[gimmeheadroom]

Can you not hear a difference? Why would you need it 'checking'?

Are you saying measurements have no value?

 

[Bytor123]

Are you saying measurements have no value?

No? I'm asking why does s/he need something to show what s/he's listening to is 'better'.

 

[Dannemand]

MQA are the ones who stripped him of any further chance to test the codec, so it would be a bit big of anyone to declare that he's not given them a fair chance. Furthermore, it is a common fact of life that test signals that are good at testing equipment / codec performance and fidelity are not musical, they are not musical because test signals have test roles to fulfil that are not fulfilled by regular music, not because test signals are unrepresentative of music reproduction performance. I mean, basically the only test you can run using an unstructured music signal is the null test... which MQA, raw or unfolded, will just as well fail, digitally if not audibly, and which FLAC will predictably pass 100%. Is that the takeaway then?

If a "synth mode" exists for MQA, it also wasn't up to him to select or not select it. And if the codec adapts dynamically, would he / the operator needed to do anything different at all?

None of this is intended to endorse GoldenOne's claims, but just my way of saying "if there's something wrong with them, this is not it and that is not it"...

Again, my argument in this discussion is LESS about whether MQA is good or bad, and mainly about the validity of GoldenOne's analysis, which (A) seemed to me like a hit job disguising as objective analysis, and (B) was in actual fact based on tricking the encoder into accepting signals outside its defined performance envelope -- evidenced by it rejecting his test tracks, and next by the warnings it gave even when he managed to hide his tests inside music tracks. If you insist on running you car far above its red line, despite all the warning lights and alarms, don't be surprised if you break the engine.

As I wrote, there is nothing wrong with synthetic signals when testing a DAC or a lossless codec. But this codec is officially designed to encode "other information" (ultrasonic samples) in the bits normally used to preserve noise and extreme amplitudes in the high harmonics. Its premise being that none of those exist in actual music. It IS a lossy codec specifically for music, and useless for anything else. So a test looking to preserve those bits serves no legitimate purpose.

I only read somewhere that the encoder has a mode for synthetic music, but I don't know how it is implemented or whether Tidal offers this mode to artists who upload their tracks. I am guessing that Tidal exposes less than the complete encoder toolbox. You are right that if it were adaptive, it should have switched mode instead of rejecting the test tracks.

But my main argument against GoldenOne's analysis is that he never bothered performing the same before-and-after analysis on any of the readily available 2L tracks, where he could have compared their MQA encodes to original DXD tracks or a number of resampled versions. We can assume that their encodes were made with good understanding of how to use the MQA encoder and its settings. I recall @Currawong saying that those DXDs contain lots of ultrasonics (as one would expect in a 352.8kHz track) so there is plenty of opportunity to test the MQA encoder's performance on challenging material that is still within its defined envelope.

I fully agree that more openness from MQA would be enlightening. I don't know if GoldenOne requested their cooperation before he began his project or only after. As many journalists (and "journalists") have found over the years: Once you have published a hit-job on a company or individual, the target tends to be less cooperative afterwards

 

[Dannemand]

Agreed and I think this is more than reasonable. Seems like the issue is getting around MQA's checks and figuring out how to upload and encode tracks with the criteria you mentioned. Honestly, I would prefer if someone other than GoldenOne did this to rule out any potential bias, flawed processes, etc. It would be more in keeping with the scientific method and disputing the results of one person is much easier than disputing results of many people, whatever the results may be.

See my reply to @Joe Bloggs above.

The most relevant test GoldenOne could perform now to convince me of his sincerity, would be analyzing those 2L tracks the same way as his test tracks. This doesn't require access to the MQA encoder (they're already encoded) nor does it allow him to (deliberately or not) misuse the encoder or its settings to obtains skewed results.

 

[Dannemand]

Can you not hear a difference? Why would you need it 'checking'?

That's what audiophiles do. And why they spend so much money instead of just listening to music.

 

[Joe Bloggs]

Joe - do you listen to MQA in your downtime or stick to non-MQA?

Hi. You may know that I conceptualized the Magesound 8-ball on our players (now simply known as MSEB), and in my downtime these days I'm trying to program a dynamics recovery plugin that lets you get full dynamics back from crushed 90s-00s masters and any other recordings that were victims of the loudness war. If you're a dynamics addict like me you may even want it on all your recordings. Now if you know what MQA allows and doesn't allow on playback I think that answers the question

 

[vanhalen26]

Hi. You may know that I conceptualized the Magesound 8-ball on our players (now simply known as MSEB), and in my downtime these days I'm trying to program a dynamics recovery plugin that lets you get full dynamics back from crushed 90s-00s masters and any other recordings that were victims of the loudness war. If you're a dynamics addict like me you may even want it on all your recordings. Now if you know what MQA allows and doesn't allow on playback I think that answers the question

Ok, I’m moving on with my questions about that. If you find a miracle cure for brickwalled recordings, I’d love to hear about that. Insanely, some producers continue to ruin good music with that!

 

[Joe Bloggs]

Again, my argument in this discussion is LESS about whether MQA is good or bad, and mainly about the validity of GoldenOne's analysis, which (A) seemed to me like a hit job disguising as objective analysis, and (B) was in actual fact based on tricking the encoder into accepting signals outside its defined performance envelope -- evidenced by it rejecting his test tracks, and next by the warnings it gave even when he managed to hide his tests inside music tracks. If you insist on running you car far above its red line, despite all the warning lights and alarms, don't be surprised if you break the engine.

As I wrote, there is nothing wrong with synthetic signals when testing a DAC or a lossless codec. But this codec is officially designed to encode "other information" (ultrasonic samples) in the bits normally used to preserve noise and extreme amplitudes in the high harmonics. Its premise being that none of those exist in actual music. It IS a lossy codec specifically for music, and useless for anything else. So a test looking to preserve those bits serves no legitimate purpose.

I only read somewhere that the encoder has a mode for synthetic music, but I don't know how it is implemented or whether Tidal offers this mode to artists who upload their tracks. I am guessing that Tidal exposes less than the complete encoder toolbox. You are right that if it were adaptive, it should have switched mode instead of rejecting the test tracks.

But my main argument against GoldenOne's analysis is that he never bothered performing the same before-and-after analysis on any of the readily available 2L tracks, where he could have compared their MQA encodes to original DXD tracks or a number of resampled versions. We can assume that their encodes were made with good understanding of how to use the MQA encoder and its settings. I recall @Currawong saying that those DXDs contain lots of ultrasonics (as one would expect in a 352.8kHz track) so there is plenty of opportunity to test the MQA encoder's performance on challenging material that is still within its defined envelope.

I fully agree that more openness from MQA would be enlightening. I don't know if GoldenOne requested their cooperation before he began his project or only after. As many journalists (and "journalists") have found over the years: Once you have published a hit-job on a company or individual, the target tends to be less cooperative afterwards

As far as I can tell, all GoldenOne had done publicly by the time his recordings were all pulled from the catalog and him barred from publishing any further MQA recordings was... publish some MQA recordings.

As I was going to say, there's nothing special about test tones (other than the ultrasonic ones) that should make them unencodable. The only special thing about test signals compared to music is, well, they're designed to give test results. You can't design a psychoacoustic encoder that totally failed to encode test tones if you tried, and I think codecs like AAC are more psychoacoustic from the ground up than MQA would ever be. As I said, the only test I can see done with music signals (as you propose) is the null test, which MQA would just as well fail. Can you enlighten me as to the tests you would propose to have him run?

Finally, did anyone ever get anything out of MQA from saying "hey I want to run some test signals through your system and publish the test results in public"?

Not that I predict anything bad would come out of it , but it just seems to be something that institutionally can't happen

 

[hmscott]

John Lennon - Plastic Ono Band, The Ultimate Collection - 2021- 120 Tracks - 07:26:10... 192khz MQA
https://tidal.com/browse/album/180827516

John Lennon - 35 videos - 04/28/2021 - unrelated to "The Ultimate Collection" Videos
https://tidal.com/browse/playlist/2db30213-ab33-4b9a-946c-6ff654e4aab3

Youtube Playlist

Imagine - John Lennon & The Plastic Ono Band (w The Flux Fiddlers) (Ultimate Mix 2018) - 4K REMASTER


What Are You Listening To Right Now?
https://www.head-fi.org/threads/wha...w-rules-please-read-them.253245/post-16322495​

 

[Dannemand]

As far as I can tell, all GoldenOne had done publicly by the time his recordings were all pulled from the catalog and him barred from publishing any further MQA recordings was... publish some MQA recordings.

As I was going to say, there's nothing special about test tones (other than the ultrasonic ones) that should make them unencodable. The only special thing about test signals compared to music is, well, they're designed to give test results. You can't design a psychoacoustic encoder that totally failed to encode test tones if you tried, and I think codecs like AAC are more psychoacoustic from the ground up than MQA would ever be. As I said, the only test I can see done with music signals (as you propose) is the null test, which MQA would just as well fail. Can you enlighten me as to the tests you would propose to have him run?

Finally, did anyone ever get anything out of MQA from saying "hey I want to run some test signals through your system and publish the test results in public"?

Not that I predict anything bad would come out of it , but it just seems to be something that institutionally can't happen

Please read my full posts again, as I already answered these questions several times. I completely acknowledge your smarts and credentials, but it looks a bit like you are replying to what you thought I was saying, not what I was actually saying.

The test signals GoldenOne used (A) didn't contain noise, and (B) contained amplitudes in the high brilliance band where no music has such amplitudes. MQA was designed on the premise that those bits are wasted in the PCM stream to bit-perfectly preserve content that isn't present in any "natural" music (synthesized music is a different story). And it uses those bits to store other information (ultrasonic samples). Therefore it makes no sense to test it on such signals. I could have told him the outcome before he posted them -- and I strongly considered doing so.

That is not to say that test signals aren't useful. They are fantastic for testing lots of equipment. But not on a lossy codec that was deliberately designed with a narrower envelope, specifically targeting musical recordings from microphones, sampled by ADCs.

As I've now posted several times, GoldenOne should perform the same analysis using master tracks and "best effort" MQA encodes, such as the publicly available ones from 2L. If those also show poor performance, then he will have proven something. Showing poor performance on test signals that obviously break the encoder -- obvious with even the most rudimentary understanding of MQA's premise -- proves nothing.

We really, really should stop this, as we are trashing this thread, burying perfectly good posts from good members discussing Tidal tracks and albums they like. Discussions they are entirely permitted to have without constantly being bombarded by people telling them they are wrong.

 

[Joe Bloggs]

Please read my full posts again, as I already answered these questions several times. I completely acknowledge your smarts and credentials, but it looks a bit like you are replying to what you thought I was saying, not what I was actually saying.

The test signals GoldenOne used (A) didn't contain noise, and (B) contained amplitudes in the high brilliance band where no music has such amplitudes. MQA was designed on the premise that those bits are wasted in the PCM stream to bit-perfectly preserve content that isn't present in any "natural" music (synthesized music is a different story). And it uses those bits to store other information (ultrasonic samples). Therefore it makes no sense to test it on such signals. I could have told him the outcome before he posted them -- and I strongly considered doing so.

That is not to say that test signals aren't useful. They are fantastic for testing lots of equipment. But not on a lossy codec that was deliberately designed with a narrower envelope, specifically targeting musical recordings from microphones, sampled by ADCs.

As I've now posted several times, GoldenOne should perform the same analysis using master tracks and "best effort" MQA encodes, such as the publicly available ones from 2L. If those also show poor performance, then he will have proven something. Showing poor performance on test signals that obviously break the encoder -- obvious with even the most rudimentary understanding of MQA's premise -- proves nothing.

We really, really should stop this, as we are trashing this thread, burying perfectly good posts from good members discussing Tidal tracks and albums they like. Discussions they are entirely permitted to have without constantly being bombarded by people telling them they are wrong.

I could point out how you pick and choose which things I say to reply to and ignored the key question I asked you, that goes to the the premise on which everything you said in this reply stands on, the not answering of which makes it all moot; but then I am not trying to support GoldenOne's claims, just disputing your analysis of it, which I suppose gets us further off topic than ever. So I'd just drop this. I agree that music content is more important than anything, although bolder souls than me had already pointed out the implications of MQA in that regard.

 

[xand]

I could point out how you pick and choose which things I say to reply to and ignored the key question I asked you, that goes to the the premise on which everything you said in this reply stands on, the not answering of which makes it all moot;

 

[Dannemand]

Can you enlighten me as to the tests you would propose to have him run?

As I've now posted several times, GoldenOne should perform the same analysis using master tracks and "best effort" MQA encodes, such as the publicly available ones from 2L. If those also show poor performance, then he will have proven something. Showing poor performance on test signals that obviously break the encoder -- obvious with even the most rudimentary understanding of MQA's premise -- proves nothing.

I could point out how you pick and choose which things I say to reply to and ignored the key question I asked you, that goes to the the premise on which everything you said in this reply stands on, the not answering of which makes it all moot.

Test. MQA. using. music.

It's not that difficult a concept to comprehend, and I've repeated it umpteen times now.

But yes, time for the Ignore button.

 

[Joe Bloggs]

Test. MQA. using. music.

It's not that difficult a concept to comprehend, and I've repeated it umpteen times now.

But yes, time for the Ignore button.

And there I thought you were claiming that the right test signal would have been just fine?

It's impossible to test using music, that was the point. The analytical features you use with test signal output to calculate performance metrics are not there. Unless you're happy with it failing the null test as the only possible test, as I pointed out. Bye.

 

[Currawong]

I recall @Currawong saying that those DXDs contain lots of ultrasonics

They contain high amounts of DS modulation noise above 48 kHz, so they can't capture very high ultrasonics from instruments (assuming the mic used in a recording can, which is somewhat unlikely anyway).