[jagwap]

Which part of "from a Redbook definition" did you not understand?

So from your very singular position, decoded MQA is lossless?

 

[eio]

  Here is Benchmark's critique of MQA:
https://benchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa

benchmark talked about 18bit flac being smaller than MQA, which can be a death sentence to MQA if it is evaluated properly and generally applies to various musical content.
 
but benchmark haven't said much about the 18bit source: is it dithered (i bet not)? on what content?
as of the unknown 18bit pcm having better quality than MQA...well, that's their opinion, i don't know...
those things are crucial, MQA guys must have some reason to use such a complicated scheme afterwards. it all comes down to compression ratio.
 
personally i would consider MQA really revolutionary if it can achieve ~2x compression ratio than flac on the same 17bit ditherd (or 24bit source) pcm stream, given the quality loss is as small as they claimed.
 
anyway, it is an interesting compressor 

 

 

[watchnerd]

So from your very singular position, decoded MQA is lossless?

 
From a Redbook POV?
 
Probably (or nearly so), but where the lazy filter (necessary for that reduced ringing) of MQA starts to kick in matters, as well as the ability of the origami algorithm to restore flatness.  
 
Graphs like this make it looks like it starts to kick in at around a sample rate of 30khz, which is below Redbook standard:
 

 
 
Which ends up shaving a teensy bit from the top octave:
 

 
Which is then supposed to be fixed by the magic origami unfolding process:
 

 
I haven't seen any independent tests yet on just how well this works and if there are boundary conditions in the algorithm solution set that cause problems.
 
And how true in audio terms can we regard the results given the appearance of aliased components that were not present in the original recording? 
 
This is all highly theoretical and I'd like to see more direct head to head measurements.

 

[watchnerd]

  benchmark talked about 18bit flac being smaller than MQA, which can be a death sentence to MQA if it is evaluated properly and generally applies to various musical content.
 
but benchmark haven't said much about the 18bit source: is it dithered (i bet not)? on what content?
as of the unknown 18bit pcm having better quality than MQA...well, that's their opinion, i don't know...
 

 
Did you click on the link the article?  It would have lead you to:
 
http://www.computeraudiophile.com/blogs/miska/some-analysis-and-comparison-mqa-encoded-flac-vs-normal-optimized-hires-flac-674/
 
There is discussion in the comments section regarding dither.

 

[jagwap]

From a Redbook POV?

Probably (or nearly so), but where the lazy filter (necessary for that reduced ringing) of MQA starts to kick in matters, as well as the ability of the origami algorithm to restore flatness.  

Graphs like this make it looks like it starts to kick in at around a sample rate of 30khz, which is below Redbook standard:





Which ends up shaving a teensy bit from the top octave:




Which is then supposed to be fixed by the magic origami unfolding process:




I haven't seen any independent tests yet on just how well this works and if there are boundary conditions in the algorithm solution set that cause problems.

And how true in audio terms can we regard the results given the appearance of aliased components that were not present in the original recording? 

This is all highly theoretical and I'd like to see more direct head to head measurements.

My understanding is that the "lazy filtering" or apodizing filter only effects those frequencies when the sample rate is 44.1 - 48kHz. When reconstructed to 96kHz the low pass filter poles are doubled in frequency so the 20kHz passband is untouched.

Also it is generally part of the DAC, so may not be there in the digital file. Not sure on this, but it seems likely.

 

[watchnerd]

Some simple comparison testing here:
 
http://www.computeraudiophile.com/f8-general-forum/mqa-ces-27127/index18.html#post501224
 
"As evidenced by the graph, the original recording contains nothing of value about 16kHz, only sigma-delta modulator noise. The MQA encoding has filtered this out and replaced it with ... something. Here the MQA version has lower noise level well into the (somewhat) audible band, so it's no surprise if it sounds better. However, it would probably sound better still if it was simply filtered with a cutoff at 16kHz. Also of interest is that the difference at lower frequencies seen in the first sample is pretty much absent here.

From this I would say the claims that MQA preserves full CD quality even without a decoder are clearly bunk. What an MQA decoder might do is anyone's guess at this point."

 

[eio]

   
Did you click on the link the article?  It would have lead you to:
http://www.computeraudiophile.com/blogs/miska/some-analysis-and-comparison-mqa-encoded-flac-vs-normal-optimized-hires-flac-674/
There is discussion in the comments section regarding dither.

sorry i missed that link, thanks!
 
Miska did a 3x (surprise!) downsample on the 18bit file, and dithered (probably, Miska didn't explicitly say it) with TPDF, which is not an outstanding dithering algorithm and introduces noise into audible band.
personally i don't think it's a fair comparison.

 

[jagwap]

My understanding is that the "lazy filtering" or apodizing filter only effects those frequencies when the sample rate is 44.1 - 48kHz. When reconstructed to 96kHz the low pass filter poles are doubled in frequency so the 20kHz passband is untouched.

Also it is generally part of the DAC, so may not be there in the digital file. Not sure on this, but it seems likely.

Looking on my phone again. Redbook is the standard for CD. It does not include sample rates other than 44.1kHz or more than 16 bit. So 30kHz is way above the filtering requirements of redbook.

 

[eio]

i know this is a funny idea but maybe it can be subjectively evaluated by amplifying the very quiet moments in music, and slow it down to make ultrasonics audible?

 

[Morkai]

  i know this is a funny idea but maybe it can be subjectively evaluated by amplifying the very quiet moments in music, and slow it down to make ultrasonics audible?

 

 

[watchnerd]

Looking on my phone again. Redbook is the standard for CD. It does not include sample rates other than 44.1kHz or more than 16 bit. So 30kHz is way above the filtering requirements of redbook.

 
Yes, it was ridiculous late at night. I conflated sample rate and filter in my haze. Fixing post.

 

[jagwap]

   
Yes, it was ridiculous late at night. I conflated sample rate and filter in my haze. Fixing post.

It's time for sleep here.
 
It was a slow day, so thanks for the entertainment.

 

[castleofargh]

 

  Here is Benchmark's critique of MQA:
https://benchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa

benchmark talked about 18bit flac being smaller than MQA, which can be a death sentence to MQA if it is evaluated properly and generally applies to various musical content.
 
but benchmark haven't said much about the 18bit source: is it dithered (i bet not)? on what content?
as of the unknown 18bit pcm having better quality than MQA...well, that's their opinion, i don't know...
those things are crucial, MQA guys must have some reason to use such a complicated scheme afterwards. it all comes down to compression ratio.
 
personally i would consider MQA really revolutionary if it can achieve ~2x compression ratio than flac on the same 17bit ditherd (or 24bit source) pcm stream, given the quality loss is as small as they claimed.
 
anyway, it is an interesting compressor 

 

 
 you wouldn't just convert your 24/96 to 24/96MQA, because the MQA doesn't pull the extra samples from a hat, it needs the higher sample version PCM as encoding basis. so forget your idea of a better flac, it's not that at all. it's a different arrangement of resolution ratio that ends up in a container that doesn't look like a different arrangement of resolution ratio. it's the only rel confusing part really. they decide what they want to keep (or can keep depending on the amplitude) and that ends up hidden in a PCM file, that file can be encoded in any lossless format you want, the same way wave can. for the converter it's factually the same format. so you will get your MQA as a FLAC, that's no problem at all.
 
I see it as a different resolution. no more, no less. more samples, less bit depth. there is nothing revolutionary.
12/384 stereo would be about 67mb for a minute
24/192 would be about 67mb for a minute.
look I made a revolutionary format of 12/384 that has higher "resolution"(not really), and my impulse response looks much better than a 24/192 file so I can claim to solve "time smearing" and misguide people in thinking that it really has better definition. as long as I don't talk too much about the noise floor going up in my marketing, surely I have the format of the future.  reminds you of something?

 
 
we can pretend to argue forever, if you believe ultrasound content will save your life, then go for stuff with high sampling rate and be happy(DSD 128 must be the real bomb). if like me you know for a fact that you can't hear the difference past 16/44 on correctly implemented DACs, ultrasounds will always be a waste of space however you encode them. it's really that simple IMO.

 

[TheTrace]

I'm extremely sceptical about the benefits of the MQA technology itself to the point where I'm uninterested. It seems if there are benefits it's pretty much for steaming, most of the time I like using native files on my device and stream at home.

However what I am interested in is if they're using better masters (or remastering) for the MQA albums and if that's what the improvement is. If the masters are better than what I can find on CD and Hi res due to remastering, I may have an interest to obtain some albums. Been searching high and low for the answers to that.

 

[eio]

i need much more understanding and testing to know it better.