MQA is Vaporware

[Jud]

Here are the frequency responses of the 16 (I misremembered 17) MQA filters. These are taken directly from the filter coefficients as determined by a logic analyzer applied to the internals of a DAC:

 

[Jud]

I had mentioned previously that MQA processing might result in compressed dynamics. Looking back through old MQA posts, I found measurements of an average of 3dB compression vs. non-MQA, with a maximum of a 10dB(!) difference. It is important to note this is not the result of remastering, but of MQA processing.

 

Thus MQA processed music would have noticeably louder quiet passages, and would sound quite different for that reason alone, even disregarding all else that we know.

[Jud]

2 hours ago, AudioDoctor said:

@Fx Studio I am sure you're aware of this, but choosing to ignore it, but...  There is extensive research into MQA done and all of the results are available here on this website. All you have to do is as Jud suggested, go read it. However, as I suggested at the start, I am sure you're already aware of that.

 

We've pretty much laid out everything he demanded (and more) in the last couple of pages of this thread.

[Jud]

2 hours ago, Shadorne said:

Yes, of course it compresses, this is simply the result of apodizing. I have pointed this out before but most folks don’t understand the mathematics of apodizing. 

 

When it comes to filtering I have a few broad mathematical concepts, but certainly don't "understand the mathematics of apodizing." So my first request would be if you know of something on a site like https://src.infinitewave.ca/ that would visually illustrate what you're saying, that would be very helpful to me.

 

Second, one of the surprises regarding the MQA filters that was remarked on in the technical thread is that they really don't cut sufficiently at low enough frequencies to work as classic apodizing filters. So how would that work with the mathematics of compression?

[Jud]

3 hours ago, Shadorne said:

By definition an apodizing filter will reduce resolution. This means peaks are reduced and broadened. This is a mild form of compression. You can tell by looking at the half height width of the impulse response. Apodizing filters are always broader as higher frequencies are removed. 
 

On infinitewave website compare Apple Tiger Vs Apple Leopard to see how much compression occurs from apodizing versus brickwall.

 

Difficult for me to tell any difference in width from the graph, but it's evident the peak is lower with Tiger.

 

What aspect exactly of apodizing filters results in this compression, because as previously noted in the technical thread there was surprise that the MQA filters did not in fact appear to be apodizing?

[Jud]

30 minutes ago, Tsarnik said:

IIRC, the encoder patent doesn't comprise a dynamic compressor.

Scrambling the lowest bits and the highest frequencies would certainly influence transient peaks but not lead to an overall gain change of quiet parts.

I would assume that said '"remastering" effect' is an effect of an actual remastering (with a real dynamic compressor) in the process of creating MQA files.

If not, how would manisandher's delta with differences down at -90dB even be possible?

 

Or maybe I am missing something here...? 

(It's been a while for me to get upset about MQA and trying to grasp every real information about it; in the past couple of years I was just quietly praying that, eventually, it will die. Happy to read that this seems more or less to have happened.) 

 

I'm thinking somewhere between the lossy compression and the need to use 1-3 bits to tell the DAC what to do instead of for music, the quiet bits may get left off. But that's sheer speculation from someone who's definitely a non-expert.

[Jud]

1 hour ago, Tsarnik said:

furthermore heavily cut the high frequencies at fc = 16 kHz (with approx. 15 dB/octave) then the resulting audio has a higher quantization error (noise, distortion) and sounds dull, but it is still not dynamically compressed, its peaks and average loudness deviating merely a maximum 0.2 decibels from the original.

 

 

This is exactly what MQA *doesn't* do (heavily cut high frequencies), which is known from the actual filter coefficients. So it can't be the result of cutting high frequencies. That's an apodizing filter, which MQA's aren't. The filters don't even cut appreciably in the low ultrasonic range (thus they can't remove ringing), let alone 16kHz.

[Jud]

2 hours ago, botrytis said:

Since we really don't know WHAT or HOW m-qanon works on a file

 

We know exactly what the interpolation filters are (that except for #15, which is seldom if ever used, they're incredibly leaky), we know about the lossy compression, we know that bits are devoted to telling DACs which of the filters to use (mostly #4 for material originally Redbook, #8 for hi res, if I recall correctly). I'd say that's quite a reasonable amount of knowledge.

[Jud]

9 hours ago, Tsarnik said:

I would assume that said '"remastering" effect' is an effect of an actual remastering (with a real dynamic compressor) in the process of creating MQA files.

If not, how would manisandher's delta with differences down at -90dB even be possible?

 

Can this be accomplished by batch processing? Because that's something else I seem to remember from these pages upon pages of information, that MQA doesn't give the vast majority of recordings individual attention. (Makes sense, since it would be rather expensive if it did.)

[Jud]

22 hours ago, vortecjr said:

I would like to see MQA or the new owners clean up the leaker filter. I know that would be admitting there was an issue, but that seems like the logical next step. This assumes a fix is even possible and I have no clue on that. 

 

That would be called "garden variety digital audio with lossy compression." 🙂

 

Since no Gibbs effect (commonly called ringing, or in MQA-speak, "blurring") was the primary reason behind MQA, and the only way that can be accomplished is with very leaky filters, or with garden variety apodizing filters that MQA has no patent on, cleaning up their filters would be the same as giving up a good part of their IP.

 

The rest of their IP is the lossy compression, which doesn't exceed the performance of plain old FLAC or ALAC.  So why use it? Easy, lossless compression is a cinch to reverse engineer. The only reason for the lossy proprietary compression was to try to preserve the rest of their IP, the leaky filters.

 

In other words, the only fix that would involve decent filters and a lossless format would no longer be MQA.

[Jud]

16 minutes ago, vortecjr said:

I would curious to see a lossless compression scheme - 16bit or 24bit at 44.1kHz or 48kHz file that contains hi-resolution content. No compression for the sake of compression. BTW I have no particular reason except to see it can be done. An engineering exercise if you will.   

 

Having purchased a classical album from Qobuz in 24/96 resolution, I'm offered this: "Additional audio formats available: FLAC, WAV, ALAC, AIFF and more." FLAC and ALAC, of course, are lossless compression schemes. I can scan this with Audirvana to see if it's "true" hi-res if you're curious, though the artist (Jordi Savall) is the label owner and I know the label is conscientious about good sound.

 

Edit: I see I've slightly misread what you're after. But why would you care whether the resolution was reduced if the file size was the same as achieved with MQA's lossy compression?

[Jud]

2 hours ago, Currawong said:

The second was the Gibbs Effect, which I believe has long been mis-represented in how it relates to in-band audio signals, since it is something that appears on extremely out-of-band signals

Gibbs effect occurs at the frequency where the filter begins cutting sharply, so in digital audio that’s the near ultrasonic range.

[Jud]

4 hours ago, vortecjr said:

4 hours ago, Jud said:

Cool...I'll check him out. The additional formats are additional purchases? 

I don’t know that they are, I just choose one. Since they’re lossless they readily convert to each other so I’m not sure what Qobuz would actually be stopping anyone from doing.

[Jud]

50 minutes ago, vortecjr said:

This is not a conversation defending them or comparing the tech to flac or being worried about storage. I’m curious if a file can contain both low res and high res both playable depending on how they are used.  Right now it’s close to possible and disagreeing won’t change the fact. 

 

Although MQA is not "low res and high res both playable." It doesn't do hi res as well as "normal" high res (because of the leaky filters and lossy compression), and it doesn't do low res well because some bits of the limited resolution have to be given over to telling the DAC which of the 16 leaky filters to use.

 

NativeDSD includes all lower resolutions in the purchase price. Streaming services like Qobuz do have many albums in various versions, both RedBook and hi res.

 

If I run Qobuz through my home system I get recordings in up to 24/192 resolution. The same albums running over my iPhone will be played at 24/48 unless I deliberately run the phone's digital output through equipment designed to avoid that. So again, I'm not seeing the problem that needs to be solved by a "two-in-one" file.

[Jud]

40 minutes ago, vortecjr said:

Lol. I’m just curious if it can be made to do what it was suppose to do. 

What it “was supposed to do” was pure marketing, so certainly MQA can’t be made to provide both a reasonable hi res layer and a reasonable lossless Redbook (CD) layer.

 

Could you do both these things in the same file with a non-MQA process? Sure, you would just have to figure out a means external to the music files that would select hi res or Redbook based on DAC capability.

[Jud]

48 minutes ago, vortecjr said:

That is irrelevant to me in this discussion. This is just a curiosity about what is possible. 

Yes, that was the second paragraph. Certainly you can package the two resolutions and compress them losslessly. For all I know the means currently exist to select the proper resolution based on DAC capability.

[Jud]

5 minutes ago, vortecjr said:

Now let’s chat about the white glove treatment:)

Not happening except for the biggest artists because it’s too expensive.

[Jud]

3 hours ago, vortecjr said:

I’m looking at this like the JVC process.
 

The gear is easy to measure and if you look at some of the gear out there some could any help they can get. We have tuning files for headphones…

That strikes me as problematic. I’ll get to the reasons in a moment, but just to review: MQA marketed selection of filters on a per-DAC basis, but what really happened is there were 16 filters of which only two were usually selected (one for Redbook source material, one for higher res, both similarly leaky).

 

Now why is filter adjustment on a per-DAC basis not a great idea?

First, the room and the speaker crossovers will have a large impact on the final sound, so concentrating on the DAC can cause as many problems as it solves. Software offering multiple filters or adjustment of filter parameters leaves the final choice to the consumer rather than forcing a choice that may not be appropriate for that listening situation. (Thus headphone tuning is a lot more common and less expensive than room EQ software, and can be much more standardized.)

 

Second, trying to package that in the music file itself makes the music more expensive.

 

Third, software maintenance for uncounted numbers of DACs is a tremendous pain.

 

So this would be both ineffective, and unattractive to consumers and labels.

[Jud]

7 hours ago, The Computer Audiophile said:

One of my law professors used to say

 

Unless you wound up doing both professions, I think the one you've chosen is the right one. 😉

[Jud]

1 hour ago, The Computer Audiophile said:

When I got to the part about Mike Jbara’s “excellent people skills” my eyes were like 😳

Here’s a reminder of my view on that from RMAF (link).

 

Wow, life is rich. 

 

 

 

 

 

JANUARY 4, 2024, CONSUMER ELECTRONICS SHOW, LAS VEGAS – The Lenbrook Group of Companies, owner of globally recognized hi-fi brands like NAD Electronics, PSB Speakers, and Bluesound and the renowned BluOS hi-res multiroom platform, has created a new business unit, Lenbrook Media Group. The new entity will more effectively enable the commercialization and development of Lenbrook’s content management solutions, as well as pursue new opportunities that fit Lenbrook’s vision to offer increased choice to music fans and enthusiasts for quality audio. With the BluOS content management platform at its core, this group will also oversee the growth of content encoding solutions MQA and SCL6 which Lenbrook acquired in the autumn of 2023.

“The acquisition of MQA’s assets complemented our existing BluOS platform and supports creators in the recording studio with capturing and delivering directly and in the highest quality, their art into the homes of music lovers. Creating the Lenbrook Media Group allows us to fully capitalize on this unique position by giving it focus, and putting the right strategies, structure, and resources in place,” explains Gordon Simmonds, President and CEO of The Lenbrook Group of Companies.

 

Lenbrook Media Group will be led by newly-appointed Vice-President and General Manager, Mike Jbara, the former MQA CEO and Warner Music executive.

“We worked closely with Mike when he led MQA and were impressed by his understanding of the complete content chain from creation through playback,” Simmonds adds. “His excellent people skills, his deep understanding of the music industry and his knowledge of music technology and licensing-oriented business activities, makes the leadership role at Lenbrook Media Group a perfect fit for both parties.”

Jbara says, “Once you combine Lenbrook’s vision for advancing choice and innovation in the performance audio industry with the assembly of some of the best minds in audio hardware and software engineering, it results in substantive opportunity to empower creators, distributors, fans, and broadcasters in addition to hardware manufacturers. It is an exciting list of possibilities that we believe will benefit both the music and the specialty audio industries.”

Jbara and Lenbrook Media Group colleague, Andy Dowell, Head of Licensing, will attend the Consumer Electronics Show in Las Vegas, January 9-12, 2024, and will be taking appointments to discuss licensing, integration, and other opportunities.

Notice also he’s still putting out the “empower creators” BS. (MQA tried to say it would help pay artists more in some way.)