MQA pt II

[adamdea]

Now I broke my neighbour's irony meter!

I'm afraid that if this goes on much longer the irony police are going to have to seal off the whole area to prevent escape of the unthinkably high levels of irony, leaving us all trapped here forever.

 

[Jim Audiomisc]

Jim, I did this in Feb 2017 and apparently I did not take many notes, or I just deleted them after losing interest.

I captured the MQA-unfolded output from Tidal in the digital domain with Audacity. I also captured the fully MQA-decoded output of the Explorer2 using a Tascam recorder running at 192kHz (a DV-RA1000, these days I use the more handy DA-3000). But a warning with the latter approach: as you know nearly all audio ADC chips these days are delta-sigma and have shaped quantisation noise rising rapidly once above 20kHz. This obscures of course that which one wants to investigate in an MQA capture!

It is much cleaner to capture the first unfold in the digital domain, and then add oversampling with MQA-rendering filtering oneself afterwards. Mans published the render filters that are used in the wild.


Yes, Britain's coastline is infinitely long. But the border of Colorado is not.

If magic triangles are good, then magic hexagrams might be better.

I have various ADCs inc the Benchmark ADC1 which seems quite good. Its USB outlet only goes to 96k/24 but the spdif and AES go to 192k/24 and I do have a Tascam portable with inputs for those. So can probably cobble something together.

That said, my hope is that the MQA DAC I'm wanting will be class compliant. if so I may be able to use it with Linux *and* RISC OS. And a while ago I wrote a RO prorgram that can act as a 'test bench, driving an ADC and DAC in parallel playing whatever test waveforms or files I fancy. Would be convenient to use that! If not, I'll put something similar together for Linux, or just hack it using two progs running together.

I assume Colorado is less chaotic than Britain, thanks to map-makers using straight lines. We used the Sea. 8-]

 

[Jim Audiomisc]

(OT)



You will remember that I told you that we visited that very Concorde three years ago!

No! Memory failure on my part, alas! I suspect I may also was forgetful in the past, but can't remember...

The seats in it were probably more comfortable when you saw it, and the right way round so you didn't hang from the straps at take-off. 8-]

 

[Jim Audiomisc]

I'm sorry to hear that mansr, on a number of levels. Amir has as long as I have been aware of him, been a disappointing mix of admirable and non-admirable tendencies. Leaving aside the behaviour, his approach on this issue is quite baffling since I struggle to believe that he is not aware of the inconsistencies.

I did have a brief look at the ASR thread yesterday and noticed that he seemed happy with my 'part 1'. Didn't read more than a few pages, so have missed other comments there. (1) But got the impression that some of the possible implications hadn't been picked up. And no idea if others have read it. Fair enough as this stage, though.

(1) As you get older, life gets shorter. :-/

 

[Jim Audiomisc]

Audio coding using the I Ching?

Name of a Chinese shop-till cash machine manufacturer? Or is that joke too old?

 

[adamdea]

I did have a brief look at the ASR thread yesterday and noticed that he seemed happy with my 'part 1'. Didn't read more than a few pages, so have missed other comments there. (1) But got the impression that some of the possible implications hadn't been picked up. And no idea if others have read it. Fair enough as this stage, though.

It struck me that your article had ostensibly been praised but had rather selectively been quoted, without regard to the delicate and playful understatement of your "mostly harmless" conclusion.

 

[Jim Audiomisc]

It struck my that your article had ostensibly been praised but had rather selectively been quoted, without regard to the delicate and playful understatement of your "mostly harmless" conclusion.

I've assumed that given the 'science' in the forum's name people there would follow the RS approach of "Take no-one's word for it" and read it for themselves. But I can't tell.

Maybe the've never heard H2G2.

 

[March Audio]

I have various ADCs inc the Benchmark ADC1 which seems quite good. Its USB outlet only goes to 96k/24 but the spdif and AES go to 192k/24 and I do have a Tascam portable with inputs for those. So can probably cobble something together.

That said, my hope is that the MQA DAC I'm wanting will be class compliant. if so I may be able to use it with Linux *and* RISC OS. And a while ago I wrote a RO prorgram that can act as a 'test bench, driving an ADC and DAC in parallel playing whatever test waveforms or files I fancy. Would be convenient to use that! If not, I'll put something similar together for Linux, or just hack it using two progs running together.

I assume Colorado is less chaotic than Britain, thanks to map-makers using straight lines. We used the Sea. 8-]

@Jim Audiomisc I have a meridian explorer 2 sat in a drawer doing nothing. It's class compliant. Happy to send it through to you.

 

[Cesare]

The basic problem with the whole approach is that there is no unused 'space' in a sampled signal for you to insert another signal so as to not in some way damage the original. The idea that you can losslessly add another signal, then remove it and return the original signal *only* works if you know ahead of time what signal was added (as in, it's deterministic) so in this situation there is no other information conveyed. What is confusing to me is that the standard used to encode this impossible signal is quite capable of transmitting higher sample rate formats, so if that's the aim, why not use it? It can't be a problem with the bandwidth required, as audio is so low compared to something like video, and people are streaming video all over the place.

It's impossible technically. Just to add another point, the best mics in the world don't capture >20Khz, some struggle above 15Khz. The point is, good mics are large diaphragm, and hence the moving mass means they don't like high frequencies. A neumann U87 is down 10db by 20Khz, and that's with an artificial presence boost around 8Khz to try and stop it sounding dull (I would guess it's rolling off around there otherwise and would be down 15db by 20Khz otherwise).

Screenshot 2021-06-01 at 15.01.46 by Cesare Ferrari, on Flickr

 

[mansr]

MQA is an elaborate money-making scheme poorly disguised as a technical solution to a made-up problem.

 

[Jim Audiomisc]

The basic problem with the whole approach is that there is no unused 'space' in a sampled signal for you to insert another signal so as to not in some way damage the original.

To be fair, it would be possible to hide some added HF into in 44.1/ or 48k *24bit* files as the bottom few bits/sample tend to be noise anyway. So they might well be able to do a decent job as a trade-off if the extras were encoded to sound like noise as they'd be lower than the *actual* noise level, so would make little difference. The problems arise with trying to get away with more than a tiny amount of non-LPCM by 'hiding' it at HF in 16bit low-rate material.

Thus you tend to run into the "Does little harm if it doesn't anything much" situation.

All that said, noise shaping to 96k/16 would work fine and not need a 'magic decoder ring' or pay anyone for the format.

 

[Jim Audiomisc]

@Jim Audiomisc I have a meridian explorer 2 sat in a drawer doing nothing. It's class compliant. Happy to send it through to you.

I'll PM you. That sounds interesting

FWIW I happily used a Meridian 200 + 263DS and then + 563 for many years and thought the DACs sounded excellent. (Found the deck a bit of a pain at times, though.) So I have liked Meridian DACs in the past.

 

[Rosewind]

Thanks to March Audio for this condensed version of MQA grievances:

Any rational, objective, subjective or user benefit analysis concludes MQA is a solution looking for a problem. It's obviously a land grab attempting to monetise the music distribution and hardware chain. Technically it just damages the recorded signal. There isn't anything about it that benefits the consumer. Nothing, Nada, zip.

 

[mansr]

Jim Collinson of Linn was right on the mark back in 2017:
MQA is an attempt to not simply sell the same content again at a higher margin, or to maintain audio quality in streaming ecosystems: it is an outright land grab. It’s an attempt to control and extract revenue from every part of the supply chain, and not just over content that they hold the rights for. It really is quite extraordinary.

https://www.linn.co.uk/uk/blog/mqa-is-bad-for-music-heres-why

 

[March Audio]

I'll PM you. That sounds interesting

FWIW I happily used a Meridian 200 + 263DS and then + 563 for many years and thought the DACs sounded excellent. (Found the deck a bit of a pain at times, though.) So I have liked Meridian DACs in the past.

I used to have a 563 and 500 transport and agree it sounded fine. The 500 drawer did get a bit tempramental with age and would stick requiuring cleaning /lubricating.

Sent the explorer info through. It has latest firmware, works with my raspberry pi running linux so indeed looks class compliant. Also worked with some 2L tracks from Qobuz played by Roon. Blue light came on hahahahahah.

https://photos.app.goo.gl/Zh1P9CEGAG4qXpAk6

 

[John Phillips]

Thanks to March Audio for this condensed version of MQA grievances:

Any rational, objective, subjective or user benefit analysis concludes MQA is a solution looking for a problem. ... Technically it just damages the recorded signal. There isn't anything about it that benefits the consumer. Nothing, Nada, zip.

This is how I see it. But I see a lot of very intelligent people with admirable curiosity discussing the technical trade-off between information discarded by the encoding vs. information added by the coding.

That surely distracts from the fundamental issue that no trade-off is necessary these days. There is no need to discard any information. EDIT: no technical need.

Being inveigled into technical argument about the trade-off simply perpetuates the fundamentally flawed idea that the trade-off is a legitimate matter to discuss in the first place.

 

[Jim Audiomisc]

Well, one possible benefit of this is that it may have provoked more people into considering two issues:

1) So what is the *real* amount of audibly-significant content in the region above about 20kHz?

2) Can we simply use methods like noise shaping to fit what is needed efficiently into existing formats like flac containing 96k/16 LPCM payloads?

Personally, I'm interested in MQA for various reasons. One being that it helps us to think about the above questions and also about just what 'music' may be in those terms. A factor I keep in mind is that we lose any sense of 'pitch' once we go well above 5 - 10 kHz and take it from lower components.

 

[mansr]

1) So what is the *real* amount of audibly-significant content in the region above about 20kHz?

None. We can't hear those frequencies.

2) Can we simply use methods like noise shaping to fit what is needed efficiently into existing formats like flac containing 96k/16 LPCM payloads?

96 kHz, 16-bit with noise shaping and some pre-emphasis could probably capture anything remotely signal-like coming out of a microphone. Then again, 48 kHz, 24-bit (or 20-bit) can easily capture everything _audible_, so why bother with anything more complicated?

 

[John Phillips]

Well, one possible benefit of this is that it may have provoked more people into considering two issues:

1) So what is the *real* amount of audibly-significant content in the region above about 20kHz?

2) Can we simply use methods like noise shaping to fit what is needed efficiently into existing formats like flac containing 96k/16 LPCM payloads?

Personally, I'm interested in MQA for various reasons. One being that it helps us to think about the above questions and also about just what 'music' may be in those terms. A factor I keep in mind is that we lose any sense of 'pitch' once we go well above 5 - 10 kHz and take it from lower components.

With apologies if my comment (actually from reading the ASR thread) had unintended consequences.

Yes. A long while ago I similarly asked myself about what end-user-delivery digital audio parameters might be regarded as clearly "blameless" (that's not the same as "sufficient"). And from a moderately well informed PoV (but not formally educated) it seemed to me that Red Book CD did fall a little short of being clearly blameless. My conclusion then was 48 kHz 20 bit would get there.

I do see your questions as perfectly relevant. For my part I certainly favour not discarding audio information whenever possible, and ISTM that using modern psychoacoustic modelling as in codecs like AAC seems very successful to these ears for (few I hope) situations where you do.

 

[Zombie]

Oh, no! The thread is grinding to a halt. You need to fire up DimkaZ again