will cause imaging and distortion gets refected into the audio range. Many perceived this as ‘analog sounding’ but in fact this is actually distortion caused by imaging from bad implementation of filters.
I assume that you have been listening to the MQA version of the track without the full MQA unfold - were you using the first Roon unfold?
I have just listened to this track on my second system (Sonore microRendu/Mytek Brooklyn+) and certainly cannot replicate your findings. I found MQA and standard 16bit/44.1 ‘versions’ of the album on Tidal, and found that the two versions in this instance sound very similar indeed.
Unlike your description of the two versions, I actually found the MQA version to sound very subtly better. However, the difference was so small that I almost certainly would not have noticed a difference at all had I not been concentrating on listening for one. Not an outstanding result or glowing praise for MQA (in the case of this album) I grant you, but I certainly do not recognize your description at all in my own listening tests.
What equipment were you using to listen to the album?
Aha, of course.
You are correct. I have MQA completely disabled.
On the other hand, I have several other Tidal albums that are available only in MQA and they aren’t effected, at least as far as I can discern, in the way that “Relish” was.
I guess not a legitimate criticism of MQA, but rather of Tidal’s MQA only philosophy on some albums.
Dunno.
Here’s the more analog like wave form after final MQA rendering/usampling in a MQA Dac.
Later confirmed by John Atkinson.
As expected the waveforms from the renderer is quite ‘distorted’ If this kind of distortion produces the typical signature of MQA sounding, we will probably not listening to the actual PCM masters. If this is the kind of ‘preference sound’ to put it into the masses, we are in serious trouble 
I had the same reaction and deleted all the MQA albums.
Then I decided to try QOBUZ and their “HiRes Audio” masters - and I am glad I did! My Audio system never sounded better to my ears, with a presence rivaling being in a concert hall (I usually listen to classical music, but the jazz albums were equally awesome).
As a result, I cancelled my TIDAL subscription and I am now listening to Qobuz through Roon.
In short - Audio Nirvana!
3 Likes
Here the thing, Qobuz Hi-Res is actually derived from the final stereo mixed down masters; so you get unaudultrated sound quality. With Qobuz Hi-Res, you can add DSP and tune to your liking.
MQA version will have go through another process done by MQA Ltd, this of course alter the sound signature (Think of another DSP). Some people like it some don’t.
I prefer the unaudultrated version, it sounds more livier, more clarity and better soundstaging. This is I called truly unaudultrated Masters.
I have absolutely no issue or quarrel with anyone who states that they prefer Qobuz (or other) hi-res albums to equivalent (same Master) MQA encoded albums on Tidal - either from a sound quality perspective or from a technical perspective. This probably (just) correlates with my own listening experiences of MQA Master albums on Tidal.
However, I simply cannot understand posts which state that “all MQA albums are unlistenable” or that “MQA encoded albums sound worse than 16bit/44.1 versions” or even some who postulate that “MQA sounds ‘worse’ than MP3”. Irrespective of what technical ‘evidence’ is published (or regurgitated) on this forum, these views simply do not correlate with my own by now quite extensive listening experience.
3 Likes
Actually MQA albums can sound worse depending what dac or upsampling filter you use. Chord dacs work better with either cd or hires vs MQA unfolded. Linear filters in HQPlayer sound worse with MQA than with original cd quality.
I ended up switching back to Qobuz again. I was using Tidal but avoiding MQA. First thing I noticed with Qobuz was the depth of soundstage when listening to new Tool album. Qobuz has many hires albums that aren’t even available in MQA.
1 Like
To complement your observation, I read this about the Hegel H590 in a SoundStage! Hi-Fi review:
“This is Hegel’s first DAC to natively decode and render MQA. The lazy way to implement MQA would have been to route all incoming signals, including PCM and DSD, through the MQA engine. But because doing so is suboptimal for non-MQA content, Hegel’s new DAC has separate signal paths for MQA and non-MQA signals. Furthermore, unlike Hegel’s previous DACs, the H590 does not upsample the incoming signal.”