MQA is Vaporware

[soxr]

I came to the conclusion that in case of the available 2L.no test tracks which exists in a 352.8K real DXD studiomaster file and their decimated MQA counterparts, it's now possible to derive a file using the MQA file as only input, that sounds indistinguishable from the DXD original studio master that was used to encode this MQA file.

I used the open source SOX tool. No code from MQA was used to do this.

Anyone can peer review this research as I decided to open source my findings. Archimago, this topic and several gearslutz's MQA topics were a great help to come to this method:

I have been following these topics for months, but now the time has come to disclose my findings.

[soxr]

4 hours ago, crenca said:

 

Ok, I will be the bad guy  Your just making a "sounds like" assertion.  I could do the same thing with any DXD file, "decimate" it to 320mp3, use Sox to get a DXD file, and then claim it "sounds indistinguishable" and of course I would be right.

 

All of this is quite besides the point...

No it does not work like that. Minimum phase does not introduce pre-ringing, and does not accumulate it, so is basically a band limited lossless version of the original with in case of MQA limited to a dynamic range of approx 102dB.

Fully decoded MQA (with one of the 32 correction filters applied) does not even sound like the master, it sounds different. More echo and reverb than the original. Calling this "authenticated" and saying it sounds exactly like what the mastering engineer heard .... marketing lies. I bought a mytek to be able to test these claims.

Undecoded MQA contains 17 bits of resolution + dithering noise which contains a lossy version of 20-40 Khz with very limited dynamic range (this part is like MP3 but we ignore it). The 0-20 Khz band should give 102dB dynamic range.

I don't know many systems that can fully exploit this range within a normal living room, normal sized speakers and neighbours. Futhermore, as this file has been created with minimum phase, transients are very similar to the master, but band limited. This is not like MP3. The proprietary lossy part of the MQA stream is below the noise floor and not decoded, containing data that on normal systems will be below the noise floor of your system. Maybe with PA gear you can reach noise floor of room + 102dB, but with most consumer hifi: forget it. You would still go deaf anyway trying this.

I agree that you are missing the lossy 20-40 Khz content which has been limited in dynamic range, but I challenge that you can even hear this with our sox method, as you will not be able to tell the DXD vs upsampled MQA via sox apart.

The audible part of the undecoded MQA flac file is not MP3 but lossless.

It is like the 16/44.1 part of our little DVD-A test, which could be upsampled back to 24/96 and nobody except one could hear the difference between original 24/96 and band limited to 16/44.1 and upsampled back to 24/96.

But that only works with minimum phase. When using linear phase, the pre-ringing errors would accumulate with each step. Our method works because MQA has chosen minimum phase.

This link is also useful:
https://arstechnica.co.uk/gadgets/2017/05/high-res-audio-guide-mqa-explained/7/

MQA does not want to comment on minimum phase vs linear phase, see the end:

"My final question to MQA Ltd about its balance of priorities, between preserving phase linearity and achieving a compact impulse response, remains unanswered "

I did not find any post where MQA / Bob acknowledges that they use minimum phase. It was on fora like this, gearslutz and archimago that we figured it out.

 

[soxr]

On 30/6/2017 at 6:17 PM, mansr said:

The entire premise of MQA is that high resolution audio is superior to 44.1 kHz sample rate. In other words, there is an implied assumption that ultrasonic frequencies are in some manner perceptible. If we accept this as valid, is it not then strange to start dumping garbage at those frequencies? The only way this distortion can be deemed harmless is to concede that ultrasonics are inaudible, but then what do we need MQA for in the first place?

Yesterday I was trying to determine the noise level of the "secret" closed standard MQA metadata embedded in their 24/44.1 distribution files from 2L.no via audacity. I was looking at ways to make it audible.

2L-053_04_stereo-DXD is interesting as it has room noise and it takes a second before the music starts playing. In the first second you can hear the musician take 2 breaths, but before that it's dead silence if you play at normal listening volumes. But the waveforms don't look like silence.

So I cropped the silence out of the DXD version and the upsampled MQA version and started listening in loop mode. I had to boost the volume of both by running the amplify effect (approx 35dB before clipping) and then I also boosted the gain slider for the two tracks until it started to distort, which was +5dB so I set it to +4dB, so total gain approx 39dB (almost 8.000 times louder).

I listened via my internal laptop speakers for this test, as they also tend to exaggerate high frequencies. I discovered the double breath taking in 2L-053_04 thanks to my laptop speakers, which I never noticed on my real system before. Now that I know it's in there, I hear it every time. So our hearing can be trained. It's like errors in video codecs and video processing: once you notice them, you will see them every time. So crappy laptop speakers still have their purpose

I put audacity in loop mode and started pressing the solo button of each track: upsampled MQA and switching to DXD and vice versa.

It largely confirms this video:

Both boosted noise parts sound very similar, where the room noise of the upsampled MQA version and DXD sound identical, but MQA adds a little bit of HF noise. The added noise of MQA does not sound like broadband noise, but like high frequency hissing.

It's still subtle but once you hear it there is no way back. But as we had to boost the background noise by insane levels, I believe this will be masked under normal playback.

So I'm questioning the audibility of 24 bit resolution. I have so many tracks in 16/44.1 which are recorded so well that listeners during a demo will ask: is this DSD? is this high-res? No it's just redbook.

I'm pretty sure MQA knows what dynamic range in a typical audiophile room will be obtainable because they know the noise floor of the average room and the dynamic range of typical recordings, and encodes their "secret" format that contains the unfold data and upsample parameters of the renderer into those LSB bits that are below the typical noise of the room. They also describe in their AES patent that there is no content just above 40 Khz ~ 50 Khz, which is the reason MQA only recovers a 88.2K or 96K samplerate after the first unfold and the second unfold is just upsampling that adds no new content.

It's also noteworthy that most MQA dac's don't offer true 24 bit resolution, and most don't even specify the noise floor. Bluesound is advertizing "24 bit in every room" while the SNR of their node2 is -110dBA = 18 bits. The new AudioQuest is not any better, these cheap entry level devices are not capable of 24 bit. Not even 20 or 21 ENOB like many older 24/192 DAC's. False advertizing.

Mytek does not specify the SNR either, which is strange as it is a studio product:
https://mytekdigital.com/hifi/products/brooklyn/

So I have serious doubts if MQA is able to restore the true 24 bit resolution on any delta-sigma dac. I doubt it. ESS is typically -126dB ~ -129dB, so max 21.5 ENOB.

One bit in the MQA chain is already lost for the signalling between the output of the core and the second unfold (renderer), so I believe it is correct that best case MQA can do 23 bit? But does their closed format unfolding data contain enough entropy to unfold to the source 24 bit material? I doubt that.

The fact that upsampled MQA (using minimum phase) which includes some extra added HF  noise below the noise floor of a typical audiophile playback room sounds (almost) indistinguishable from the DXD from which it was derived, confirms the 24 bit audio myth for me.

While there may be very small differences on extreme high-end systems, you will need to listen very focused and  it still will make you doubt. I did that the last couple of days. This is what the feedback was of my shootout with several listeners, tested on a more extreme system that most MQA dac owners will have.

The amphion krypton III system is my workhorse for product development, and my other systems with real studio monitors are not so extremely revealing and level out these differences that make you guess without being able to make consistent conclusions.

[soxr]

19 hours ago, mansr said:

The actual MQA data is in the bottom 9 bits. The remaining PCM audio is heavily dithered with noise shaping creating quite a hump up against the 22/24 kHz limit. This dither noise is probably what you heard.

 

I wonder what will happen with MQA CD. So if something like LSB 3 bits are assigned to the secret MQA data, what entropy does that contain to even restore something highres?

There is one engineer doing such releases:
https://www.gearslutz.com/board/12716272-post1049.html
 

Quote

In fairness no DAC reaches full 24-bit resolution. The very best chips manage 132 dB, at least if measured in just the right way.

 

I own the Metrum Adagio, an R2R NOS DAC, but that's an exception as it's not a chip DAC but using custom built modules by Metrum. Adagio does true 24 bit SNR just like some overpriced MSB dac's. But no delta-sigma is going to cut it and reach 24 bit SNR in a chip.
 

Quote

It can't. There is no way to store that amount of information in the space allotted. Besides, even if it did, the renderer dithers it down to 20-bit at best, 16-bit in all actual files I've seen.

I wonder where they get the 20 bit resolution, if 9 bits in their baseband file are lost to encode the HF spectrum, they only have 15 bits left. Unless off course some of the secret MQA data also contains audio via dithering, so the "apparant" dynamic range can be a little bit higher, just like correctly dithered CD can do more than 96dB dynamic range.

If dithering is partially correlated to the audio data, that means they don't have 9 bits of entropy for the HF part.

[soxr]

Someone sent me this link:
 

https://www.youtube.com/watch?v=Cdlmp7zpXLM

 

An interview with none other than Bob Stuart, at the time of DVD vs Blu-ray/HD-DVD, where he debunks the need for super formats and high-res, and also states:

 

"We support DVD Audio. But the most important thing I think we have to do, is to produce the sound from CD, which is so good, and so exquisite, that the difference to high resolution is very small."

 

He also states "CD is good enough for 95% of the people" and laughs at audiophiles and their private recordings / special audiophile versions of the same recording.

 

So why did Bob make a 180 degree turn?

[soxr]

21 hours ago, Rt66indierock said:

I’ve discovered is MQA Ltd is recruiting mastering engineers to be area representatives. So when you hear a mastering engineer support MQA you have to ask two questions. One are they an area representative and did they receive equipment or other forms of compensation to say what they did.

 

So MQA is hiring paid shills.

I also believe MQA is paying reviewers to debunk the MQA criticasters, example:
 

 

Then there some extreme fanboys, who are on every forum, blog, facebook group, always hammering on how good MQA is and why we need deblur and fixing the timing errors, which were never an issue before MQA.

It's like these people have a fulltime job just promoting and pushing MQA and battling with proponents. They always deny any affiliation with MQA, but they post comments like "good work, team" ....

Nomal people don't do that, unless they are part of the team. One of them is also member here.

 

21 hours ago, Rt66indierock said:

Finally, opponents of MQA have formed a group to collate, organize and disseminate information about MQA. The group includes people from high end audio manufacturers, major labels, studios and me.

MQA pushers claim this group fears MQA. To the contrary, I think this group can cut through the marketing BS, and figure out that the A in MQA is a marketing lie, confirmed by a lot of real studio engineers not paid by MQA.

The fact that MQA is now hiring engineers shows how desperate they must be.

On Linkedin, MQA is spreading their usual marketing lies, e.g.:

"MQA is an award-winning British technology that delivers the sound of the studio in a file that’s small enough to stream or download. "

Studio's don't master for MQA. They don't even have the tools to do that. All encoding is done at MQA, not in realtime in the studio. MQA seriously alters the master. So MQA is not the sound of the studio. Just follow engineers such as Brian Lucey on Gearslutz and it is now confirmed several times that MQA is not the master.

 

[soxr]

Mytek Brooklyn vs Asus Essence STX II

 

Archimago has posted measurements which now prove what I have been hearing in the lab for months.

http://archimago.blogspot.be/2017/07/measurements-mqa-filters-on-mytek.html

Basically what MQA tries to do is to kill post-ringing via a programmable filter in the ESS chipset, which seriously alters the sound. On cheap devices like the Dragonfly, this filter is only active when playing MQA. On the Mytek Brooklyn, this filter is ALWAYS active!

 

Now with the Mytek Brooklyn that I own, there's something strange going on. When I play through an Asus Essence STX II sound card in a linux box powered by a 12V linear supply, for which the Asus card does 24/192 max via PCM 1792A, and connect that card to a very high-end amp on studio speakers, and also run SP/DIF from this card to the Mytek and run a second analog cable set to the amp, I can do shootouts between the 220 € Asus card and a 2000 € dedicated DAC by switching the input on my amp.

 

No matter what we do to the Mytek, like adding a very decent linear 12V PSU, or whatever cable we try: the Mytek sounds so much less than the cheaper Asus card. It's less musical. This was very frustrated and we did not know why. This has been confirmed by local members of the hifi press which were present during the shootout.

First we believed that MQA is degrading NON-MQA on MQA dac's, to make the difference bigger. But now with the measurements revealed, it actually confirms that the MQA efforts to kill post-ringing just do not sound right.

 

I bought the Mytek to do some validation testing, but never suspected it would sound less than a 220 € Asus computer soundcard.

 

I hope Mytek offers some option to disable this MQA alike filters on normal PCM.
I certainly do not like this.

 

For me this is blind evidence that something is wrong with MQA's filter choice, as I have been annoyed by the sound of the mytek for weeks now, because you would expect that a 2000 euro standalone DAC would sound better than a 200 euro soundcard.

 

Other ESS dac's such as auralic vega don't have these problems, so it's not related to the ESS chipset, but because of MQA's filter choices, which seem to infect non-MQA playback as well.

[soxr]

4 minutes ago, AJ Soundfield said:

Maybe you just got so used to "smearing" all your life, that you now prefer it.

 

All audio playback colors and thus smears when playing through speakers in a typical room. With our without MQA.
 

4 minutes ago, AJ Soundfield said:

I'm holding out to hear every artist in history confirmation of authenticity.

 

Dream on.

[soxr]

4 minutes ago, mansr said:

IIUC, the Mytek offers other filter choices if MQA is disabled. If you're playing non-MQA content, you're probably better off disabling the MQA decoder and using one of the other filters that then become available. It's a bizarre design for sure.

 

H'mmm need to check if my MQA decoder was active, I think I once disabled it.

 

4 minutes ago, mansr said:

Do you by any chance have a decent recording interface capable of 192/24 or better?

I have two very old but good Onkyo PCI 200 SE LTD cards which I used in the past for audiodiffmaker.

[soxr]

49 minutes ago, mav52 said:

 

Trouble is AJ, Stuart is more famous than all of us

Coca Cola is also more famous than most other soda brands, even the organic soda brands that nobody knows about.

That does not make it the most healthy soft drink choice.

[soxr]

4 hours ago, Rt66indierock said:

Finally you need to reread what you quoted. Not everyone at the majors is supporting MQA. In fact many oppose it.

This story is very similar to 4G in my country. One foreign provider jumped on the license and paid insane amounts to get it. This provider has done nothing with his assigned 4G frequencies, except hold the license so that no other party can occupy the spectrum, just to be ready when they want to be ready.

Well it never happened ... so signing a deal does not mean that the license will be used.



 

[soxr]

3 hours ago, Rt66indierock said:

 

Considering what can be done in software I see endless opportunities in the near future for you to review, compare and write how to articles. This is type of software is only beginning. I seem to remember writing about wanting control of the filters a while back.

The good thing about MQA is the fact that the competition also starts to think about improving or changing filter design (e.g. minimum phase vs linear phase), ways to improve compression (e.g. blank noise bits in flac so the entropy encoder has less work to do = smaller files), ways to deblur pcm while not licensing anything from Bob.

This whole MQA discussion was interesting as I learned a lot in the process. It made me test ideas that were on my todo list, but faster than what I was planning.

This drives competition which is good. I'm happy we don't need to license MQA decoders but we can process MQA in a different way and we can benefit from the lessons learned and also apply that knowledge to non-MQA files


 

[soxr]

1 hour ago, esldude said:

MQA is really its evil twin AQM.  Authenticated Quality Missing.

 

or Missing Quality Assurance
http://www.6moons.com/industryfeatures/mqa/1.html

 

Quote

On the technical side, we requested their decoder to develop proper unit-to-unit QC protocols as we have always done on our end. They only promised a few test tones. Those are insufficient to measure distortion, bandwidth, impulse response, linearity and noise to mention just a few. They seem to be very afraid to divulge more about their algorithm to us, their intended business partners. If we can't properly test our MQA-enabled product, how can we be confident to ship it around the world?

 

[soxr]

44 minutes ago, Shadders said:

Hi,

Yes - my interpretation is that the master file you receive has had the de-blurring already implemented. The filters in each device is an inverse filter (MQA requirement in code/hardware) of the DAC IC FIR filter (assumed FIR) such that the smear from that FIR filter in the DAC, is minimised.

Deblur is done in the renderer. The original master waveform (eg DXD) vs upsampled decimated MQA with minimum phase: no evidence that the decimated file has been pre-processed. But it contains a metadata field so that the renderer knows which one of the 32 filters to use.

[soxr]

Just now, mansr said:

What is that figure supposed to show?

See my other topic:
 

"studiomaster" is one the original 2L.no files
The other is the related mqa file with sox rate -vsM 352800

 

[soxr]

1 minute ago, PeterSt said:

 

Hi soxr,

 

I have been following your work via your reporting about it in CA. Great effort you did there.

But do notice that showing /observing / analyzing a graph at this zoom level can never tell you anything.

Also, by these means (and in that other thread) you claim that there's no difference in the upsampling vs what MQA makes of it. And without looking I tell you : no sample is the same.

 

I zoomed in much further than this, and all I see are ultrasonics in the DXD file which are not in the MQA file.

MQA file also has HF hissing noise, but you need to boost a quiet passage with 45dB and turn up the volume to extremes to hear it. This part is the lossy HF frequency spectrum which goes to x2 rate, so the lossy data for the first unfold.

I don't see the plot of the waveform being seriously altered to make the transients sharper.
 

1 minute ago, PeterSt said:

 

Additionally, upsampling vs what MQA does not sound the same at all. It even can't as the upsampling filters are never the same (so they already sound different among each other). And if you claim you can't hear the difference, when you are in the neighborhood, you are welcome. 

I have an MQA dac. I don't like the sound of MQA. so I rather buy DXD files or real lossless PCM.
So what I want is highres without the "thinning" process that MQA applies via it's renderer.

My current DAC has a perfect impulse response, so what MQA would do on that DAC is a downgrade. Don't fix what ain't broken.

There are several brands that implement MQA not to risk losing any business, not because they like MQA.

[soxr]

10 minutes ago, Digital Assassin said:

This is why you hear 95% of MQA zombies talk about software unfolding only unless they have a Meridian or MyTek DAC.

 

Furthermore, it was always claimed by MQA that full unfolding was not possible in software and you needed a DAC for this, which has been debunked by what's inside the bluesound library. The BS library that is an .so file (shared library) contains both the software for first unfold and renderer stage.

So Bluesound does full unfolding in software. It basically debunks the need for an MQA dac.
Also some of the fanboys claim the filters are analog and MQA is an analog end-to-end process, while hey listen to an NAD c390DD amp which is fully digital.

Which means MQA cheats. They do allow full unfolding + DSP on some of their own products (like the active Meridian DSP speakers) and they do also allow EQ after full MQA decoding on NAD:

https://support.bluesound.com/hc/en-us/community/posts/115007974948-MQA-and-EQ-functionality-

 

but at the same time limiting output to the first unfold in software like the tidal player and MQA software players.

So MQA maintains double standards.

 

[soxr]

1 hour ago, fung0 said:

At this point in history, it's been well-demonstrated that open standards are vital, enabling maximum progress for consumers and maximum profit for businesses. It is also obvious that open standards don't always materialize by the grace of the market's 'invisible hand.' Or survive in the face of corporate onslaught. Sometimes, consumers and experts need to speak up.

I thank my former & current career to open source, open standards & open file formats.

Allmost every player software including closed source software is based on open source decoders and/or open standards. Almost nobody is going to write their own flac decoder.

Then some company comes along and claims: all that you know about audio sampling is wrong, we have the solution. They infect open formats like flac and redbook by inserting encrypted content which can only be decoded via a licensed decoder. Without decoder it becomes high frequency noise / hissing. This is a form of hacking: hack the file formats to technically inferior versions for those who don't have the decoder, provide articles to the hifi press which are almost impossible to debunk as the technical details are hidden or simplified because the solution is a black box.

Like any proprietary format, decoders will be hacked, decryption keys leaked ... it's just a matter of time. Look at DVD and DeCSS. BluRay was hacked.

SACD was hacked because the PS3 fat models with SACD drive had the decryption keys onboard, and a custom firmware could be installed. I had some fun installing this on two old PS3's so a customer could make a legal backup of his SACD to DSD files which he could then play using open source software like MPD and open DSD decoderss. So a closed encrypted format was unpacked into an open file format, while keeping the DSD data intact.

The same applies for MQA. Input & output of MQA decoder and render stages can be rerouted to files or virtual devices. The proof is in the open source code that was revealed in the technical MQA thread. So what if technical users take matters into their own hands?

Output of the first unfold can be sent to a much higher quality upsampler without the post-ringing trickery that makes MQA sound thinner than the original studiomaster, which audiophiles may wrongly interpret as "more air" "more stage" "more reverb" "more echo".

This can even be legal: just create a vitual audio device, and send that data to some post-processor. Alsa can do that easily. You can create slave devices which write to a pipe. Under windows, such solutions can also be developed, just look at dirac live, which uses a similar setup.

I took a different route: instead of trying to crack MQA decoders, I tried to figure out what gives MQA it's distinct different sound compared to the real studiomaster DXD files and figure out how we can benefit from the smaller MQA files without hearing a very different rendering compared to what the studio engineer was working on before sending his files to MQA for offsite encoding.

Studio engineers do not listen to MQA in realtime. Even though this is claimed on the MQA site, actual discussions with real studio engineers on gearslutz proved there are no realtime tools yet. Several GS engineers provided a lot of info and exact logistical workflows, and for the technical side we have to thank mansr and archimago to cut fact from fiction and expose the inner workings of the black box that the hifi press cannot grasp.

The problem with the hifi press is that they just copy paste the official versions, or do not understand them, which leads to video's like this, where the admit they were wrong:

This is why I became member here and decided to disclose the sox method so others can benefit from my research. I also became member because I want to have obective fact vs fiction, and I do not take the canned material from the MQA marketing for granted.

There needs to be a counterweight to bring everything in balance. I am part of that counterweight.
I'm not 100% against MQA, as I can still benefit from MQA smaller file sizes which are free on Tidal (I don't have to pay extra for MQA) and throw a minimum phase upsampler at it, and hear virtually what the mastering guy was working on, not the MQA guy in some remote encoding facility which makes the choices of encoding parameters.

[soxr]

8 minutes ago, Fokus said:

And how would you go about verifying a 'drastically improved impulse response' without access to undoctored-with before and after files?

 

Even manufacturers don't get test tones from MQA so they can't verify transient reponses and do basic QA testing:
http://www.6moons.com/industryfeatures/mqa/1.html

 

Quote

On the technical side, we requested their decoder to develop proper unit-to-unit QC protocols as we have always done on our end. They only promised a few test tones. Those are insufficient to measure distortion, bandwidth, impulse response, linearity and noise to mention just a few. They seem to be very afraid to divulge more about their algorithm to us, their intended business partners. If we can't properly test our MQA-enabled product, how can we be confident to ship it around the world?

 

8 minutes ago, Fokus said:

It would have been so easy for MQA to promote their process by putting up a website with short before and after clips. That is, if their invention amounted to anything at all...

 

They won't do this. But you can do this blind test:

http://archimago.blogspot.be/2017/07/internet-blind-test-mqa-core-decoding.html

The transient responses of the MQA bluesound renderer, Mytek and AQ Drangonfly are known and seem to be the same. So far no proof of DAC specific tuning. The renderer is a minimum phase upsampler (so yes the 2e unfold is upsampling, no new information is extracted) with some post processing to kill post-ringing.

It's also possible to do the full unfold in software. Custom tools were written by FOSS devs.

So those who claim nobody is interested to verify the claims from MQA are clearly wrong.

 

 

 

[soxr]

22 minutes ago, PeterV said:

Well Fokus..let's ask MQA themselves, but also find a way to do proper A/B comparisons in the analogue domain. Is this possible or impossible? Even though loudspeakers and microphones have their own intrinsic (time-smear) flaws, why not test, measure, compare? 

Why are you changing the subject? As soon as a mastering engineer applies reverb or some similar effect, or edits a recording of instruments in a room or hall with natural reverb, the original signal from the instrument is combined with reverb / echo's and other distortions (either natural or electronically), smearing the audio. It is this smear that allows the ear to decode the real or virtual positioning info. It is this smearing that gives certain bands their signature sound. Imagine Metallica without guitar pedals? Without that fat sound?

All time smear in MQA products is "killed" using one of 32 predefined filters in the MQA render stage:

[soxr]

14 minutes ago, PeterV said:

I am not interested in measurements or graphs or reversed engineered software programs. All I care about is that MQA is contributing positively to the sound and that is just my personal observation.

 

All you care about is confirmation bias, almost in an evangelistic way. Any independent research you overlook and reject. At the same time you want the time smear proven and always hammer on your fallback argument.
 

Quote

What I want to know is if this is due to the improvement in impuls response and as a result also to the reduction of time-smear in air.

 

 

How are you going to do this without measurements? You contradict yourself.
 

Quote

If my ears are capable to notice this en MQA claims are very clear about the reason and call this a 'paradigm shift' then all I care about is to get professional and objective confirmation. 

You want OBJECTIVE proof but you reject measurements and any independent research? You contradict yourself.

This is not very scientific, as the scientific method requires peer review and those who make extraordinary claims, to prove them and allow others to duplicate it

[soxr]

3 minutes ago, Fokus said:

That is not a review but a regurgitation of MQA promo material.

 

It's one of his fallback articles. So the discussion always goes around circles.

 

[soxr]

14 minutes ago, PeterV said:

I will take the challenge immediately to do A/B comparison just by listening

 

http://archimago.blogspot.be/2017/07/internet-blind-test-mqa-core-decoding.html

[soxr]

50 minutes ago, Fokus said:

 

The irony...

 

MQA claim that the ringing of ANTI aliasing filters is detrimental to sound quality. They tackle this by introducing aliasing effects in their masters. The opposite of what you stated.

 

They trade mistake A with mistake B. A trade-off. It's in their AES paper. In the hope that one is audible and the other not, without providing scientific proof, just anecdotal proof. Smoke and mirrors.

 

50 minutes ago, Fokus said:

MQA solves a non-problem.

 

I just talked to the CEO of a well known audio company. Confirmed that it is a non-problem.

 

[soxr]

49 minutes ago, Fitzcaraldo215 said:

Actually, there are multiple AES papers outlining their theories over a number of years, all peer reviewed, by the way. Now, that does not mean that peers attempted to dublicate experimental test results or that peers agree with all conclusions.  But, many of those earlier experimental results were done by others, and that research is meticulously cited, and those citations will have been checked via peer review for consistency.

 

Yes, the papers cite some MQA experimental results, and those experiments may not be fully described,   But, I do not recall seeing any simple anecdotal result cited.  I do not think that would have passed peer review.

 

"There is also upward aliasing introduced by the reconstruction process: here we rely on plausibility arguments, verified by listening to the final result that these alias products, lying above 48 kHz, are inaudible and low enough in level to avoid slew-rate or other problems."

So MQA introduces new mistakes (aliasing), which they claim are inaudible, and they provide only anecdotal proof. How did they test that? Did they use double blind methods? It's not documented.