MQA Deep Dive - I published music on tidal to test MQA

[amirm]

@amirm , can you keep us updated on how the MQA music encoding effort is going? I think many people here and elsewhere are ready to contribute, myself included.

It is not currently as no one in the industry has reacted to my quest.

 

[voodooless]

That would be fine too. I think @voodooless is citing 20/96 because it would be sufficient even by Stuart's own standards: 20 bits' worth of depth, plus a sample rate that permits encoding of frequencies at 26kHz (per Stuart's AES paper) and has a Nyquist of 48kHz (MQA's maximum).

Indeed @lucretius is missing the point here. If the renderer only actually gets 20 bit resolution audio at best, then 20/96 is all we need to be on par with what MQA delivers. And we already knew that it compresses about the same as a 24/48 MQA file. This basically negates the whole point of having MQA’s lossy encryption the first place.

Personally, I’ll take 24/48 just as well.

.. and before anyone asks.. no I can’t sleep anymore (it’s 4:35 here).. it’s bloody hot and the birds outside are driving me crazy Maybe I should put on an undecoded MQA silence track at full volume and drown the bird sounds with pleasing noise shaped ultrasonic bliss

 

[amirm]

Indeed @lucretius is missing the point here. If the renderer only actually gets 20 bit resolution audio at best, then 20/96 is all we need to be on par with what MQA delivers.

You can't sell "20" bits to audiophiles who expect "24 bit" content. And you would be back to relying on dither to produce that although ways that could go wrong would be less.

 

[voodooless]

You can't sell "20" bits to audiophiles who expect "24 bit" content. And you would be back to relying on dither to produce that although ways that could go wrong would be less.

Looks like you can sell it.. it’s called MQA It’s just that nobody told the audiophiles as much. They could have made MQA with far less effort: just make a new file format derived from FLAC that uses 20 bit audio internally, and outputs 24 bit audio after decode. They could still add the magic bits to authenticate and control the renderer in the decode phase. But of course this would not have been a good story to sell.

 

[tmtomh]

You can't sell "20" bits to audiophiles who expect "24 bit" content. And you would be back to relying on dither to produce that although ways that could go wrong would be less.

Are you saying MQA delivers 24 bits’ worth of depth, or are you saying it’s okay for MQA to let consumers think they’re getting 24 bits even though they’re not?

 

[ebslo]

You can't sell "20" bits to audiophiles who expect "24 bit" content. And you would be back to relying on dither to produce that although ways that could go wrong would be less.

Devil's advocate: "Why would you want 24-bits? That's 4 bits MORE NOISE than 20-bit. You don't want MORE NOISE do you?"

 

[Raindog123]

Are you saying MQA delivers 24 bits’ worth of depth, or are you saying it’s okay for MQA to let consumers think they’re getting 24 bits even though they’re not?

99.9% of MQA consumers (or any format, for that matter) do not know how to spell '24-bit'. The rest are too busy with their crusades to enjoy any music.

evil's advocate: "Why would you want 24-bits? That's 4 bits MORE NOISE than 20-bit. You don't want MORE NOISE do you?"

Not if you fill those bits with 0's, rather that that pseudo-random crap.

 

[DimitryZ]

It is not currently as no one in the industry has reacted to my quest.

That's too bad. Would it be crazy to call the nearest MQA-equipped mastering lab and ask how much it would take to encode a track or two?

As for a master, I am sure there are many amateur and professional recording engineers here, including @John Atkinson , who may be able to convince talented musicians to allow a track to be encoded for a non-commercial purpose (over a craft beer, single malt scotch or a vintage champagne).

I suggest that an appropriate royalty is considered to be a part of the effort. I don't think the funds are the limiting gate here.

 

[ebslo]

Looks like you can sell it.. it’s called MQA It’s just that nobody told the audiophiles as much. They could have made MQA with far less effort: just make a new file format derived from FLAC that uses 20 bit audio internally, and outputs 24 bit audio after decode. They could still add the magic bits to authenticate and control the renderer in the decode phase. But of course this would not have been a good story to sell.

More devils advocate: I wonder what what the effect on FLAC encoded file size would be of sign-extending 20-bit PCM to 24? Seems it would compress pretty well...

 

[DimitryZ]

99.9% of MQA consumers do not know how to spell '24-bit'. The rest are too busy with their crusades to enjoy any music.




Not if you fill those bits with 0's, rather that that pseudo-random crap.

I guess I am that one in a thousand MQA consumer with rudimentary spelling ability. And English isn't even my first language.
I am enjoying MQA right now.

 

[voodooless]

More devils advocate: I wonder what what the effect on FLAC encoded file size would be of sign-extending 20-bit PCM to 24? Seems it would compress pretty well...

Yes probably it would. But that way you’ll have to add dither after the decode to fill up those zeros.

 

[ebslo]

If publishers really cared about proving authenticity they would just create a signature of the PCM data with an X.509 certificate and store it in the FLAC metadata.

 

[Raindog123]

Yes probably it would. But that way you’ll have to add dither after the decode to fill up those zeros.

Why would you do that? If your PCM signal has, let's say 16 bits of dynamic range, and you add 8-bit randomized dither on top of it, you'll get a porcupine of a waveform. The one that looks like one, and sounds too... You add just enough dither to 'smoothen' digital quantization, and that's about it. The rest should be zeros.

EDIT: [And yes, you can start playing with 'shaping' the noise and 'moving it out of band', but even that you do not do by full-scale random or PN dither...]

 

[ebslo]

Yes probably it would. But that way you’ll have to add dither after the decode to fill up those zeros.

Really? Presumably the dither was added during the re-quantization from the 24-bit master to the 20-bit encoding. Since sign extension doesn't change the value, the sign-extended "24-bit" samples would have the same value as the samples in the 20-bit encoding. So why would they need to be filled with noise (other than to obfuscate what you had done)?

But never mind really, it was just a silly joke

Edit: NVM again, @Raindog123 beat me to it.

 

[voodooless]

Ah wait, you guys mean the zeros on the other side. Stupid me

But there are issues with this as well. Given a state of art DAC, it has 20 bit resolution at best, you’ll never get the last 4 bits out. You’re basically back to Redbook . So without converting your fake 24-bit to real 24-bit, this won’t work either. But obviously this needs to happen with the 20-bit file as well. The advantage however is that you know upfront that it’s a 20-bit file.

But never mind really, it was just a silly joke

While jokingly mentioned the idea is not a silly joke at all.

 

[BrEpBrEpBrEpBrEp]

Yes. Pathetic.

20 bits is pathetic? It's only the limit of modern DACs...

 

[Rottmannash]

That you’re a shameless annoying troll. Both in appearance and substance.

And it’s our tireless moderators’ job to deal with the likes of you. Or, often you guys just fade away in a couple of days, reaching your depth. So, let’s see.

Finally....

 

[pozz]

do not know how to spell '24-bit

They can't spell the word.

(Oh god what a bad pun )

 

[Raindog123]

...I'll put myself on the spot here (gladly, as it will allow true experts to throw rocks.)

But while RF-frequency ('gigahertz class") DACs' state-of-the-art today is about 16+ bits, I thought that low-frequency audio DACs (and digital filters of DAC SoCs) are at least 24bit, and might even be 32 now...? While still 'of consumer prices'..?

 

[Rottmannash]

Quite the opposite, sadly.

MQA will not encode high amplitude utrasonics, as they don't exist in music, at least with an automatic encoder.

Do you understand? If not, read the first word and see if you do. Proceed to the next word and repeat.

Your snark is getting a bit old and tired. Please try to be more pleasant in your attempt to belittle anyone who disagrees with you.