[spruce music]

   
How would reconstruction algorithms help them deal with increased noise in the stopband that result from using a minimum phase filter?

I am probably not the person to give a definitive answer about that.  Read some of the papers on compressive sensing.  Even then I may have it wrong.  My idea is they will be able to unfold this into rather high sample rates and use minimum phase filtering that has a very wide transition band so they don't get these noise artefacts below 20 khz. Then again it would nice if they weren't so secretive and just spilled the beans about what they are really doing.

 

[watchnerd]

 
I am probably not the person to give a definitive answer about that.  Read some of the papers on compressive sensing.  Even then I may have it wrong.  My idea is they will be able to unfold this into rather high sample rates and use minimum phase filtering that has a very wide transition band so they don't get these noise artefacts below 20 khz. Then again it would nice if they weren't so secretive and just spilled the beans about what they are really doing.

 
I'm actually very familiar with compressive sensing and have used it in sparse coding for RF data.
 
You might be able to sparse sample the top octave (10khz - 20khz), which would potentially reduce stopband noise inversely proportional to the sample sparsity (i.e. few samples = high reduction in noise), but one obvious problem is that the top octave may not be sufficiently incoherent to allow compressive sensing to work in a way that sounds authentic as opposed to artificial.
 
Then again, given how hearing deteriorates with age, and the rolled-off euphonia that analog tape and tubes exhibit and that some people seem to like, it might be psychoacoustically fine.
 
But one thing it wouldn't be for sure is Redbook....it would be something new.

 

[spruce music]

   
I'm actually very familiar with compressive sensing and have used it in sparse coding for RF data.
 
You might be able to sparse sample the top octave (10khz - 20khz), which would potentially reduce stopband noise inversely proportional to the sample sparsity (i.e. few samples = high reduction in noise), but one obvious problem is that the top octave may not be sufficiently incoherent to allow compressive sensing to work in a way that sounds authentic as opposed to artificial.
 
Then again, given how hearing deteriorates with age, and the rolled-off euphonia that analog tape and tubes exhibit and that some people seem to like, it might be psychoacoustically fine.
 
But one thing it wouldn't be for sure is Redbook....it would be something new.

Okay, sounds like you are in a better position to judge how much that could contribute.  I am no expert in that area and did not want to try and pretend I was.  What you described about it sounding authentic rather than artificial is something I have been wondering about it as well. 
 
So many prefer tubes or tape or vinyl and confuse that with fidelity it is a continual issue in audio IMO.  I do understand I preferred tubes and reel tape until I got hands on an ADC which clearly showed how utterly transparent digital is compared to those mediums. I can still enjoy those things I just know better than to try and make digital sound more like analog.  Has Meridian done something of that sort in the hopes of people hearing it and saying "oooh.....ahhhh! sounds like analog which we know is right". Everyone almost will quickly declare it 'better sounding'.  I am more interested in recordings being of high fidelity.  Let me massage to my preferences on the playback end if I choose to do so. 

 

[watchnerd]

 
Okay, you sounds like you are in a better position to judge how much that could contribute.  I am no expert in that area and did not want to try and pretend I was.  What you described about it sounding authentic rather than artificial is something I have been wondering about it as well. 
 
So many prefer tubes or tape or vinyl and confuse that with fidelity it is a continual issue in audio IMO.  I do understand I preferred tubes and reel tape until I got hands on an ADC which clearly showed how utterly transparent digital is compared to those mediums. I can still enjoy those things I just know better than to try and make digital sound more like analog.  Has Meridian done something of that sort in the hopes of people hearing it and saying "oooh.....ahhhh! sounds like analog which we know is right". Everyone almost will quickly declare it 'better sounding'.  I am more interested in recordings being of high fidelity.  Let me massage to my preferences on the playback end if I choose to do so. 

 
What's even more interesting is that if they do use sparse sampling, it could be equated to being partially lossy, which would create an interesting case where 'old school' Redbook would be a more true reconstruction of the original waveform than sparse sampling approximation, even if the latter sounds more 'analog'.
 
Which would create an odd world where those who like 'analog sound' might prefer the version with 'missing bits', which is close to the accusation that anti-digital people have used for decades....
 
Zany.

 

[spruce music]

   
What's even more interesting is that if they do use sparse sampling, it could be equated to being partially lossy, which would create an interesting case where 'old school' Redbook would be a more true reconstruction of the original waveform than sparse sampling approximation, even if the latter sounds more 'analog'.
 
Which would create an odd world where those who like 'analog sound' might prefer the version with 'missing bits', which is close to the accusation that anti-digital people have used for decades....
 
Zany.

Yes, and the replies to questions about whether it is lossy or not have been dodged.  They use terms like "audibly lossless".  It is clear some versions will be lossy over 30 khz from what they have said.  It isn't clear if all versions will be except Studio MQA which supposedly isn't.  Maybe it is just a lousy rollout with lousy PR handling, but it sure seems like intentional obfuscation about what it really is other than don't worry we and all these studio guys say it is wonderful. 

 

[watchnerd]

They use terms like "audibly lossless". 

 
Oh that's just horrible.
 
I guess we should call 320k CBR MP3, LAME V0, and 256k AAC  "audibly lossless", too.

 

[watchnerd]

As additional info (graph from Archimago), this is what you get if you strip out the 16 bits of musical content and focus on the lower 8 bits where MQA does weird stuff:
 

 
 
Basically, white noise, the kind you would get with dithering.  Or, taken differently, possibly a lossy ultrasonic reconstruction scheme based on parse sampling. And because it's random it's not easily compressible.

 

[watchnerd]

  What you described about it sounding authentic rather than artificial is something I have been wondering about it as well. 
 

 
"Without any idea of what I was testing for, she clearly preferred the original "DAT" 16/44, describing the MQA version as sounding "synthetic" in ways similar to my description above with ABX testing... Well, what can I say, can't disagree with my better half's ears - younger, better looking and she plays the piano as well "
 
http://archimago.blogspot.com/2016/01/measurements-mqa-master-quality.html

 

[Gringo]

Thanks watchnerd for the helpful response, I actually learned stuff. I wish I was informed enough to provide answers to your questions but alas I am not. I am going to guess that you have seen the two Hans Beekuyzen MQA analysis videos on YouTube which are the best totally independent explanation I have come across of MQA. His comments are much more insightful than anything I could offer and he does say that how Meridian actually achieve what they claim, is a closely kept secret and only time will prove how accurate their claims are.
I have absolutely no problem with your sceptisism of MQA which is a healthy position to take and I did not wish to imply you were trying to be-little me, that comment was more a reflection on other comments from some followers of this thread on the forum which were not directed at myself but rather another where they
indulged in insults and name calling.
As to Meridians reticence to divulge technical details, I guess if I was suddenly blessed with the insight to come up with a new way of doing something which could make money I would not go and tell everybody but would try and capitalise on it. Does that make me a money grabber? If it does, then does this mean that everyone who ever patented an idea is a money grabber? Trying to hood wink the consumer, yes that is totally reprehensible and if proven should be condemed. As others have said, Meridian's PR might not be all that it should be but Bob Stewart certainly comes over as an honest and pationate guy where music reproduction is concerned. As an analogy, be careful when swimming in the sea and watch out for preditors but remember despite the occasional shark most aquatic life is a wonder to behold.

 

[watchnerd]

As to Meridians reticence to divulge technical details, I guess if I was suddenly blessed with the insight to come up with a new way of doing something which could make money I would not go and tell everybody but would try and capitalise on it. Does that make me a money grabber? If it does, then does this mean that everyone who ever patented an idea is a money grabber? 

 
You're mixing up patents and trade secrets.
 
Trade secrets aren't disclosed, like the formula for Coca Cola.
 
Patents are publicly disclosed.  In exchange for sharing the invention, the patent holder is given an exclusive right to use for a limited time.  Meridian has already filed for this, so unless they're either still pending or there is some key piece that they haven't patented, their ability to be rewarded is protected.
 
I don't object at all to inventors seeking financial reward.  What I do object to is misinformation, and some of Meridian's terminology and turns of phrase are coming awfully close to the line of obfuscation, if not crossing it. 

 

[watchnerd]

Regarding "temporal blurring" and the need for a faster impulse response:
 
Here is the fastest impulse response of a real acoustical event (not an electrical pulse signal) I've ever seen recorded.  It's an electrical spark discharge.  Note that each sample period (dot) is 10 micro-seconds, so whole first complete peak-to-trough cycle is about 70 micro-seconds. Recorded at 24/96khz, 5 cm distance.
 

 
 
That's really fast!
 
What kind of filter / impulse response combo gets close to that?
 

 
Wow, that looks awesome!  Now what kind of filter gives me that?
 

 
Hurm...well that doesn't look like much of a filter at all...what does that do to the sound?
 

 
GAH!  That's horrible, and very very audible.
 
So clearly Meridian isn't takings things to that extreme.  So what might they do instead?
 
Well, they could use a filter that looks like this...
 

 
Not bad, about -4 db down at 20khz.  Which looks like this in the audio band:
 

 
Not bad at all.  With some noise-shaped dithering you mash down those peaks at the expense of a higher noise floor, but still keep it all well below the audible range.  And you get an impulse response that looks like this:
 

 
Pretty good, practically no pre- or post-ringing.  Where can you get this awesomeness?  Well, it's one of the standard filters built into ProTools, the most popular DAW software, probably used to mix and master most of the music you listen to.
 
Which brings up the key point:
 
Whatever MQA is trying to do on the playback end, everything else in the production chain impacts the lowest common denominator.  Even if MQA has impulse response as good as the ProTools example above, it doesn't really matter if the microphone, ADC, the mixing software, the DAC, and the speakers/headphones aren't equally good.
 
Combine that with:
 
1. Reduced compressibility due to noise generation / ultrasonic reproduction in the least 8 bits.
 
2. Unusual synthetic reproduction of the top octaves
 
3. A format that requires MQA-compatible playback HW/SW to reach maximum potential
 
4. A hybrid of lossless and lossy approaches
 
5. Possibly DRM-equipped
 
6. No smaller for streaming 16bit lossless
 
....this all seems like, at best, a lot of work for little gain.  And at worst a giant vendor lock-in scheme for a technology format that is likely to never really take off on a wide scale (see HDCD, SACD, etc.).

 

[watchnerd]

As I had speculated previously, using a minimum phase / gentler slope in order to achieve better impulse response runs the danger of higher noise in the audible band.
 
Well, it would appear that Meridian has not reinvented physics or signal processing math.  MQA is in fact noisier than high resolution PCM, even played back on Meridian hardware:
 

 
That's a whopping 12 dB more noise in the top band.
 
So for those who equate 'more analog' to 'more tape hiss type noise', I guess they'll like it.

 

[Gringo]

Sorry  "watchnerd" for the mix up regarding patents and trade secrets but it still does not necessarily make someone a money grabber in my opinion
 
Quote:

 What I do object to is misinformation, and some of Meridian's terminology and turns of phrase are coming awfully close to the line of obfuscation, if not crossing it. 

I'm not for one moment proposing to defend anything Meridian have said about MQA but I have seen quite a lot of misinterpretation by others about what MQA is or is not.  I do feel that Meridian have not been pro-actively enough in making it sufficiently clear is that the "lossless" tag is only in the sense of audible lossless.
Of the claims by Meridian that come to mind I have noted the following:
 
1. Meridian have stated quite clearly that the only information which is lost is the random 0's and 1's below the audible threshold floor due to that being where the "folded data" is stored, this is yet to be independently analysed and proved correct or not
 
2. They have claimed that you don't need the MQA hardware/software to be able to download and play MQA encoded material, (you just don't get the major benefits of MQA) so existing equipment is not made redundant.  They have proved this to be absolutely the case and there are a number of early adopters on line who have confirmed this to be so.
 
3.  They have claimed that the files size for MQA files is not much more than regular CD's.  2L's free online MQA files appear to bear this out and are there for any one to see and access
 
4.  They claim that some artefacts induced in the original transference from masters can be corrected.  The jury as far as I can see is still out on that one as there is yet to be any independent technical analysis. (There are a number of early adopters saying that they have downloaded and listened to un-decoded MQA material and claim it is slightly better than regular CD)
 
5. They claim that MQA decoded will give a better audio quality than CD and be on a par with "Hi-Res files" or better.  Again independent analytical study is yet to prove this to be true or not (again early adopters are already claiming that their downloaded and decoded MQA experience is providing a significant improvement over regular CD and generally on par or slightly better than "Hi- Res"
 
6.  They claim that what the artist or engineer signs off is what you get.  This clearly depends on the integrity of whoever signs of the file but the MQA decoder does not light up if the file is not bit perfect compared to the signed off file and this is proving to be the case by the early adopters.
 
Off the top of my head I can't think of a claim made by Meridian regarding MQA that is false or intentionally meant to deceive, (I'm sure someone will put me right if I am wrong)

 

[watchnerd]

For the record,  didn't refer to them as a money grabber.
 
 
Regarding everything else:
 
Quote:

  I'm not for one moment proposing to defend anything Meridian have said about MQA but I have seen quite a lot of misinterpretation by others about what MQA is or is not.  I do feel that Meridian have not been pro-actively enough in making it sufficiently clear is that the "lossless" tag is only in the sense of audible lossless.
 

 
There is nothing to make clear - "audible lossless" is a term they invented, clearly designed to allow some level of conflation with "lossless".  There is no term of art called "audible lossless.   The correct term is "transparent".
 
"Audible lossless" is clearly a made-up term designed to disguise the fact that MQA, in certain scenarios, is at least partially lossy.
 
 

 
 
1. Meridian have stated quite clearly that the only information which is lost is the random 0's and 1's below the audible threshold floor due to that being where the "folded data" is stored, this is yet to be independently analysed and proved correct or not
 

 
AKA, they're throwing away data.  Benign or not, TBD, but they're chucking out data.  Much of which they're throwing out will be noise-shaped dither data, which is put there on purpose to reduce quantization noise.
 
 

 
3.  They have claimed that the files size for MQA files is not much more than regular CD's.  2L's free online MQA files appear to bear this out and are there for any one to see and access

 
Not much bigger = bigger.  It's not as efficient as being compressed.
 
 

 
4.  They claim that some artefacts induced in the original transference from masters can be corrected.  The jury as far as I can see is still out on that one as there is yet to be any independent technical analysis. (There are a number of early adopters saying that they have downloaded and listened to un-decoded MQA material and claim it is slightly better than regular CD)

 
Slightly better in what way? When you take out the 8 bits that MQA uses for secret sauce, they're both 16bit PCM formats.
 
 

 
4.  They claim that some artefacts induced in the original transference from masters can be corrected.  The jury as far as I can see is still out on that one as there is yet to be any independent technical analysis. (There are a number of early adopters saying that they have downloaded and listened to un-decoded MQA material and claim it is slightly better than regular CD)
 
5. They claim that MQA decoded will give a better audio quality than CD and be on a par with "Hi-Res files" or better.  Again independent analytical study is yet to prove this to be true or not (again early adopters are already claiming that their downloaded and decoded MQA experience is providing a significant improvement over regular CD and generally on par or slightly better than "Hi- Res"
 

 
Did you see the noise graph posted above? It has 12 dB more noise from artifacts than 24bit/192khz PCM.  Clearly:
 
a. They haven't dealt with the noise very well
b. How is that in any way "on a par with "Hi-Res files" or better?
 
 

6.  They claim that what the artist or engineer signs off is what you get.  This clearly depends on the integrity of whoever signs of the file but the MQA decoder does not light up if the file is not bit perfect compared to the signed off file and this is proving to be the case by the early adopters.
 

 
Being bit perfect isn't the issue. I can sell you a bit perfect file that is a perfect representation of a horrible music production chain with bad editing, compression, and poor filter choices that MQA can't undo.

 

[Gringo]

Sorry watchnerd if I gave the impression that I was saying that you referred to Meridian as money grabbers, that is a term that others have used in this thread, forgive me if I don't specifically point fingers on this point.

Audible lossless is not a term that Meridian made up but rather one that I used to try and explain a point, (obviously not very well) so your statement "clearly a made-up term, designed to disguise the fact that MQA, in certain scenarios is at least partially lossless" is not really fair, plus they have gone on record to say that it is not lossless in the strictest sense.

1. I bow to your superior knowledge regarding the noise floor, I had always been lead to believe it was random meaningless 0's and 1's that served no purpose. Clearly from what you say, it does have an impact on the sound quality, though I have no idea what as I obviously have never heard a music file without the 0's and 1's

3. Yes MQA files are slightly bigger than CD's,(I see no point in getting bogged down in semantics) but this is overall minimal and they have never claimed parity of size with CD's. As far as I am aware Meridian have only claimed that streaming MQA compared to Hi-Res is more efficient because of the smaller data files

4. Those that claim MQA is slightly better undecoded than standard CD have documented their reasons for this this on line for anyone to read. This is obviously subjective which is why I used the term "claim". I personally don't know so that is why I have not offered an opinion.

5. Similar comments to point 4 apply to the decoded files they have downloaded. I agree your tests seem to say something quite different. Either you or they are wrong or the data being presented is not strictly valid in some way ( I don't know either way and I am happy that you will believe perfectly reasonably that you are correct), I don't know how Meridian have done what they claim but I now thanks to your efforts have an idea of what they might have done.

6. On this point we are clearly in complete agreement and as I said - it is all down to the integrity of the person signing off the material. Unless proved otherwise, I am happy to accept that you have integrity and what you say is true, in the same way I think it is wrong to automatically assume Meridian don't have integrity just because I don't understand the workings of their technology which seems to be the stance of some people's comments on this forum

DRM keeps being mentioned on the thread but I am unsure as to why as again, Meridian have stated very clearly on more than one occasion that MQA does not encorporate DRM

I would be interested to know if you listened to the undecoded and decoded MQA files, whether it was before or after conducting your tests and what your thoughts were regarding its musicality.

You have clearly gone to a lot of effort in the interesting tests you have conducted and that is very much appreciated, especially as you have gone to the bother of borrowing or purchasing an Explorer2

As I have said before, I am not clued up on the technical side but I have heard MQA and was very impressed. The only reason for my making comments are that I percieve there have been many misunderstandings and claims from those who have not heard MQA and some appear to have an over eagerness to dismiss something which might just possibly offer a real step forward.

I personally am sitting on the fence and am happy to let time prove if it is worthwhile and takes off or not
I'm now off to listen to some music