MQA is Vaporware

[Fokus]

10 minutes ago, Lee Scoggins said:

MQA may be advantageous as it can enhance 24/48 and 16/44 files as well.

 

So you believe ...

 

But then anyone can enhance 1x files. No need for a closed system with Trojan DRM and misleading claims.

 

[MetalNuts]

43 minutes ago, Lee Scoggins said:

Also, you make the assumption that bandwidth is plentiful.  In my discussions with people in the industry such as David Chesky and Ken Forsythe, bandwidth at the scale of streaming is a very real issue. 

 

I do not understand why 4K movie now can be streamed (Netflix) without bandwith issue but music files have.  I believe the data size of 4K movie be it in terms of seconds, minutes, hours must be bigger than 24/192 music data.  Is it the same technique in streaming data?

 

 

[Miska]

1 hour ago, Lee Scoggins said:

Also, you make the assumption that bandwidth is plentiful.  In my discussions with people in the industry such as David Chesky and Ken Forsythe, bandwidth at the scale of streaming is a very real issue. 

 

See https://help.netflix.com/en/node/306

 

I have no problem streaming Ultra HD (4K) over 4G LTE mobile data.

 

96/24 FLAC would be around 2.2 Mbps. RedBook is around 0.8 - 1 Mbps.

 

Uncompressed raw DSD256 is 25 Mbps.

 

[Norton]

1 hour ago, firedog said:

Again, why not just release the ACTUAL 24/96 version for streaming. There is no technical or bandwidth reason it can't be done. 

You seem to blind to this simple fact, and somehow see MQA as a needed innovation for hi-res streaming. It simply isn't needed.

 

You may be correct from a theoretical point of view, but that's no practical use to today's listener.

 

To repeat my previous question, other than Tidal/MQA, who is actually offering 96 kHz classical music streaming today?  I looked at Qobuz for example, as far as I can see from their site, the implication at least is that anything beyond RBCD is only offered as a purchased download.

 

The 2l site shows that a MQA a file is c. half the size of 24/96, presumably that has something to do with it's "stream-ability" ?

 

I am coming at this from the assumption that,  in the particular situation where both the MQA source file and the original non-MQA music release are at 96khz and the MQA stream decodes  at 96kHz to my non-MQA DAC,  then what I am getting is no less than what a (theoretical) non-MQA streaming service or a 24/96 download would offer at the same resolution.  

[Fokus]

6 minutes ago, Norton said:

The 2l site shows that a MQA a file is c. half the size of 24/96, presumably that has something to do with it's "stream-ability" ?

 

You should compare with 18/96. It has been shown that this compresses as good as or better than MQA.

It is trivial to generate an 18/96 copy of a 24/96 original. This does not tax the server side at all.

 

6 minutes ago, Norton said:

The 2l site shows that a MQA a file is c. half the size of 24/96, presumably that has something to do with it's "stream-ability" ?

 

I am coming at this from the assumption that, ... where ... the MQA stream decodes  at 96kHz to my non-MQA DAC,  then what I am getting is no less than what a ...24/96 download would offer at the same resolution.  

 

It is less. The unfolded MQA is equivalent to 17-18/96, not 24/96.

 

Again, MQA is a solution without a problem. Therefore it is not a solution.

[Miska]

30 minutes ago, Norton said:

The 2l site shows that a MQA a file is c. half the size of 24/96, presumably that has something to do with it's "stream-ability" ?

 

I am coming at this from the assumption that,  in the particular situation where both the MQA source file and the original non-MQA music release are at 96khz and the MQA stream decodes  at 96kHz to my non-MQA DAC,  then what I am getting is no less than what a (theoretical) non-MQA streaming service or a 24/96 download would offer at the same resolution.  

 

MQA averages about 16/96 resolution (17-18 bits on classical where you don't have much ultrasonic content and <16 bits on pop material where you have more ultrasonics), depending on material. When you actually convert the original to 16/96 or 18/96 standard FLAC, it becomes smaller than the MQA version. I've tested this with the 2L tracks and bunch of others as noted earlier.

 

[church_mouse]

53 minutes ago, Norton said:

 

You may be correct from a theoretical point of view, but that's no practical use to today's listener.

 

To repeat my previous question, other than Tidal/MQA, who is actually offering 96 kHz classical music streaming today?  I looked at Qobuz for example, as far as I can see from their site, the implication at least is that anything beyond RBCD is only offered as a purchased download.

 

 

 

 That is not quite correct - Qobuz offers the Sublime+ package which gives Hi-Res 24-bit FLAC/ up to 192kHz streaming.

 

I do not have that package, so I cannot comment on how well it streams in practice.

[mcgillroy]

2 hours ago, Lee Scoggins said:

 

I don't think you are seeing my point.  Labels feel 1. that streaming is the future and 2. MQA is an important component.  So my guess is that they don't really care about sound quality so much as getting more people paying them a monthly annuity to do streaming.  For all I know, they may just be using MQA from a marketing standpoint leveraging the sound quality to attract interest.  I have heard that mastering teams at each label spent a good bit of time evaluating and approving the encoding process.

 

Lets walk through this as your statements are good examples of half-truths that make the MQA-debatte so interesting and difficult and exposes the audio-presses shortcomings.

 

Labels indeed see that streaming is the future. But they imagine a different future than those platforms providing streaming at present. Plattform companies like Apple, Google, Amazon, Facebook and Microsoft already own sizeable chunks of the streaming market and more importantly the tools for music discovery (search, social networks) as well as the endpoints: phones, tablets, consoles and last not least speakers. The biggest DAC and speaker companies of the world are:

 

Amazon and Google...

 

Smartspeakers are access points for Music and this category is exploding.

 

15 Years ago the labels where at the mercy of Apple as the only means against Napster. With streaming the labels are even more so at the mercy of the tech industry. The tech industry owns discovery, access, distribution and to some extend the billing of music.

 

This is not the streaming future the labels want.

 

The only asset they have are their catalogues. Encoding the catalogs in a proprietary format that allows to scale quality (price discrimination) as well as to meter music consumption at the endpoints via a sophisticated DRM embedded not in the transport-frames but in the music-files themselves would be really cool. MQA offers that.

 

That is what the DRM of MQA is mainly about: shackle the endpoints (DACs in phones, consoles, smartspeakers and the few standalone DACs audiophiles entertain) to a mechanism allowing to extract licensing revenue from the platform companies. Which view music as a cheap ancillary of their business models.

 

As you rightly remark: labels don't care about sound quality, they care about control and MQA promises control in a world where they have lost such to bigger players.

 

This is also why they take interest in the encoding process. Not because people sit there and actually listen for enconding artefacts. They care that their crypto-credentials get properly embedded in the MQA-DRM. Plz go ask MQA to show you their tools for the encoding process and the training these people receive. You will be surprised how much it is about managing those credentials.

 

 

Quote

 

MQA may be advantageous as it can enhance 24/48 and 16/44 files as well.

 

This, dear Lee, is simply misleading and outright wrong if looked at in the light of MQAs own statements.

 

MQAs promises sound true to the "studio sound" as the "artist heard" it.

 

Any "enhancement" is clearly not that.

 

Those "enhancements" are artefacts of digital filters that might yield euphonic effects preferred by some listeners. Any upsampling library can provide similar enhancements and would give the listener control over them. MQA instead uses a proprietary recipe applying filters on a per track basis. A process that is not unlike MP3-algoryhtms deciding on a pre-track basis which frequencies are important and which can be thrown out.

 

Enhancement it is not, young Padawan, illusions of such it is. Cheap they are, yet significant they appear.

 

 

Quote

 

Also, you make the assumption that bandwidth is plentiful.  In my discussions with people in the industry such as David Chesky and Ken Forsythe, bandwidth at the scale of streaming is a very real issue. 

 

As other people have remarked it is apparent that successful streaming businesses  with much higher bandwidth requirements exist: Netflix and other video streaming services.

 

We are talking up to 10x higher requirements than for music.  

 

All the while Qobuz streams at 24/192 in one of the worlds biggest market. You as a journalist and a data broker in real live, perhaps could do your homework and read up and interview a few people on bandwidth control, allocation and billing. You could run a few numbers about aggregate storage and bandwidth requirements for 10 million or 100 million subscribers and compare that to the offerings available. You might find that bandwidths gets brokered both short and longer term and that at least for the US the fall of net-neutrality has lead to a frenzy of renegotiations of terms. You might find that the typical 40 million tracks a streaming service offers even if stored in 24/96 (as Apple asks for when staging) including metadata are not even 3 petabyte. You could get some numbers from AWS, Azure, Google, Joyent asking how much storing and streaming of those 3pb at different availability levels cost you. You could look at Spotify and other services annual reports and run those numbers against their expenses. You could ask labels and MQA to make sense of these numbers and check them against those business reports and offerings for photo-, video or game-streaming.

 

All this would move the needle on the MQA debate beyond the half-truths and misinformation you audio "journalists" entertain us with. It really comes down to what the likes of you, Atkinson, Harley etc choose: further bury your reputation being MQA-syncophants or actually prove your worth to consumers.

 

Last not least you could simply ask yourself why MQA wastes one fourth of that precious storage and bandwidth for crypto instead of music.

 

 

 

 

[Norton]

1 hour ago, Fokus said:

 

MQA is a solution without a problem. Therefore it is not a solution.

 

Again, we are not really discussing MQA, we are talking objective v. subjective or theory vs. lived experience.

 

I'm on this site because I love high quality music replay.  I'm currently listening to Gergiev Mahler2 via MQA (just a lowly 48kHz MQA source file).  Inter alia, the thing you describe as "not a solution" seems to be providing me, right now, with some of the best digital reply I've heard.

[Norton]

42 minutes ago, church_mouse said:

 That is not quite correct - Qobuz offers the Sublime+ package which gives Hi-Res 24-bit FLAC/ up to 192kHz streaming.

 

I do not have that package, so I cannot comment on how well it streams in practice.

 

Thanks, I seem to see different statements on different parts of their website, with various asterisked  caveats. It would be good to hear from someone with this package who can comment on what is currently available,  what labels etc.  As someone who couldn't see the point of streaming, I'm rapidly being converted.  I'm mistrustful of "credit card" trials though.

[Shadders]

4 hours ago, Fokus said:

 

Don't be naive. For an AS body to take this on it needs access to impartial expertise. Impartial almost implies people from outside the music or sound industry, and as for expertise ... my guess is that not even 10% of trained engineers totally and deeply understand sampling as it pertains to audio.

Hi Fokus,

The number of engineers that understand sampling etc., in other engineering domains such as RF, communications, and other areas of electronics, dwarfs the audio industry. Audio is just another subset of engineering and it is very small when considering numbers of true/chartered engineers working in it.

Regards,

Shadders.

[Fokus]

5 minutes ago, Shadders said:

The number of engineers that understand sampling etc., in other engineering domains such as RF, communications, and other areas of electronics, dwarfs the audio industry.

 

I was specifically targetting engineers outside of audio. I know many of these. I am two of them myself. Specific branches of RF and comms left aside, trust me that many do not properly understand sampling in all of the nuances that matter in audio. That is because of lacunes in education, and because of the little fact that sampling in many disciplines is resource-constrained and hence is implemented via shortcuts and approximations that have become totally embedded in thinking. Audio is about the only discipline where everything can be done by the book. But then you have to read the right book.

[firedog]

4 hours ago, Lee Scoggins said:

 

I don't think you are seeing my point.  Labels feel 1. that streaming is the future and 2. MQA is an important component.  So my guess is that they don't really care about sound quality so much as getting more people paying them a monthly annuity to do streaming.  For all I know, they may just be using MQA from a marketing standpoint leveraging the sound quality to attract interest.  I have heard that mastering teams at each label spent a good bit of time evaluating and approving the encoding process.

 

Not every track will have a 24/96 source file.  MQA may be advantageous as it can enhance 24/48 and 16/44 files as well.

 

Also, you make the assumption that bandwidth is plentiful.  In my discussions with people in the industry such as David Chesky and Ken Forsythe, bandwidth at the scale of streaming is a very real issue. 

Also, you make the assumption that bandwidth is plentiful.  In my discussions with people in the industry such as David Chesky and Ken Forsythe, bandwidth at the scale of streaming is a very real issue. 

 

 I made no such assumption. Again, you aren't paying attention. It's already been shown that MQA files are not smaller than FLAC files of equivalent bit depth and resolution, in most cases certainly everything 24/96 and under. That's b/c MQA files don't exceed 17 bits depth in actuality, and have less high resolution bits. So you could make a properly dithered 18/96 FLAC file from a 24/96 master - it would have all the content of the master and be smaller than the MQA file made from that same master.

 

I must say at this point it's very difficult to try and have an intelligent discussion with you. You seem to ignore facts and analysis that don't fit in with some preconceived notion you have, and keep falling back on what you were "told" in "discussions". Has it ever occurred to you that some of what you were told isn't accurate and needs to be critically examined? Just because someone you know in the industry tells you something, that doesn't make it factual. 

[firedog]

2 hours ago, Norton said:

To repeat my previous question, other than Tidal/MQA, who is actually offering 96 kHz classical music streaming today?  I looked at Qobuz for example, as far as I can see from their site, the implication at least is that anything beyond RBCD is only offered as a purchased download.

 

The 2l site shows that a MQA a file is c. half the size of 24/96, presumably that has something to do with it's "stream-ability" ?

 

I am coming at this from the assumption that,  in the particular situation where both the MQA source file and the original non-MQA music release are at 96khz and the MQA stream decodes  at 96kHz to my non-MQA DAC,  then what I am getting is no less than what a (theoretical) non-MQA streaming service or a 24/96 download would offer at the same resolution.  

Nope.

Qobuz and others stream hi-rez. No MQA.

MQA files are not actually 24/96. They are 15-17 bits in depth and less than 96 in resolution- that's because the MQA folding process throws away the parts of the file MQA considers not important. 

So you can take a 24/96 master, reduce it in size to 18/96 and it will include more information than the MQA file, and also be smaller when it is compressed to FLAC. 

 

I'm glad you like the sound of MQA. I'd guess what you are hearing is the sound of their filters, or maybe some artifacts of their process you find appealing. Nothing wrong with that. But you aren't hearing the sound of the master. 

[Shadders]

27 minutes ago, Fokus said:

 

I was specifically targetting engineers outside of audio. I know many of these. I am two of them myself. Specific branches of RF and comms left aside, trust me that many do not properly understand sampling in all of the nuances that matter in audio. That is because of lacunes in education, and because of the little fact that sampling in many disciplines is resource-constrained and hence is implemented via shortcuts and approximations that have become totally embedded in thinking. Audio is about the only discipline where everything can be done by the book. But then you have to read the right book.

Hi Fokus,

We will have to agree to disagree.

In essence, the fact that the MQA DAC for 192kHz sampling has half the bandwidth empty, and MQA are claiming that you are receiving the entire 96kHz bandwidth, is fraud.

As per the people here that have discovered this, the advertising standards authority should be notified how to discover this aspect, and they can then employ their own engineers to replicate and prove it, independently.

Regards,

Shadders.

[Fokus]

1 minute ago, Shadders said:

In essence, the fact that the MQA DAC for 192kHz sampling has half the bandwidth empty, and MQA are claiming that you are receiving the entire 96kHz bandwidth, is fraud.

 

It is not fraud, because MQA do not explicitly claim that they convey the entire 96kHz payload band of a 192kHz original. What they do claim is that they convey the sub-48kHz part plus an overall impulse response commensurate with the original 192k (or 384k) sample rate. The Blue LED is there

1) to tell you the original rate

2) to assure you that no-one has tampered with the file between creation and delivery.

 

Look at it this way: the people behind MQA are rather smart (really). They are smart enough not to claim something that can be refuted easily for a jury of relative laypeople.

[Shadders]

3 minutes ago, Fokus said:

 

It is not fraud, because MQA do not explicitly claim that they convey the entire 96kHz payload band of a 192kHz original. What they do claim is that they convey the sub-48kHz part plus an overall impulse response commensurate with the original 192k (or 384k) sample rate. The Blue LED is there

1) to tell you the original rate

2) to assure you that no-one has tampered with the file between creation and delivery.

 

Look at it this way: the people behind MQA are rather smart (really). They are smart enough not to claim something that can be refuted easily for a jury of relative laypeople.

Hi Fokus,

From the MQA website regarding the 2nd and 3rd unfold (MQA Full Decoder) :

 

"Products with a full MQA Decoder unfold the file to deliver the highest possible sound quality. At this level of playback you are hearing what the artist created in the studio – with precise file and platform specific DAC compensation and management. Partners include Aurender, Bel Canto Designs, Brinkmann, Cary Audio, Meridian, MSB, Mytek, NAD and Technics. "

 

You cannot hear what was recorded in the studio for the 2nd and 3rd unfold if that bandwidth is empty. It is no different from the 1st unfold apart from the sampling rate. Hence fraud.

 

Regards,

Shadders.

[Norton]

47 minutes ago, firedog said:

Nope.

Qobuz and others stream hi-rez. No MQA.

MQA files are not actually 24/96. They are 15-17 bits in depth and less than 96 in resolution- that's because the MQA folding process throws away the parts of the file MQA considers not important. 

So you can take a 24/96 master, reduce it in size to 18/96 and it will include more information than the MQA file, and also be smaller when it is compressed to FLAC. 

 

I'm glad you like the sound of MQA. I'd guess what you are hearing is the sound of their filters, or maybe some artifacts of their process you find appealing. Nothing wrong with that. But you aren't hearing the sound of the master. 

 

Judging by the exemplary resolution of low level detail I'm hearing, the sound of their filters and artefacts must be quite something...

[mansr]

22 minutes ago, Fokus said:

MQA do not explicitly claim that they convey the entire 96kHz payload band of a 192kHz original. What they do claim is that they convey the sub-48kHz part plus an overall impulse response commensurate with the original 192k (or 384k) sample rate.

What's the difference?

[Fokus]

12 minutes ago, Shadders said:

You cannot hear what was recorded in the studio for the 2nd and 3rd unfold if that bandwidth is empty.

 

But you also cannot hear that if that part of the spectrum is not empty. When will you see that all of their claims are in the subjective realm, and thus nearly impossible to attack?

 

Anyway, the 2014-2015 MQA papers explicitly state that humans do not hear ultrasonics, that the ultrasonic sound can safely be removed, and that this operation can be called 'subjectively lossless'.

 

 

[Fokus]

2 minutes ago, mansr said:

What's the difference?

 

The avoidance of anti-aliasing filters, knowing that the original signal has next-to-no content above fs/2.

[Shadders]

1 minute ago, Fokus said:

 

But you also cannot hear that if that part of the spectrum is not empty. When will you see that all of their claims are in the subjective realm, and thus nearly impossible to attack?

 

Anyway, the 2014-2015 MQA papers explicitly state that humans do not hear ultrasonics, that the ultrasonic sound can safely be removed, and that this operation can be called 'subjectively lossless'.

 

 

Hi Fokus,

I disagree. Their web site explicitly states the 2nd and 3rd unfold allows you to hear what the artist created in the studio. If that 2nd and 3rd unfold provides no new information, then this claim is fraudulent.

Regarding the MQA papers, if they state that one cannot hear the ultrasonics, then claiming on their website that they provide the listener with the experience as per what the artist created, then this is plain fraud. Their paper does not support their advertised offerings.

Regards,

Shadders.

[FredericV]

1 hour ago, Fokus said:

Anyway, the 2014-2015 MQA papers explicitly state that humans do not hear ultrasonics, that the ultrasonic sound can safely be removed, and that this operation can be called 'subjectively lossless'.

 

This means the first unfold has no purpose. The first unfold just recovers one more octave in the ultrasonic range, which we can't hear. Any further unfolds do not recover any more music data.
 

So what remains are MQA's leaky upsampling filters (which are similar to Ayre's implementation, they all violate Nyquist and allow aliasing) with one cycle of postringing, which can be duplicated with sox.

Undecoded MQA still has enough bitdepth compared to decoded MQA. So if we can't hear the ultrasonics, you can mimick MQA's sound with minimum phase upsampling + one cycle of postringing settings.

This filter makes the bass more tight, attacks kick more, which can be fun with electronic and EDM, but these filters also lack a sense of depth, and decays are shortened. After experimenting with archimago's intermediate phase filter, the MQA filter certainly is not a one-size-fits-all filter. Archimago's filter sounds a lot more fluent, with a lot more depth.

So congrats that an independent researcher can develop the parameters for a filter that blows away MQA's filter:

http://archimago.blogspot.be/2018/01/musings-more-fun-with-digital-filters.html

 

[asdf1000]

3 minutes ago, FredericV said:

Undecoded MQA still has enough bitdepth compared to decoded MQA.

 

So are we saying that undecoded MQA is not worse than CD quality?

 

Not better but not worse?

 

 

[Don Hills]

1 hour ago, Norton said:

 

Judging by the exemplary resolution of low level detail I'm hearing, the sound of their filters and artefacts must be quite something...

 

Indeed. Note that it proves you don't need 24 bits to get "exemplary resolution of low level detail" - 15 to 17 bits is quite adequate. But regardless, it isn't the sound that was heard in the studio by the artists and producer who signed off on it. (*) We won't get that until we start getting music which was recorded from scratch using MQA enabled ADCs.

 

(*) The original music was auditioned and approved using standard PCM ADCs and DACs. They tweaked it until it sounded the way they wanted. You need to use a "standard" PCM DAC to listen to it. Processing it later with MQA may make it sound different, and you may prefer it, but it's not what was originally heard and approved.