MQA General Discussion

I should add that in the video Bob Stuart does not say that the resolution of MQA is 17 bit he is talking about something else at that point. If you think that listen again.

Thank you for the reference - I remember reading 21 bits, but the claim here is over 23 bits; i.e., virtually 24 bits.

Well if you listen and watch the very first part before it was cut, the interviewer was asking that the dynamic range of the ‘Air’ in comparison was roughly -80dB, Bob Stuart commented ‘17 bit’, he even commented that the full range of ‘24 bit’ cannot be utilized, not to mention 32 bit is a waste of space.

There are many instances that MQA is compared to the ‘Air’ response. So my question is MQA really 17 bit resolution but based on what Bob Stuart, I’m getting the impression it is indeed so.

Exactly he is talking about noise in air not the technical resolution of MQA. You can see from his responses in Stereophile that MQA has 24 bit resolution. In the video as part of the same answer he also states that more steps (i.e. Bits) are good. Obviously for other reasons than to capture the dymanic range of air which only needs 17 bits.

To me it seems that when Bob S talkes about bits he does not mean “actual bits” he means “equivalent undithered bits”. I.e. in the same way as there is lossless and BS-lossless.

So -80 dB should be roughly 13 actual bits but then with dithering/noise-shaping this get translated into 17 BS-bits.

Looks like nobody knows what he is talking about.
I didnt hear anything on MQA that allowed me to understand the principles behind it.
Nothing really clear.
Worst marketing strategy I have seen in a long time.

I canceled Tidal.
Enough.

He corrects the interviewer and says it is spectrally down 120 dBA not 80 and that is 17 bits.

My comment was not directed at the interview, which I have not watched, rather a general observation.

If you know how low the noise floor is, basically you know the overall dynamic range, which is again translated the actual resolution of the recording. The actual output can be ‘24 bit data’ but it doesn’t fully utilise the actual 24 bit resolution.

MQA uses the last 6 to 8 bit in ‘24 bit data’ in order to code some of frequencies above 22.5k. So when one playback MQA, it will always display ‘24 bit’ in playback software but it will not deliver 24 or 20 but around 17 bit of resolution, which is slightly better than Redbook.

His argument is no recording can truly achieve 24 bit resolution therefore MQA takes this advantage and utilise a few upper bits of the ‘24 bit data’ to code the some frequencies above audio band. MQA trade off resolution to give you some high frequency contents in return it gives you smaller file size.

1 Like

Have you read the Stereophile article I linked above? In there it says that unsubstantiated assertions that MQA is limited to 17 bits along with a number of other things) arise from over literal interpretations of our origami diagrams, misunderstandings of information theory and general lack of appreciation of the novelty of in our advanced data burying methods.

Still hoping Roon has not deprioritized MQA unfolding in Roon.

1 Like

Right.

Using some of the bits in the samples to convey high frequency data instead of high dynamic range data is just a choice of encoding scheme. This is commonplace in signal transmission.

We transfer bits. It’s just a convention to say that they correspond in a fixed manner to 2 channels of 24 bit words at a 96k rate. We may convert from analog that way, but transmission is a different matter.

Insisting that the signal sample rate correspond directly to the audio frequency, and the signal sample word length correcspond directly to audio dynamic range, is unnecessarily simplified and backward-looking. It creates the rectangular data space that is very inefficient.

I don’t think there is anything controversial about coming up with a more efficient use of the data rate, of the bits. Happens all the time. Look at how pkzip encodes the data, nobody complains about that.

Once you accept that the idea can work, you can then discuss whether the particular implementation is effective and efficient and harmless. But decrying it as unnatural is meaningless.

3 Likes

Nobody complains because there’s no loss of data, it is lossless packing, mathematically identical. The same goes to FLAC. In digital recordings the effective 24 bit data in a real world exists (2 to the power of 24 combination) but we simply can’t utilise the full range of a 24 bit resolution or the dynamic range. What Bob Stuart said is correct and I strongly agreed.

17 bit at 96kHz is still considered Hi-Res but there’s more effective way to encode in pure FLAC and still produce a comparable file size, lossless and able to playback on virtually any devices out there.

Yes, I have seen Jussi’s claim for that, for example.
If you accept that compact encoding of high resolution data could be useful, then it is fair game for anybody to attempt to address this opportunity with alternate innovations.
The challenge is that a solution requires not just an encoding scheme with the balancing of various tradeoffs, but also getting the publishers and streaming services and hardware and software manufacturers to support it. The MQA team are working hard at that, and there is a lot of complaining about the slow pace of adoption.
So far, I have seen no content published in any alternate formats.

Not a good review of MQA - actually pretty bad all around. Question is do you believe his graphics and his conclusions based on those graphics???

As more tests are revealed vs what are claimed, we begin to see the real truth! There’s a trade off between implementing a good time domain correction vs aliasing effects. Interestingly, there’s more impulse response digital filters than any other standard PCM! Which one it actually use, without the knowledge of the users are still a mysterious… Re-sampling anything above 96kHz to 384kHz does not contain any real information, the first unfold up to 48kHz is the more important.

What I don’t like is someone takes a 44.1/48kHz and give you 88.2/96kHz and sell it as hi-res; obviously this is cheating. It had happened elsewhere before and now we can seeing all over again :disappointed:

Wow, you guys are still here? :joy:

5 Likes

Well you are still here, why would not anyone else still be here?

It seems to me there are basically two camps on MQA, those that accept the claims made by the MQA team and consequently like it, and those that are sceptical about it. I belong in the sceptic camp. The more info that gets put into the public domain by the likes of Archimago the more reasons there are to be a sceptic. There is some truth in the saying there is no such thing as a free lunch. MQA cannot deliver audio Nirvana with no downsides. There is no magic in what it does, just obfuscation.

A debate is a good thing if it is informed and we are not fooled by professional dressing up of flawed analysis.

Archimago often posts some good stuff, but I’ve not seen an analysis of MQA from him that isn’t incorrect in some way, incomplete, or misses some crucial point.

The latest blog, IMO, reaches incorrect conclusions on the filtering. Obviously he is not privy to the full MQA process (nor am I for that matter), but a simple patent search reveals that Bob Stuart and Peter Craven have moved beyond linear phase vs. minimum phase. In fact, the patent application reveals that they are very much pro-linear phase; it’s just the pre-ringing which is the problem. They have apparently figured out a way to have their cake and eat it: linear phase (for best spatial accuracy) with selective group delay (to get rid of existing ringing and to suppress ringing in down-stream filters).

Regarding the selection of different filters, that shouldn’t be a surprise, because (a) it’s public information that the mastering engineer has options in the encoding, and (b) Bob has always said that MQA is an end-to-end process. My take on this is that, if you encode a track one way, you need to select the appropriate filter during final unfold and rendering.

In terms of aliasing, what matters is whether or not it’s audible. MQA says not. The pro-camp are likely to trust MQA; the anti-MQA camp are likely not to…

5 Likes

MQA’s claims actually matter a lot more to the sceptics than the pro crowd. They are the ones constantly trying to pick those claims apart. It is probably also misleading to structure the argument in a way that suggests people ‘like’ MQA because they ‘accept’ the marketing. People like or dislike MQA for a myriad of reasons, but mostly people like it because it sounds good to them. A lot more are simply indifferent.

3 Likes