Stereophile Series on MQA Technology

[FredericV]

Another article in the series:

https://www.stereophile.com/content/mqa-tested-part-1

They must be very desperate to have a paid series written by stereophile to attack the criticasters.

I don't care about their short post-ringing tail. The enforced MQA renderer even with regular PCM makes this dac sound worse than some much more cheaper but excellent dacs. One is the Asus Essence which won EISA awards and exists in standalone and computer version. Sells for a fraction of the brooklyn.

So we tested this PCIe card's analog outs and also ran an ultra expensive SPDIF cable to the brooklyn, and have a second run of analog to a second input of a high-end vitus amp. Then we switch inputs.

The enforced MQA processing was very obvious, and I had members of the Dutch hifi press present, and they did not like the forced MQA processing either. At the time we did not know 100% for sure the MQA decoder was always processing, but we suspected it. Music did not flow naturally through the brooklyn.

Also the Manhattan 2 is not as good as the mk1. mk2 has MQA, mk1 doesn't.

So I now believe MQA is deliberately crippling standard PCM when going through an MQA dac by enforcing it through their leaky renderer. I will never buy an MQA dac again. One distributor even send back his mk2 and continues to play with the mk1 as this distributor also found the mk2 harsh and sterile.

Last year we won best of show with a mytek manhattan mk1 as part of the system, exact same system with mk2 this year and everyone was complaining that it did not sound natural. So after one showday with a lot of complaints. Same system with a Metrum Adagio and Antelope Platinum + 10M clock and nobody complained.

So MQA has successfully crippled DAC's with their forced processing. It's no longer an extra checkbox, but a serious degradation of good products, which have become bad after MQA started to infect these product with their fake highres codec which also tries to fix PCM.

[FredericV]

Interesting comment:

https://www.stereophile.com/comment/570532#comment-570532

 

Quote

 

Look at figures 5 and 6 above.

 

So this one:


 

Quote

Now, open a new browser tab or window to here: https://www.stereophile.com/content/ayre-acoustics-qx-5-twenty-da-proces... and look at figure 1.

If the scale in figure 1 is hard to visualize, try page 4 here: https://www.ayre.com/white_papers/Ayre_MP_White_Paper.pdf (dated February 4, 2009)

Who says that you can't learn a lot about audio by reading Stereophile?

I'm having a hard time determining the ratio of irony to tragedy in this one.

 

https://www.ayre.com/white_papers/Ayre_MP_White_Paper.pdf

 

And this was in early 2009.

[FredericV]

AYRE / MQA / Mytek filter with sox

This thread is becoming very interesting.

Just obtained a 96 Khz impulse response testfile from another member which was used to measure MQA dacs. Thanks @mansr

I was playing with the sox filters to change the length of the postringing tail and changing the allowed aliasing parameter, which lead to the bottom plot, but not as short as the Ayre filter.



I found some extra sox recipes on the slimdevices forum which duplicate exactly the Ayre / MQA filter:



Interesting recipe here:

http://forums.slimdevices.com/showthread.php?97046-Announce-Squeezelite-a-small-headless-squeezeplay-emulator-for-linux-(alsa-only)&p=785601&viewfull=1#post785601
 

Quote

The recipe for Ayre's Listen filter is something like this "-u vM::1:28:25:180:0"

Decoding the above squeezelite upsample string, Ayre's / MQA's / Mytek minimum phase filter can be duplicated with the following sox settings:

recipe
v = very high
M = minimum phase

flags
none set

attenuation
1 = -1dB

precision
28 = 28 bits

passband_end
25

stopband_start
180

phase_response
0 = minimum phase

Filter will allow aliasing and be leaky.

So this is proof that sox can do the exact same filter as what Ayre and MQA used.

So who remembers some former members who laughed at sox

[FredericV]

Listening to the Ayre / MQA / Mytek minimum phase filter via sox. Just plotted several impulse responses to be sure.

Using Track 11 from the Nerve soundtrack by Rob Simonsen, trackname "verrazano".

My impression is that bass kicks are getting artificially tight. There's an illusion of more detail.
But with regular minimum phase upsampling, there's more body after the bass kick, but it also sounds a little bit brighter.

So again, MQA's filter kills post-ringing, making everything more tight, but if this is more musical is to be debated. Some audiophiles will kick on this, others will not like it.

I have that same experience when I listened to DXD vs MQA played on a Mytek Brooklyn. At least now I got to test this anti-postringing filter with my own music, as the 2L.no catalog is not my main genre which I like to test equipment with.

[FredericV]

I'm having a lot of fun changing sox parameters, and obtaining an at least 15x improvement in time domain resolution

17 samples vs 204 samples

 

Now that I control these, and having carefully listened for several hours, the whole time domain argument is overrated.

[FredericV]

Is sound on sound posting incorrect measurements? Do they have technical staff who understand what they publish?

So 17 samples for total impulse duration, at 192000 samples per second, leads to 88,5 microseconds.



This is also what I see in the blue curve here:

 

So why are they claiming 100µs for conventional systems and a reduction to 10µs, if the duration of the blue response is certainly not 10µs but very similar what I measured with the sox version of Ayre's filter.

And this sound on sound article is always being repeated by the MQA shills, while clearly it is wrong.

[FredericV]

Just now, mansr said:

It's nonsense, so there is no right or wrong.

 

Agree. Their whole time domain argument only applies when a signal is being resampled / oversampled. Redbook already has picosecond timing resolution.

It all boils down to misusing Kunchur's research:

http://forums.slimdevices.com/showthread.php?105070-A-Look-at-MQA&p=860274&viewfull=1#post860274
 

 

Quote

The funny thing Mnyb is that they know 16-bits 44kHz will give us time domain accuracy in the picosecond range. They seem to be basing all this on impulse response plots...

Quote

They may know it, but it doesn't stop them from quoting guys like Kunchur who started out saying that since 44/16 samples are 22 microseconds apart, it must not be able to accurately reproduce anything smaller.

http://archimago.blogspot.be/2017/02/musings-discussion-on-mqa-filter-and.html?showComment=1487090152241#c649077476121042324

 

Quote

Remember, at 16/44, for frequencies below Nyquist, time domain performance in on the order of 60 picoseconds which is what Monty's video is showing for that ability to place the signal between sampling periods. (See the old post: http://archimago.blogspot.ca/2015/10/musings-meditations-on-limitations-of.html). This time resolution is already beyond the 5-10µs threshold the Kuncher papers speak of.

 

[FredericV]

7 hours ago, Shadders said:

Hi,

Didn't the user "witchdoctor" use the term MQA haters ?. GUTB uses this too.

Or is this the usual term as per the facebook group - to call people exposing the truth, as a hater of MQA ?

Regards,

Shadders.

Peter Veth was using the word hater, even attacking me via personal messages using the "hater" word all the time.
I also encountered a user Bob Okeson on facebook which I blocked, as he was also using the "hater" word.

These MQA opinion makers which are everywhere on the internet like it's a full time job, always use the same arguments.
It's like arguing with flat earthers.

I have a love/hate relationship with MQA:

- thanks to MQA, I did a lot of research into PCM and learned a lot = LOVE
- thanks to MQA, I also invested a lot of time in researching resamplers = LOVE
- thanks to MQA, we have the Ayre filter implemented via SOX = LOVE

This was possible thanks to Archimago, Mansr, Xivero, Dr Lesurf' and several others. The topics on CA were also very enlightening.

-> so we benefit

but what I truly despise about MQA:

- selling a modified version of the master as Master Quality
- claiming this is what the mastering engineer heard -> it's not (see Brian Lucey's masterings and the MQA encodes without his permission /  approval)
- secretly pushing the minimum phase filter which degrades PCM when implemented in a "cheap" way on hardware with very limited cpu power (like addressed by Auralic's CEO and many others)
- mk2 versions of existing DACs with sound like SH* when MQA started to mess with it's design, and some MQA module is in the chain in the mk2 and not in the mk1 version
- fake authentication of a file, where the mastering engineer did not agree to his work being altered
- claiming PCM is wrong, and MQA solves everything
- the DRM aspects of MQA including phone home and deliberate degrading features
- the misleading marketing
- not fully compatible with open source, but upsampling with SOX is an alternative without even using the first unfold

-> this is what Auralic is also doing: they upsample MQA with minimum phase and they also figured out how to modify SOX
-> we came to the same result, without looking at the exact code Auralic is using, and the sox recipes they have configured or modified

I'm very happy that I connected the dots and have a working SOX version of  Ayre's / MQA's upsampler, where I could duplicate some of MQA's claims, but with the knowledge that redbook PCM already has picosecond timing resolution, which is ignored by MQA's marketing.

I would never enforce this filter, but leave it open as a choice. For some music it may improve things. But it's not something you can enforce on everything. It will actually degrade certain music, which is why the blind test of archimago came to a mixed conclusion.

Without MQA, there would never have been such a deep discussion about PCM and why it's still something good In the end, MQA's time domain hocus pocus is implemented ..... via PCM.

[FredericV]

11 minutes ago, psjug said:

Back to "MQA Tested, Part 1"...

Here is the response of the linear phase DAC, when input is MQA encoded impulse:

Fig.4 Benchmark DAC3 HGC, impulse response (one sample at 0dBFS, MQA-encoded, 48kHz sampling, 100µs/horizontal div.).

 

1. Jim Austin writes "This response is mostly linear-phase, though the asymmetry suggests some nonlinearity in the phase response."  WTF is "mostly linear-phase"?  Wouldn't it be more honest to say the response has lost the linear phase symmetry while at the same time retaining the pre-ripple?  In other words it looks kind of f'd up?

2.  Why didn't he show the response of the minimum phase and slow roll-off filters to the MQA encoded impulse?  I don't see how it could be an oversight that these were not included.

3. Can the people on here who are much smarter than me figure out how what modification MQA is making to the impulse such that it that would result in the response shown in Figure 4?
 

Is this deliberate degrading / making the transient more complex, and being undone back to normal by MQA's renderer filters?



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

[FredericV]

So why does the article about Mytek's impulse response is being compared to upsampling? Same applies for all those canned articles which explain MQA's time domain hocus pocus, like the sound on sound article which is used by all the MQA key opinion makers.

Shouldn't this ring a bell to those readers not following up on CA and Archimago, that MQA is based on upsampling?

[FredericV]

I noticed Brian Lucey's posts are being deleted in the comments section of the Stereophile article.
He also was not mentioned in the article.

Brian exposed the fake authentication of MQA, as MQA encodes existing masters not approved by the mastering engineer, and the MQA user will think it's authenticated when the blue light shines. He also exposed the audible degradation, making MQA not master, not quality, not authenticated.

This indicates:

_ stereophile articles are a paid series by MQA
_ otherwise they would not censor the truth

MQA also hired Hans Beekhuyzen for very similar reasons. Hans even pasted emails from Bob in his video's, as if this was proof MQA does not contain any DRM. Guess he did not read the patent which includes the deliberate degrading which MQA offers as a feature.

[FredericV]

10 minutes ago, Ryan Berry said:

 

Not trying to be too particular, but SOX's filter is a filter somewhat representative of Ayre's, though it's not Ayre's filter exactly.  I just wanted to point that out before everyone starts calling the SOX filter the "Ayre" filter as there's other nuances of the filter that makes it what it is that isn't seen in an impulse response, but make surprising differences to the sound.  Perhaps just calling it a minimum phase filter would be more appropriate.

 

Thanks for the clarification. The "standard" minimum phase setting in sox gives a long postringing tail (upper track), but when using specific parameters for passband_end and stopband_start, it can result in a very similar impulse response (lower track) as MQA and Ayre's filter:

Not saying it is in any way the exact filter Ayre or MQA is using. I found this filter on a forum and decided to test & plot it.

[FredericV]

@GUTB What is there to hide?



Speakers we test with:
- Amphion Krypton 3 flagships for farfield
- Amphion Two 18 studio flagship for nearfield
- Amphion One 18 for nearfield
- John Watkinson Legends, active omnidirectional speaker
- Aktimate active speakers for testing if a server is correctly playing after staging, combined with ifi

DAC's:
- Mytek Brooklyn
- Metrum Adagio

- lot's of the little iFI dacs for testing machines after staging, and for zone2 tests

and access to Aries Cerat, Mytek Manhattan mk1 and mk2, Zodiac Platinum with 10M superclock, MSB analog dac and many others in any price range

AMPs
- vitus RI-100 integrated
- vitus SS-025 power amp
- and a diy LM3886 amp for the One18 which I also use for my desktop system sometimes
- access to a lot of others including Metrum Forte

 

Source is always some variation of the music server we produce, see signature. While we are also in the DSP business, our source can also be used as a high-end bitperfect source. Used at Munich High-End 2017 in the Aries Cerat system, our first Munich show as exhibitor.


 

 

[FredericV]

13 minutes ago, plissken said:

 

When they wouldn't take 24/192 from Mark Waldrep (even though Bob Stuart said they would) and then going non-responsive to him...

 

A theory: being a lossy codec, MQA throws away so much entropy that with Mark's high quality masterings, it would sound less than the real master, so it would not serve a demo track for MQA. Just like a Bose demo, MQA demo tracks are most likely cherry picked.

[FredericV]

11 hours ago, firedog said:

https://www.stereophile.com/content/more-mqa

 

https://www.stereophile.com/content/mqa-tested-part-2-fold

 

Interesting that they seem to think the first MQA article showed that the time response of MQA is doing what it claims-I’m not sure it did.

 

At least JA does understand that there are DRM and consumer choice issues involved.

 

 

What a load of crap:

 

Quote

But as with every aspect of the codec, there's a serious philosophical perspective behind MQA's commitment to data-rate reduction. For Stuart, efficiency in the delivery of musical information is an aesthetic, even an ethical commitment.

 

Let's deblur this to:

 

Quote

The MQA encoder, he told me, can detect "spectral components above 48kHz," and has "several strategies" for dealing with it, including choosing from among more than 2000 encapsulation algorithms. The encoder will choose the option that will "allow the decoder to most accurately reproduce the signal 'envelope' and slew-rate." Does that mean that this musically relevant information is preserved—and how well? As I said, I'm not sure—Stuart never quite committed on this point—but it seems a lot of trouble to go to just to throw it out.

From the measurements we have seen so far, MQA throws away everything above 48 Khz. It replaces everything above 48 Khz with aliasing. Our reviewer has no means to validate Bob's claim. The reviewer tries to shine some light into a black box and fails. The secrecy of the black box allows this kind of BS to be published.

 

Quote

To sum up the losslessness issue: In its folding and unfolding, MQA distinguishes between music-correlated data and noise, tries hard to retain the music-correlated data, but sensibly worries much less about preserving the noise bit by bit. This allows MQA to achieve their goal of preserving the benefits of high-resolution data without the burden of large, weighty swaths of pointless noise.

 

This can also be obtained using the Xivero solution or similar strategies to blank the noise bits therefore reducing the entropy and thus file sizes. This achives a higher compression ratio than MQA's implementation.

And there is no need for a proprietary decoder. So which solution is than more ethical?

The last part where he can't hear the noise added by MQA basically debunks the need for high-res:

 

The noise level is similar to redbook noise. This again confirms MQA is somewhere around 15 ~ 17 bits of actual resolution. But it's enough for MQA to get away with it.

Image used in this post under fair use.

[FredericV]

10 hours ago, rickca said:

So beautiful.  His motivation is pure as the driven snow. Makes my engineering soul soar.

So ethical to use crypto DRM.

https://media.ccc.de/v/34c3-9113-mqa_-_a_clever_stealth_drm-trojan

and not use "bit freezing" / entropy reduction, compatible with all exsting DAC's:
http://www.audiomisc.co.uk/MQA/cool/bitfreezing.html
https://www.xivero.com/xifeo/

 

MQA oh so fair

[FredericV]

3 hours ago, mansr said:

I'm saying that a) you don't need MQA to deliver higher quality than 24/48, and b) MQA being higher quality than 24/48 is debatable, it having somewhat better high-frequency extension but lower bit depth and added ultrasonic noise.

 

And this extra ultrasonic noise doesn't seem like a problem for Stereophile:

https://www.stereophile.com/content/mqa-tested-part-2-fold

(c) Stereophile, used under fair use

They tried hard and could not hear MQA's higher noise floor (compared to the studio master in DXD) which is hardly any better than redbook. It's 20dB higher than DXD. So if you can't hear the noise at all, why not cut it out? What's the purpose of the first unfold anyway?

When capturing actual concert halls, like I did in deSingel in Belgium with K's Choice in choir, a full 240 piece choir with band singing at full volume vs the three band leaders singing a song a capella unamplified (which was quiet), all these variations fit nicely in 16 bit dynamic range as captured by my camera. We recorded the hall a second time with a professional 4 channel surround recorder which has much better microphones that my Nikon D750, studied the dynamic range, and again, 16 bits was more then enough. This was from the 7th row in a 940 seat hall.
 

The RT60 of this concert hall is about 2 seconds, which is the time going from a crescendo back to background noise. In some MQA talks it is said that you need an extra 10dB of headroom for details buried in the background noise, so 70dB range should do the trick for this hall, which is about 12 bits.

So if MQA can do somewhere around 15 to 17 bits, they have some extra headroom and more than what is needed for most situations, but they basically debunk the need for 24 bit highres, as they use 24 bit flac for other reasons, mainly to bury their secret DRM lossy non nyquist data in the noise floor.

So what is the real purpose of the first unfold, if a higher noise floor in the ultrasonics is not being detected by the reviewers anyway except in a plot, but not by their hearing ?

[FredericV]

8 minutes ago, Evangelist said:

I was Not even invented to the MQA press conference and I was never payed any amount or granted any privilege by MQA, Bob Stuart, Meridian or any other company or person for my opinion on MQA. I have never been payed for influence on ANY journalistic work I have ever done and I will never have money influence my opinions.

 

So please elaborate why you registered as Evangelist.

 

8 minutes ago, Evangelist said:

So, FredericV, do your home work and while you're at it, read the MQA publications again (I have read about 125 pages of it) and you'll see you're wrong there too. Hans Beekhuyzen

 

We did. Did you study all reverse engineering work on MQA? Probably not.

[FredericV]

4 minutes ago, Evangelist said:

that's the wrong way around. Please proof that I was payed by MQA or associates. You're the one that made the acquisition.

I get your commercial emails advertising your paid services. You even were outside our XFI room where I overheard what you are being paid for your video articles and such.

You should haven been smarter, instead of registering as Evangelist.

[FredericV]

3 minutes ago, Evangelist said:

Yes, and?? Did you see any MQA commercial on my channel? Further more adds are adds and will never influence my journalistic work. I think you're dutch, so ask around. Some distributors hate me for it.

MQA has stated in their year report they have key opinion makers.

You have shown to be very uncritical towards MQA, just copy pasting the marketing of MQA into video series, like other magazines who put it in magazines or the web. You also made clear mistakes not understanding the dragonfly, but at least you admitted your fault later.

You also have a lot of video's where you try to debunk the MQA criticasters. You even have video's where you hammer on the need for 24/192 for MQA's time domain argument (which was already debunked by Archimago and other researchers).

Did you even study the reverse engineering work by Mans Rullgard, Archimago, Xivero, Soxr and the Chaos Computer Club?
All of MQA's filters have been dumped, we know exactly what is happening in devices like the dragonfly.


 

[FredericV]

1 minute ago, Evangelist said:

show me then when and where I am payed by whom

 

So you make all these video's trying to debunk arguments against MQA (like the ones in this thread) just for free? As a hobby?

Mot evangelists are being paid for their work.
 

[FredericV]

10 minutes ago, Evangelist said:

Don't come back with rhetorical questions but show me your proof. You can't for there is no.

 

 

So you work for free at hifi trade shows doing presentations about MQA?

 

I still have your reply from the XFI show. It's not the only event where you gave MQA presentations.
You also did an MQA presentation at Chattelin Audio Systems, and this was in colaboration with the official Meridian distributor.

So again, convince me you are NOT affiliated with MQA / Meridian. You can't. You can't hide from google. Some way or another, you are being paid to be an MQA evangelist and/or to do these talks, video's .... you don't work for free. If it's direct on indirect does not matter.
 

[FredericV]

There are more lectures by Hans Beekhuyzen about MQA, for several distributors and resellers:
 

MQA lecture by Hans Beekhuyzen, by the NL branch of AES:

https://www.facebook.com/events/1305380532856375/
http://sound-engineering.nl/nl/lezing-aes-mqa/
http://www.aes-section.nl/lezingen/project 247/project247.html

Tijdens de lezing gaat audiojournalist Hans Beekhuyzen in op vragen als:

    hebben we meer bandbreedte nodig dan 20 kHz en zo niet,
    waarom dan hogere bemonsteringsfrequentie?
    Hoe word de hogere informatiedichtheid in een MQA bestand bereikt en
    wat zijn de vooruitzichten in de toepasbaarheid?

-> do we need more bandwidth than 20 kHz, and if not:
-> do we need a higher sample rate? (only for the fake time domain claim in the renderer / upsampler)
-> how is higher information density reached in an MQA file?
-> what are the prospects and applications?

XFI lecture by Hans:

https://www.audiovideo2day.com/x-fi-2016-lezing-mqa/
Wordt MQA het nieuwe audioformaat?
"Is MQA the new world format"?
-> includes the typical MQA lies " bieden MQA-bestanden gegarandeerd de kwaliteit uit de studio "
 

"MQA files are offering the quality of the studio guaranteed" -> sorry Hans but this has already been debunked

Lecture for Dali / Bluesound by Hans:

https://www.audioclub-limburg.nl/dali-benelux-bv-presenteert

 

He is advertising his services to us via email. You can't convince me he is NOT being paid to give these MQA lectures. Hans is not doing a very critical review of MQA, but quite the opposite. Furthermore Hans was outside our room at the XFI show where I overheard him discussing the remunerations of the hifi journalism business.


 

[FredericV]

5 hours ago, crenca said:

 

Chris,

 

It's not Hans affiliated with the industry? His YouTube channel is a for-profit business, just like a Blog is.

And several video's are Patreon only. So for profit. In the past he wrote for Vi-Fi, which I respected as he was one of the few to show measurements of gear, and he was a pioneer in streaming reviews. Stuff like the early squeezeboxes, ... This was 10y ago. His magazine went bankrupt, official reason because the bank pulled the plug.

Years later he rebooted his activities via his video channel and the HB project, which are for profit business. His company is Beekhuyzen Beheer BV

https://drimble.nl/bedrijf/nieuwegein/5852404/beekhuyzen-beheer-bv.html

 

I receive his commercial emails signed Beekhuyzen Beheer BV:

The Hans Beekhuyzen Channel en TheHBproject.com are proporties of  Beekhuyzen Beheer BV

I expect no longer to get receive them in the near future.

I stopped respecting HB when he started doing MQA lectures, and responded very uncritical to my technical remarks, which I hoped he would integrate them in his presentations.

[FredericV]

5 minutes ago, crenca said:

 

Thus, his is an audio related business not different than ML's, except that it might very well be $successful$.  Industry affiliated indeed...

It's a business:



This was sent to a lot of email addresses, so no privacy is being violated as this is not a personal message.
His rates are not absurd or overpriced, but it shows it's an actual business.