MQA at CES

[testikoff]

In addition to Miska's spectrograms here are the waveforms of MQA-encoded 24/44 sample from 2L & my home-made downrezzed/downsampled to 16/44 version of the 24/352 DXD master (which utilizes SoX95 linear-phase re-sampling & moderate noise-shaped MDA VST dithering):

 

- MQA:

 

- 16/44:

[jeffmudrick]

FWIW MQA material is live in the Onkyo Music Store :

 

https://www.onkyomusic.com/GB/releases/5044517

[R1200CL]

The interesting here is that MQA is released also in 16 bit.

 

So then we at least have a confirmation that MQA is delivered in 4 resolutions. (Or 6 if we add 32 bit)

 

That also tells me that most likely the record company can choose how to implement MQA.

 

A possible problem with Onkyo music, is that they do not tell us the original resolution of at least those 16/44.1 tracks. Or it could mean that is the original resolution.

[TubeLover]

Honestly, virtually all of this extended discussion is pointless. Since it clearly appears that MQA will require new DAC's that support it, this format is dead in the water. There was a "chance" when it appeared that it could be handled with software but that chance is null and void. As someone mentioned earlier, it would take a miracle for MQA to succeed and there aren't many of those in existence. And, even in the beginning, far too many people were far less impressed with it anyway.

 

Add in the way they appear to have misled or, at the very least, misinformed partners like Auralic, and others, there goes any credibility, in addition.

 

JC

[francisleung]

When I set out for the listening comparisons, there was no offer of MQA tracks in the test bench. I went into the two albums and paid for downloading the 049 and 106 MQA tracks.

Now there are 11 MQA tracks in the test bench. After getting them I compared 038 (Mozart Violin Concerto) with the DXD I bought previously.

This time, the MQA pushes the soundstage slightly backwards and expands it discernibly to two sides. So the soundstage becomes too wide for an ensemble of 22. The DXD has it better, with proper layering front to back and the violin forwardly separated from the ensemble. Well I know this may not be the actual setting during the recording but conforms more to the usual perspective in a live concert. Also the "cathedral" acoustics I mentioned in an earlier post regarding the 106 track, is not provided by the MQA this time; neither the DXD has it.

The MQA processed 038 track is worst than the original DXD in my view. The score now is 2:1 in MQA favour.

[Doak]

Honestly, virtually all of this extended discussion is pointless. Since it clearly appears that MQA will require new DAC's that support it, this format is dead in the water. There was a "chance" when it appeared that it could be handled with software but that chance is null and void. As someone mentioned earlier, it would take a miracle for MQA to succeed and there aren't many of those in existence. And, even in the beginning, far too many people were far less impressed with it anyway.

 

Add in the way they appear to have misled or, at the very least, misinformed partners like Auralic, and others, there goes any credibility, in addition.

 

JC

 

Totally agree with this assessment.

[francisleung]

Honestly, virtually all of this extended discussion is pointless. Since it clearly appears that MQA will require new DAC's that support it, this format is dead in the water. There was a "chance" when it appeared that it could be handled with software but that chance is null and void. As someone mentioned earlier, it would take a miracle for MQA to succeed and there aren't many of those in existence. And, even in the beginning, far too many people were far less impressed with it anyway.

 

Not quite.

It is said that MQA encoded tracks would still bring about sonic improvements even when played via traditional DACs. And MQA dedicated DACs will give even better results.

It was based of this that I set out to make comparisons. Otherwise if a dedicated DAC is a prerequisite, I would have dismissed it altogether.

[francisleung]

Another comparison: 2L 050 simple symphony

This was recorded with the ensemble sitting at a round table of sorts with microphones placed in the centre of the table. Some of the players were seated with their backs facing the audience.

When playing just now the DXD track from the album that I bought previously, I was able to figure out the back facing players somewhere in front of the plane of my speakers during loud passages.

The MQA blew these phantom images away.

Removed the MQA and the score is 2:2.

Up to now it becomes dicy as to whether MQA is to go for or not.

To popularize MQA, I think instead of charging a royalty, Bob Stuart should urge Morten Lindberg to give away MQA as a bonus when customers buy DXD or DSD albums.

With MQA being charged on its own, I would take DXD instead.

[R1200CL]

The MQA Flying Circus

Is it only me ? Please correct me if I'm the insane one.

 

First it the was that poor HW manufacturer. Now we pass it over to the webshops and record companies.

 

Onkyo is first out.

The 2L MQA stereo files are sold as FLAC 24/352.8. Other MQA are sold as 16/44.

 

Costumers may start to wonder what they are buying. If they buy at all ? Ever ?

 

The 2L store was very clear about the different formats sold, except about resolution of the MQA file. And it is a very nice gesture to allow people to test the formats.

 

Bob now seems to copy Apple's communications strategy, and only gives interviews to people who do not ask critical questions. Bob says MQA is a philosophy, it's many thing. You can't purchase that. (Sell maybe to a CEO?)

 

Before we used to understand the meaning of MP3, CD format, and 24/96. You knew what you purchased. Now we have a something with hidden requirements in both ends, in an end to end game changing technology. Are there a Guinness in confusion ? You are hereby nominated.

 

Now they (Onkyo/7Digital) do not tell you that the 352.8 actually is a 44.1 compressed file.

Or that the 16/44.1 is a Redbok encoded CD. (I guess, cause are the master really in that format). Remember 2L sell Redbok as 24/44.1

 

They do not tell you that the transport in this philosophy can take 6 forms, and that even if your MQA DAC is only 192, it can eat 384, cause that is our (MQA) faith.

 

Original Redbok can be both 16 and 24. Proven so by the two webshops.

What are the requirements for a record company to be allowed to use the MQA logo?

Is it a commercial secret ?

 

We for sure discovered one requirement on FB! when it came to transporting and decoding the format.

 

Will an US solicitor claim that the webshop are selling something not according to the description. Hey, these guys has money. Lets sue them.

 

For us (me) that has tried to understand the format for some days, it maybe clear, but does the average person understand that purchase a MQA FLAC 24/352.8 or a 24/88.8, will in both cases give him a 24/44.1 unless he has a MQA DAC ?

 

Roon may has a solution to some of this as they would like to implemt a DAC multiprofile SW. (Not a default one). Possible based on auto negotiation and/or user select.

 

Will they be allowed ?

[testererr]

Hi, Miska.

 

In fact they only "keep" the 13 bits of the original sound.

They compress (perhaps lossily - they can claim to be "lossless" based on the fact that it's a new recording) the rest of the high-freq bands and put it into separate data streams.

Bob with his partner created a few years earlier and presented at AES http://www.aes.org/e-lib/browse.cfm?elib=7964 a technique to hide data (watermarks/DRM/etc) inside audio, using the highly-compressed data as a dither pseudo-random source and some sideband data to allow extraction of the dither-prng stream.

So they use 3 bits on the original PCM stream plus the 8 extra bits, plus whatever they can get from the dithering technique above, giving them ~10-16bits per sample = 400-800Kbit/sec per channel, 0.8-1.6Mbit/sec stereo data stream to put all the compressed data into.

 

Actually, dropping the stupid goal of playing a compressed stream on unsupported devices, your idea of compressing higher-frequency bands with vorbis is basically what they're doing. But to keep the "layered" approach to quality, they have the extra layers hierarchically divided: the "legacy" 44/16 stream, then a layer of extra 22KHz up to 96/24, then a layer of extra 44KHz up to 192/24, then a layer of extra 88KHz up to 384/24, up to the "top" quality being recovered using a special "touchup" layer, being the compressed difference between the output of all layers decoded and the original file.

[granosalis]

>> In addition to Miska's spectrograms here are the waveforms of MQA-encoded 24/44 sample from 2L & my home-made downrezzed/downsampled

>> to 16/44 version of the 24/352 DXD master (which utilizes SoX95 linear-phase re-sampling & moderate noise-shaped MDA VST dithering):

 

 

So, 2L is worst than you and Miska to encode their own masters?

No, excuse my frankness but for me it is unthinkable.

[kravi4ka]

Which of the current DAC ICs support MQA? And what does the DAC chip need to be compatible with the MQA format, I did try to understand but I do not see a HW requirement mentined anywhere, just "MQA is supported by X, Y, Z companies..."

[testererr]

Reading up some more on the tech, it seems there is one more trick, which might explain the differences in quality people perceive.

Bob is talking about doing sampling not using square "windows" but B-splines of order 2 (triangular over 3 samples) and higher. There is convolution with triangle or higher-order function during sampling, and interpolation during playback.

So even at 44KHz, each sample will be based on several (4+) real samples, with different weights (by default, pyramid-shape triangular). This reminds of DSD somewhat, though Miska might be able to clarify the similarities and differences better.

[francisleung]

MQA lost out to DXD again.

2L-064 Come away death, mezzo soprano singing with accompaniment by piano.

The DXD track was one of my favourites some time ago.

Now a chance to compare with its MQA processed offspring.

The MQA did not screw up anything, for example, the size of the mouth and smoothness of the voice with no trace of compression or distortion, but did not improve anything either.

The DXD has the voice and piano in more 3D, hence better perceived separation, and a deeper front to back soundstage.

So the MQA goes to archive and the score is now 3:2 in DXD's favour.

[testererr]

Thinking about it some more, this "new" sampling method in principle is the same as in DSD (valuing timing information higher than bits), only more "convoluted".

[Miska]

For us (me) that has tried to understand the format for some days, it maybe clear, but does the average person understand that purchase a MQA FLAC 24/352.8 or a 24/88.8, will in both cases give him a 24/44.1 unless he has a MQA DAC ?

 

It looks like 24/44.1 because that's the amount of bits it has, but it doesn't really have that much useful data because the lowest bits are just encoded MQA data that appears as noise and only useful for the MQA decoder.

 

So compared to a 16/44.1 file, you get more than twice as big 24/44.1 file spiced up with extra noise. It is more than twice larger because the FLAC encoder cannot efficiently encode noise due to it's randomness. (just like if you ZIP a file with random data it doesn't compress practically at all)

[Miska]

Which of the current DAC ICs support MQA? And what does the DAC chip need to be compatible with the MQA format, I did try to understand but I do not see a HW requirement mentined anywhere, just "MQA is supported by X, Y, Z companies..."

 

None, they offer a firmware module for the XMOS chip commonly used to implement USB interfaces for DACs. So typically MQA decoding wouldn't work for example if you send the data to a DAC through S/PDIF.

[Miska]

Bob is talking about doing sampling not using square "windows" but B-splines of order 2 (triangular over 3 samples) and higher. There is convolution with triangle or higher-order function during sampling, and interpolation during playback.

So even at 44KHz, each sample will be based on several (4+) real samples, with different weights (by default, pyramid-shape triangular). This reminds of DSD somewhat, though Miska might be able to clarify the similarities and differences better.

 

Typically ADC/DAC chip sampling uses a FIR filter with number of taps in couple of hundred. IOW, there is convolution with few hundred weights. So every sample is based on few hundred surrounding samples. This allows to realize steep "brickwall" filters, but side effect is "ringing" in time domain. This is the Nyquist-Shannon-Fourier law that frequency resolution is inversely proportional to time resolution.

 

B-spline and similar are very fast and light weight interpolation compared to FIR, because they require only few operations. But the down side is that filtering performance is really poor, meaning lot of frequency leakage (images/alias). It also usually leads to early HF roll-off because the roll-off curve is so gentle.

 

HQPlayer has two different spline-type interpolators, "polynomial-1" and "polynomial-2".

[Nikon Nutter]

if manufacturers pay meridian. then we the buyer pay a bigger price. Everyone pays meridian. this is smart business. It is I think.

[testererr]

These slides show the theory quite clearly, if you read only one of these, read this one:

http://icms.org.uk/downloads/BtG/Dragotti.pdf

 

Bob references these papers in explaining the novel sampling process:

 

Main:

http://bigwww.epfl.ch/publications/dragotti0701.pdf

 

Secondary:

http://research-srv.microsoft.com/pubs/69188/sampling.pdf

CAUSAL RECONSTRUCTION KERNELS FOR CONSISTENT SIGNAL RECOVERY | Fanny Yang - Academia.edu

http://arxiv.org/pdf/0812.3066v1

http://webee.technion.ac.il/people/YoninaEldar/files/mainFormat.pdf

[matthias]

Do you think this papers are suitable for a non mathematical forum?

 

Matt

[james45974]

Can someone clear up a little confusion (?) on my part about MQA. Is there just a one size fits all file or are there different MQA files for each different resolution. For example, Tea for the Tillerman can be had at 64 & 128 DSD, & 192/24, does that mean there will be 3 different MQA files?

[Jimmypowder]

So I'm I right in reading that Merdian has changed their mind about allowing streamers without a dac to use MQA ?

 

If this is yes , then do they want dacs specially equipped with MQA decoding capability ( theirs mostly? ) to be the only ones to get the MQA supposed benefit ?

[jzahr]

All of this MQA stuff is surely pretty confusing.

 

After much reading in threads like this and others, I will expose my opinion on this, for what its worth. Given the lack of comprehensive technical info from the proponents of MQA, it is possible that my assestment contains a fair amount of speculation: I'll try to keep this to a minimum, but keep in mind that all of us, to a certain extent, are trying to infere things from the scarce info given.

 

In short, I find this MQA thing not convincing at all. Also, I believe it will fail in the long term if it continues to be presented to the industry in the way it has until now. It will fail not because it is a "bad" thing per se, or badly enginered, or technically flawed per se, but because MQA is a too big "technology", encompassing too many different aspects (three at least I can think of) all of them presented in a single pagake and directed, apparently, to a global non-differentiated target (a target which includes not only the end music lover but also the music distributors and the music producers).

 

I'll try to explain my rationale and, for this, I will first explain how I believe this MQA thing works:

 

From what I can deduce, MQA comprises three, more or less, separated elements (correct me if I'm wrong or if someone find better info available somwhere), which I'll call PHASE 1, PHASE 2 and PHASE 3.

 

PHASE 1: This is the part where the proponents of MQA claim that they have a process or algorithm or dsp thing that can correct for defficiencies or time errors (time smear, in their words) that were comitted at the time of the original recording and AD conversion. The persons who could use this part of MQA are the music producers (probably, the mastering engineers in charge of the final mastering for new music releases, or the engineers in charge of remastering tasks for re-releases of old material). The claim is that this algorithm improves the sound quality, but it must be noted that (1) there are many tools offered to audio engineers by different software companies that also claim improve the quality of masters or remasters, and (2), that the music lover at the end of the chain do not need to know what production tools were used in the production of his favourite music.

 

PHASE 2: This is the part that takes place after the music production stage, and involves a form of "preparation and delivery" of the musical product. This preparation consists in a form of "compression" and encoding which (it is claimed) can pack a Hi-Res pcm format in a 24/44 wrapper (this, of course, if I'm understanding correctly what they claim). This second part of the MQA thing should be applied by the "music distributors" (being them providers of downloadable files or streaming companies). The end result of this preparation is a digital stream (or file) that needs to be decoded during playback of the contet in order to achieve the "full benefit" of MQA, which leads us to PHASE 3 of the MQA thing.

 

PHASE 3: the end user should have a dac equipped with a MQA decoder to recover the original Hi-Res pcm format (if the original was, indeed, Hi-Res pcm) which includes the supposedly excellent improvements added during PHASE 1. However, it must be noted that it is not really mandatory to have a dac equipped with this decoder, as it is claimed that the MQA encoding is backwards compatible with all available dacs in the sense that if a decoder is not present during playback, then the DAC used by the end customer is presented with plain old redbook pcm (at 16/44.1 or 24/44.1, not sure at this point, but inclined to believe that is 16/44.1); also it is claimed (again, my interpretation) that this non-decoded digital stream still contains "some" of the benefits added during reparative filtering of PHASE 1.

 

One thing I don't know at this point: are the PHASE 1 and PHASE 2 aspects of MQA indeppendent of each other or are they indissolubly tied together? Can the repair filtering of PHASE 1 be applied in a stanalone fashion? Can the encoding scheme of PHASE 2 be used without the reparative filtering?

 

Now to my rationale:

I think that this way of trying to encompass the whole end-to-end chain is wrong and has many drawbacks. For one: why not present three different products to three different targets? One product for "music producers", other for the "distribution chain", and other for the "DAC manufacturers". This, of course, would imply compete in three different markets. Also, avoiding the marketing wording of the like of "hey, we have the best and the only filter in the world which can repair the great damage made by ADCs, and have the only encoding in the world that can transport and deliver this to the end user so, if you, the end user, do not buy a decoder-equipped DAC you'll loose the incontrovertible benefits and if you, seller of musical content, don't provide MQA, will loose music buyers"

 

For instance, ragarding PHASE 1, there already exist an entire industry of software makers who design and sell DSP algorithms to the music producers and audio or mastering engineers, in the form of plugins to be used inside Digital Audio Workstations (DAWs like protools or others) and Audio Editors (like soundforge, wavelab or others). Among this plugin market you can find equalizers, dynamic processors and digital filters (most probably including apodising ones). The thing is: why not release a separate software product to this already existing market, containing the digital filtering of MQA which supposedly corrects the errors and time smear introduced by the ADCs used in production? The normal thing to do regarding PHASE 1 would have been just this IMO, and let that market decide: I can't assess the technical merits of the MQA correction algorithm, perhaps it is very good or even truly excellent, but in my opinion the music producers or artistic responsible persons in charge of a music production should be able to freely decide on the most suitable production tools and, in this context, some producers may elect the MQA filters while others may prefer offerings from competitors (izotope or others who offer "repair tools" or "mastering tools"). Also, the end music lover need not to be concerned (unless he really wants to) with what production tools were used and can concentrate just on the pleasure of critical music listening.

 

With the current marketing strategy (which, as said, tries to encompass the complete end-to-end chain) and loose technical wording by the MQA proponents, and above all, with the choosing of presenting MQA in "consumer oriented" scenarios like CES and others (thus involving the end user in a product that should be aimed to professionals), the only thing that will happen is a thing that is already hapening: confusion, confusion and more confusion on the part of the end users, of whom 99% can't understand the MQA technolgy or process. This music lover now will be left with a sense of not knowing or understanding what he is buying, which will cause him anxiety. This end customer, feeling that otherwise he will not be receiving the hole product, will start asking the labels to include the MQA processing of PHASE 1 in the production stages ("hey I'm not buyin anything that is not MQA approved because, even if I don't understand it completely, it is supposed to be good and better"); with this, the afore mentioned freedom of choice of production tools suffers for music producers, leading to loss of musicality in the released musical product.

 

Even more confusion will arise from the difficulty in choosing what to buy as many different versions of a music track will be offered in different sites, or even by the same seller, creating doubt in the music buyer, as suggested by R1200CL in post #359

 

Do you remember when microsoft tried to encompass the complete "or lage parts" of the software bussines, not only providing operating systems but also offimatic tools, design tools, internet browsing tools, and so on. At some point the market regulators step in and the company was forced, for example, to offer choice of other internet browsers as alternative to IE. In some countries, they were forced even to offer versions of Windows without the mediaplayer preinstalled as default, in an effort to give the user some freedom of choice.

Even if at a smaller scale, I see this MQA stuff very similar, except that here no market regulator would ever be jumping in (because they won't be able to understand the technology to be regulated).

 

Additional things:

• Lets suppose that a person buys an excellent DAC, that happens to lack the MQA decoder. Lets suppose that it also happen that such dac has in-built an apodizing filter. Then, if a person buy and play a MQAed audio stream, dual apodizing filtering may be in effect. Could this have detrimental effects? Who can judge this technically?
  
• Regarding the decoder at the DAC, it is needed only for the decompression in PHASE 3? Or, on the other hand it has some relation with the corrective filtering of PHASE1? SO, what is truly the decoder needed for? To improve on the filtering at the dac stage or only to "decompress" the encoded MQA stream?
  
• Regarding correcting the time-smear damage impaired by the original ADCs, how can this be done in the case of multi-track recordings? There are studio recordings which there may be 48 or more tracks, probably not all of them recorded at the same time, not even with the same ADC equipment. Some studios have very complex productions methods, and claim that all this can be corrected by a certain filtering at mastering stage sounds a little too pretentious, to me at least.
  
• If the proponents of MQA found a good way of correct the damage impaired by ADCs, why not design and expose to the pro market their own ADC which includes this filtering. This would be a killer product in that market if it can demonstrate superiority over well-known converters like Meitner, Grimm, Digital Audio Denmark, Apogee and others who impair a "clear" damage to their recordings. Meridian certainly has the knowledge to make a great ADC. Of course, this would leave out the possibility of controlling what happens on the DAC end.
  

 

As a summary: a technology which is very complex presented to a too broad and undifferentiated target, using loose descriptive language and with the ambition of becoming a tie for the entire music industry with legs in its production, distribution and hardware aspects, which may result in confusion and axiety.

 

Sorry for the long and tedious post. Cheers

 

Jorge

[tallica1]

MQA vs DSD 64/128/256 there is no competition..

 

DSD wins