MQA software decoding in Roon

I hear Roon is coming out with the first unfold in the software like Audirvana. My trial will end soon with Roon, so I am thinking of not signing with them, Audirvana, if this is the case. I will stay with Roon since Roon does so much more and would not need them, Audirvana if you are a some time going to first unfold. Please advise.
PS I think I am on the right track/page.

Yes confirmed ! It is coming (Roon)

It seems to me Roon is taking ‘forever’ to come out MQA decoding…The only information I see in this blog is in ‘discussion in progress’. Meaning implementation, certification and testing is still a far away.

Not sure what secretive is going unless both Roon and MQA signed a NDA. MQA is around for more than 3 years now I think Roon should give us, the users a timetable when it is expected to roll out. People are tired of waiting…

I’m more interested in an improviement in the Internet Radio facility than MQA to be honest. I’m still enjoying listening to FLAC streams.The reports of MQA across forums has been very mixed so I’m not desperate to hear it. It’ll come when it comes.I can wait. Also, marketing wise, it’s not good business to push out all your updates asap. Bugs need to be kept to a minimum and carrots need to be spread out.

I understand there are many mixed reviews out there and I’m one of them…Though at first I’ve a negative outlook on the format, over time I’ve softened my stand through listening and more exposure to the format. As a Tidal Hi-Fi subscriber on Roon, it makes sense to start decoding as Tidal has already started streaming MQA at no additional charges.

And here’s why. The more we bug the Roon team about this kind of stuff the less they will share.


MQA started out being pitched as the new cake, then the new icing and In reality it seems that MQA’s real value-proposition is as a cherry on top of the particular cake and icing that you already love. Software decoding gets you 99%+ of that cherry (i.e. 24/96), without a costly spend to get it or losing the downstream benefits of room correction, DSP or filters already optimized for your digital front-end and system overall.

As has been said already, potentially the biggest benefit is to the streaming companies in reduced file size and therefore cost to stream hires audio files.


As an IT guy (software developer) I don’t understand how you suddenly can pack substantially more information into a lossless file-format. Either flac is a shitty implementation/specification, or something fishy is going on.

And there are supposed to be 2 unfolds, which means the audio information needed is 4 times more.

It’s not lossless - no one has ever claimed it was. It is lossless up to 20 kHz or so, then it only encodes actual content above (i.e., no noise or digital silence).

Also, it can encode normal Redbook content (i.e., no “folds”), 88.2 / 96 content (one “fold”), 176.4 / 192 content (two “folds”), or 352.8 / 384 content (four “folds”) - or higher, I believe, though I scarcely believe there would be any audible content at that bitrate.

Hi Magnus

This was originally discussed as far back as Dec 15


The simplest answer is that the classical specs are indeed horribly inefficient. It is a rectangular format: in a 24/192 file, it allocates the full 24-bit resolution all the way up to 192 kHz, and you certainly do not have any full-volume signal at 192 kHz, and you don’t have any meaningful signal in the least significant bit at 192 kHz either. By discarding these unused portions of the signal space, the transmission is lossy in a data transmission sense, i.e. you can’t recreate the signal including random noise in the bottom bit. But since these portions of the signal space are unused, discarding them can be lossless from an audio perspective.

The complete story is more complex, but this is the simplest answer to your specific question about signaling efficiency.


Cheers for the answer, I am not an audio-tech guy so I don’t know exactly how the audio is D/A and A/D converted. But the ‘192 khz’ is sample rate right, and not the maximum frequency sent to the amplifier? And 24 bit is the resolution of the amplitude for a frequency? Or is 192 khz both sample rate and maximum frequency?

1 Like

192k is the sample rate, by classical theory it can transmit half of that frequency, or 96k.

As discussed here and there, we want that not because we can hear that far but because of other improvements it gives to the sound, directly and indirectly.

But the rectangular data space was always a ridiculously inefficient way of transmitting that.

Although some people get hung up on the word “lossy”, equating it to the Antichrist.


It’s a good bet that MQA uses non-Shannon sampling (information from adjacent sampling periods convolved with a kernel, and reversible) which means that the information density per nominal sample is greater than that from conventional regular sampling.


Reading that made my head hurt! But it sounds fishy :slight_smile:

1 Like

Section 4.

1 Like

Thanks for (re)posting the link Joel. The Stuart/Craven MQA AES paper makes a bit more sense now than when I read it a year ago.

Do you have any predictions as to when Roon will be able to unpack via software the first part of MQA files?

when its done