MQA is Vaporware

[John Dyson]

2 minutes ago, Ishmael Slapowitz said:

John, don;'t bother. It's wack a mole with these MQA bots.

I truly cannot believe it -- I haven't heard such disconnected nonsense as from this GUTB guy.  Yea -- my previous post was a goodbye to him.  The statements from GUTB aren't worth the bits.

 

John

[John Dyson]

4 minutes ago, Ishmael Slapowitz said:

So. NEVER in your home. And you don't even own a tape machine? Ol. 

I couldn't even do my project without audio from master tapes.  In reality, master tapes aren't that big of a deal -- except for security -- it is easier to listen to material that has been properly mastered anyway.  A person really doesn't want to listen directly to the 1960s through 1980s master tape audio  -- sometimes the tapes by themselves don't sound very good unless correctly processed...  Consumers might mistake a tape distribution copy as being the same as a master tape, which they are not.  Even the common, improperly mastered material has been tweaked to be listenable.  (This is especially true of the digital copies made from original analog masters -- by Library of Congress procedures aren't necessarily NR decoded -- and that is a real problem for consumers getting partially mastered materials.)

 

I am usually an open book -- examples too long/etc, but one thing that I can not/will not do is to share master tape contents.  That could cause real trouble, and I want to keep my relationships in good standing!!!  I treat that kind of material very similar as when I worked at a military installation. (I even keep seperate storage locations/etc.)

 

John

[John Dyson]

12 minutes ago, KeenObserver said:

For those in the know, are not all tapes being archived as PCM?

DSD type stuff is more of a consumer thing.  The recording people that I deal with tend to live in the PCM world...  They seem pretty much settled on 192k/floating point, but often archive in 24bit.   I sometimes get floating-point digital master copies and sometimes 24bit, but I think 24 is more common right now.  Floating point is more for the 'processing' where there might be an overshoot or dynamic range issue -- and it is nice to be able to ignore those things when really busy on more important problems.

I always use floating point for my own processing -- never store in anything but 24 bit.

 

John

[John Dyson]

1 minute ago, Ishmael Slapowitz said:

Well said. 

 

Occasionally a master mix is just right and needs to polishing. But in general they can sound "flat".

For older material, they can sound 'encoded', with poor documentation (guess which NR to use.)   Luckily, when I need one, they have been in good shape -- I know the calibration levels or the tones are intact/etc.  I have heard horror stories though.

 

John

 

[John Dyson]

3 minutes ago, Ishmael Slapowitz said:

Well said. 

 

Occasionally a master mix is just right and needs to polishing. But in general they can sound "flat".

Oh yea -- I really shouldn't talk down master tapes -- I have some REALLY REALLY beautiful ones also.  Mine are all DolbyA encoded, for some reason...

 

John

[John Dyson]

16 minutes ago, Ishmael Slapowitz said:

I am guessing from 1872 on....nothing before that from what I understand. 

DolbyA started appearing in about 1966, and some of the older 3 track stuff was converted to DolbyA masters.   The first DolbyA units were the A301 (actually used diodes for gain control), but the up/downgrade to JFET gain control happend a few years later.  It would seem that the FETS would be better, but they were more 'tweaky'.   All of the Carpenters are on DolbyA, ONJ, ABBA, Bread (beautiful).   I have some Simon and Garfunkel, ca. 1966 recording (these aren't master tapes in my posession, but were originally done using DolbyA.)  Very luckily, I do have some playouts that leaked into the consumer realm (I mean REALLY CLEAN copies -- they just copied them!!!  Crazy!!!  Consumers have been putting up with some bad undecoded stuff.)  I have some ABBA that doesn't even sound like ABBA at all (NO 'ABBA' distortion!!!)  The hard part -- trying to figure out the decoder calibration without tones.

 

  Itis a real historical downgrade when they copied the 3 trks to DolbyA.  Nowadays, if you can find them (and don't have my decoder), it  is best to go back to the 3 track masters.  DolbyA is the proverbial deal with the devil -- and you know I am trying to fight that devil.  (Back in the '60s/'70s, the deal with the devil seemed like it was good tradeoff -- but, now since our quality demands are higher, the DolbyA modulation distortion is just not a good thing.)  The DHNRDS DA gouges out as much modulation distortion as possible -- really tricky math -- and I am not trying to siphon money from people :-). 

 

 

John

[John Dyson]

56 minutes ago, FredericV said:

I'm not saying I don't believe in hi-res. But my theory is that hi-res sounds better because of better mastering. I did some experiments with sox on extreme high-end systems, and nobody could detect when my sox 16/44.1 "filter" was in the loop. So recordings from AIX in 24/96 vs 24/96 downsampled with extreme sox settings to 16/44.1 and then upsampled back to 24//96. I even once posted these files on some fora and nobody could do it, except when looking at the spectrum, and one Meridian dealer claimed to be able to hear the difference. All other users who tried it, failed at it.

 

(Much of your comment elided)

 

IMO (not actually humble :-)) -- the major difference between almost any recording releases IS mastering.  A very slight difference in mastering will overwhelm the effects of almost any change in pure resolution beyond 48k/17bits (I goosed up the numbers becuase of being VERY conservative in my claim.)  Very important in this comment -- I am ONLY speaking of dithered resolution with proper anti-alias filtering.  I am NOT speaking here of purposeful defects in the audio signal (e.g. lossy encoding/decoding methos.)

 

Even a filter tolerance issue can make an audible difference in mastering -- much greater modification than the slight difference because of >=48k sample rate or 17bits resolution.  (I think that 44.1k/16bits is too close to the edge to make the claim for everyone - in my case, even 40k/15bits would be sufficient, however.)

 

Per my work while trying *right now* to produce some very clean/precise test material for the C4 decoder project -- I can make mastering differences that you cannot likely detect in a spectrum analysis, yet the material sounds singnificantly different & totally overwhelm increases in resolution beyond 48k/17bits.)

 

In this note, I am not making a claim that 192k/24bits might not provide a small increment improvement in audio quality -- that claim isn't the purpose *of this note*.  The claim is that small differences in mastering will often be much more obvious than the small differences in results from digital resolution.

 

Probably most importantly -- when comparing something that is of mp3 vs MQA vs accurate PCM -- make sure that the mastering and treatment of the signals are absolutely the same.  Tweaks need not apply.   Purposeful sweetening before comparison is deception and would be misleading if being presented as a an accurate comparison between the signal encoding methods.

 

I sure *hope* that marketing people and advocacies aren't using 'fake-superior' (sweeter) masterings for the purpose of comparing digital encoding/compression methods.  Also, I *hope* that the methods themselves don't distort the results with a built-in sweetening/distortion.  I have found that signal modifications need to be made very carefully on a somewhat conservative basis -- very often, some change that sounds good on first listen can sometimes be very tiresome and fatiguing in the longer term..

 

*People doing mastering should think of themselves as final clean-up on an already finished product -- not to be super-creative themselves unless actually fixing a problem.

 

John

 

[John Dyson]

8 minutes ago, Ishmael Slapowitz said:

I'm telling you he may actually be sincere in his beliefs.

 

It is a fact that the most preposterous theory can be introduced to the public and there will be a percentage of people who will believe it. For example..911 was perpetrated by the US Govt or the Israelis, that the Lunar landing was a hoax,....

 

..and the mother of them all...the theory of Time Smear and De-Blurring...😎

I agree --

My guess is that MQA needs de-blurring because it 'blurs' the signal.  There is no 'blurring' in normal filtering unless renaming normal filtering behaviors with eccentric/incompetently used terminology -- I use linear phase filters all of the time, and they all marry together perfectly without weird timing/smearing problems (well, gotta understand DSP filtering to do it correctly.) :-).

The worst misconception from my own opinon (my opinion only) is the idea that 'Gibbs effect' is a kind of energy storage ringing, but as we all know it is just a residual effect.  Sure, it can cause clipping, but that is only because of the missing negative halves of higher frequency (rolled off) constituents of the signal.

There are lots of misconceptions out there -- and I find it odd that more people who actually know about signal processing haven't joined in more often to make the corrections.  Maybe they assume that some people are ripe to be taken advantage of by misleading and inaccurate 'ad-hoc' terminology.

 

John

 

[John Dyson]

14 minutes ago, Ralf11 said:

 

You did not say that however.

 

I have far better ways to "assist the world" than by editing wikipedia.

Best to quit feeding the troll -- he is obviously not intellectually honest or honest about what he/she says.  I have problems when dealing with people who have no integrity -- it is often painful.  My suggestion is to 'shun', and answer with a 'disagree' response.

 

John

[John Dyson]

I have an answer to the above 20kHz content matter.  Just like the actual DolbyA HW, I can disable some of the anti-IMD code in the DHRNDS DA, and decode DolbyA stuff (even stuff limited to 20kHz), and produce all kinds of material above 20kHz.  There *are* cases where there is above 20kHz material -- especially in the recording studio, but as the signal goes further down the chain, there is loss of information, and increase of distortion splats.

 

Nowadays, the situation CAN be better where there are no NR splats, but 'normal' compressors -- FET based compressors, and poorly designed digital processors can create messy material above 20kHz.   Such energy ABOVE 20kHz can very possibly create below 20kHz (audible) signal/distortion energy because of real-world analog amplification and nonlinear digital software (using limited precision in certain cases.)

 

There needs to be controlled tests, with ACTUAL above 20kHz materal -- nothing touched by NR processing or processing electronics or software that generates it's own above 20kHz components...  Do full ABX analysis (any micro-sized change in mood will show differences) on PURE material where VERY LITTLE of the signal above 20kHz is distortion components.  Also, make sure that the analog electronics and transducers do not create below 20kHz signals by above 20kHz signals (usually creating distortion wrapping back into audible range) -- this is a major cause of the ability to detect diferences in above 20kHz content.  Any digital filters used to limit the frequency range MUST be done at least to 24bit resolution and 32bit/64bit would be best if possible.  (some kinds of apparently linear math can create distortion when implemented in software -- BE PARANOID.)

 

The experiment MUST BE done *scientifically*, with equipment that has been tested to be low enough distortion that PURE above 20kHz sine-waves and IMD between two reasonably low above 20kHz frequencies (25kHz/26kHz type things)  cannot be detected by golden ears.  There is equipment out there, but would probably disqualify a very large amount of even very highly regarded transducers and analog amplifiers.  (Refer to the op-amp document that I have made available in the past as to how much distortion even highly regarded amplification devices can produce.  The distortion in ACTUAL circuits can be greater than superficial spec sheets might show.)

 

So 1) control the experiment so that the 'golden ears' are not able to detect above 22kHz material.  This might be tedious because of the general behavior of equipment (distortion components reaching down into the audible range in many -- BUT NOT ALL -- cases.)  Make sure that IMD is also checked for, e.g. 25kHz/26kHz.   (If there is no success in finding the qualified equipment, then this experiment fails as a 'nothing learned'.)  Make sure that the transducers CAN pass the above 20kHz signals, just that they don't create FAKE/DISTORTION audible components below 20kHz resulting from the above 20kHz actual signals.

 

2) run the experiment with pure material (no auto-distortion creators like DolbyA HW, FET compressors, Limiters, expanders, effects devices touching the signal.)  Use only equipment qualified by phase (1) above.

 

3) perform ABX tests, using something like 18kHz, 19kHz, 20kHz, 22kHz, 25kHz linear phase filters (must be linear phase -- the relative phases and timings are changed by non-linear phase filters, even though they might sound better to some people.)  Also, verify that these filters do not create their own distortion components (use of 16bit precision math is not good -- should be at least 32bit floating point -- 24bit signed MIGHT be good enough.)

 

The above set of conditions for the experiments aren't 100% complete, but you get my drift.  ALL variables must be controlled, but without excluding success for the detection of a difference.  That means, if equipment cannot be selected -- then the test must be declared a failure either way.  It is so easy to miss the effects of any component in the test.  It is even a good idea to make sure that connections (yes -- CONNECTIONS) dont have nonlinearities that might be caused by oxidation/bad connections.  (RF distortion can be created by bad metallic connections in wiring near a transmitter, for example.)  Any source of IMD and distortion must be controlled.

We are talking about the need to control very small relative distortion levels in this kind of experiment.

 

 

* A critically important part of the test is when using 'headphones', that the sound energy output from the headphones does not include significant BELOW 20kHz components when the headphones are provided with ABOVE 20kHz components, including in IMD tests (multiple ABOVE 20kHz frequencies.)  So, we don't want a 25kHz+26kHz signal input to the headphones then produce any below 20kHz energy.  This would show NONLINEARITY, and therefore potentially disqualify those headphones in the test.   Of course, REASONABLE levels (yet to be defined) would be allowed.

 

 

John

[John Dyson]

8 minutes ago, sandyk said:

 

I note that you didn't even have the common decency to reply to my in depth reply to your  PM.

That says far more about you than it does about me.

 

 

 

 

I like Alex, and most of the other people on the forum seem to be good guys.  The ONLY answer is to do a careful experiment.  My earlier suggestion is a *first step*, but if we can qualify the tests adequately, then the answer can be obtained once and for all.

There *are* cases where transducers (headphones) and amplifiers can distort results.  This can make these discussions UNFAIR for all.

 

Even if above 20kHz energy cannot be heard, it is VERY likely that a *difference* of some kind can be heard.  We need to control the experiment so that everyone ends up being shown to be honest, forthright and etc etc.   The answer can be had, but it is so important for those with interest in the issues to 'sit back', 'carefully control the experiment' so that EITHER hypothesis can succeed.

 

It is tricky to do the tests correctly -- either of 'can hear' or 'cannot hear' must be able to succeed, but the test setup must also reflect exactly what is being disagreed about.

 

John

[John Dyson]

1 minute ago, FredericV said:

 

Ad hominem. You still haven't explained why the internet would be degrading my audio files ... while I have shown it's not the case.

I really believe that a good experiment needs to be done (controls), or all it will be is hurt feelings and butt heads.  I have my own opinon on the matter...  Rudeness is awfully hard to avoid when two parties really believe in their case.  Linus and I used to have knock-down/drag-out fights.

 

Maybe, if I can get up the energy and spend the several hours, create an experiment that is FAIR (both sides COULD result in success), but also finds out what is going on.   I believe that Alex IS hearing a difference -- but it needs to be analysed.  I dont' think that the situation is made up.

 

John

[John Dyson]

Just now, Ralf11 said:

John, I'd rather see you put that effort into USB cables.  Might be something going on there...

I don't do 'USB' cables :-).

 

BTW -- I misunderstood the context (wrong experiment), but the experiment idea is valid.  I do believe that there is a difference in the way that the software/hardware is working for differing file types (for Alex.)   I think that I trust Alex enough to know he is hearing a difference -- but most of us aren't EE/analog/DSP types like I am.  SO, I would take it for a different situation, and might not even had mentioned what difference I would have heard -- because I would have understood the source of the problem.

 

* I now understand that it wasn't the res issue, but more of a file format issue.

 

It is SO WRONG to misinterpret Alex as being 'off the wall'.  These crazy computers can have AMPS/many Watts difference in draw for differences in programs being run.   I am not claiming this to be the issue - but there IS an honest difference.

 

John

 

[John Dyson]

5 minutes ago, Ishmael Slapowitz said:

Please, c'mon now...perhaps a few pints before this last post?

No, I have seen too much good from Alex to immediately assume that he is not hearing a difference.  I know that on my system, I wouldn't hear a difference ---  I interchange 24bit flac/24bit wav, 32bit fp wav, etc freely.   There are no audible (or almost any effects as to the values -- minor for fp difference) differences.

There is something going on in the equipment/setup (I am 10kmiles away -- cannot evaluate it) that I dont' understand.

 

If Alex is hearing something 'wrong', there is a reason for it -- but it would be wrong to claim that it is a personal issue.

 

John

[John Dyson]

Off topic:  when people have strong opinons and butt heads -- after a while, it is best just to cool off and accept that there are sometimes extreme disagreements.

I still see some skeptics about my own projects make inaccurate claims -- but I try NEVER let it become personal -- *I have had erroneous opinons too often in my life/career also.*   My method is to present facts, but if in the end, there isn't consensus agreement, then I try to pull back on my arguments without losing my integrity or forcing others to lose their own integrity.

One comment:  part of integrity is to accept the situation where one's mind can be changed!!

 

John

[John Dyson]

17 hours ago, firedog said:

There is a difference with audio: One can claim that no bits are changed in transmission, but noise conducted along with the data can effect the DA conversion adversely. There’s not a lot of evidence with this other than anecdotal listening observations, but it isn’t impossible. 

The raw digital data cannot transfer the 'noise'.  However, ground-bounces (for example) or coupled (when strong) or radiated (in the extreme) can add noise to the results of digital to analog conversion.  This is mostly related to grounding/shielding/etc, and very critical with the current state of technology (speeds involved, circuitry thresholds/etc.)   Imagine using the point-to-point wiring on the original color TV sets?  No-way that current circuitry could work well with that kind of wiring technique.

 

So, a schematic nowadays is NOT enough to describe a circuit, and certainly not enough to build a circuit unless coupled with a competent person doing the layout.

 

When working with raw audio, non digital, nothing above 40kHz, then the grounding/shielding issues only require good common sense and understanding of ground currents.  (Back in the day, hum and distortion could easily increase because of poor grounding -- shielding was not much of a problem unless RF was involved.)  Nowadays, there is so much crazy stuff in the air and on the grounds -- it is really difficult to separate all of the issues.  It *almost* becomes metaphysical (but not really -- just good physics!!!)  Also, in the old days most old audio preamp-circuits were fairly well separate from the power circuits -- the old audio preamps might have had class AB type output circuits with mild/audio-rate-only changes in ground currents (such audio-rate current changes easy to deal with.)   Nowadays, we have all kinds of fast digital noise bouncing around, and in the case of A/D & D/A conversion, there has to be a careful marriage between the analog ground world and digital ground world.  Those evil noise sources (and admittedly 'jitter') can and do mostly happen at that 'marriage point' -- not in the digital circuitry itself.

 

John

[John Dyson]

2 minutes ago, firedog said:

Yes, but there are Ham engineering papers talking about analog interference being conducted on digital cables, along with the digital signal. Not a claim the bits are changed, a claim the conducted noise has a negative effect on the results. Don't ask me for links, I looked this up years ago and aren't going to spend time on it again. 

Electrical interference can definitely be conducted -- and it is a real problem.  However, the actual noise  phenomenon that would normally be manifested on audio circuits would be analog.   That is, noise on the digital signal itself -- if the analog & digital circuits were perfectly separated would only create errors, not noise.  If the noise on the digital wires was allowed to interfere into the analog side of the circuit -- which it CAN if not carefully designed, then that can cause analog noise.  Also, the noise on the digital side of the circuit could theretically cause noise in the D/A itself -- which would be a design flaw, and mostly minimized.  However, again the syndrome is NOT digital, but is an analog interference that leaks from the digital part of the circuit into the analog side.  This scenario is usually well managed by the chip manufacturer.

 

The problem most often would be ground noise, where the analog ground is negatively affected by digital noise (digital currents allowed to flow through the analog ground.)   Coupled (not to be confused with radiated) interference can happen also.  This is where the noise on the digital wire is coupled through capacitance or inductance directly to the analog signal.  Radiation noise SEEMS the same as coupled noise, but actually isnt -- this includes different kinds of shielding.  It is best to try to avoid the need for shielding, that cannot always be done.

 

The ground noise issue is managed both by the circuit board designer and the chip designer.  (The chip designs are optimized even to the point that pinouts dont' make the problem worse.)   Ground noise worries me A LOT -- coupled noise is easy to visualize.  Ground noise requires seeing the whole circuit at once, making sure that there aren't any grounding shortcuts taken.  (Problem like -- oh, it 'goes to the same place' on the schematic...  Going to the same place and being at the same place are two different things at high frequencies -- much like relativity.)

 

The ham radio docs ARE a good source of conceptual info about the noise problems.  The noise problems are real.  There can also be noise problems that seem like jitter, but they don't happen in the way that the con-artists (profiteers) might suggest.

 

John

[John Dyson]

4 hours ago, rickca said:

Tencent is negotiating to buy 10% of UMG with a one year option to double its stake.

https://www.marketwatch.com/story/tencent-in-talks-to-buy-universal-music-stake-2019-08-06?mod=mw_theo_homepage

I'll bet you that the country China and Chinese countries will invest heartily into technologys that can be bent to protect their own intellectual property while still manipulating (both ethically and unethically) to drain Western originated IP.

[John Dyson]

11 hours ago, FredericV said:

 

Auralic seems to have written their own lightning player which relies on sox (as analyzed by myself while intercepting a firmware file during download when I launched an update of my Auralic device, and later confirmed to me by their CEO that they use the sox library), while they also had a dormant version of MPD in an old version which they later removed.

More evil is adding a closed source MQA decoder to open source software by means of a bidirectional pipe sharing intimate data structures between both. According to the GPL FAQ such (ab)use of a pipe is problematic. The now gone MPD mailing list had some fine examples of these violators ... at least I feel a little bit more enlightened after seeing those source as posted on the mpd-devel list.

If using pipes with GPLed software is problematic -- then GPLed software becomes useless on Unix-like OSes.  I understand the matter, and frankly, if/when I do that, it will be using json (not fond of XML like schemes for various esoteric & low level protocol reasons) with a defined protcol so other programs can use the interface.  At that point, the GPL religious become (ab)surd in their complaints.

 

John

[John Dyson]

As someone who ONLY has a slight technical interest in the math games done by the MQA algorithm, and a dislike for anything that messes up audio quality -- here is my own impression of that scheme.   I am sometimes a bit confused -- so I'd like for both pro-MQA and anti-MQA people to correct my impression:

 

Proposed sales disclaimer for MQA 'processed' product:

Here, I am selling you a modified quality (less dangerous) digital recording than what you used to purchase, and this new, safety improved version is what you now get by default!  However, if you want to *listen* to the less protected quality closer to you have previously purchased and as a legacy & still might expect, now we help you do so with safely 'approved' adjunct technologies.   The previous dangerous, direct digital version of the recording, not having safety 'authorized' processing,  without the important special 'folding' of any kind, is no longer available to cause anyone harm.  With the more sophisticated MQA technology, you will not even have mistaken access to the mostly still dangerous, mostly corrected and carefully unfolded digital signal.   Additionally, you will now benefit from ONLY indirectly (no direct payments needed!!! yay!!!)  paying the MQA license holders for the mechnanisms that improve  the lesser quality digital signal  to a 'better' version, which is in the safety listen-only 'lockbox'.  That is, these sophisticated algorithms will return the recordings (for listening, safely in a lockbox) to a state is just about as good as the more dangerous, original digital signal.  When further, dangerous, unlicensed consumer signal processing is needed, you'll be protected by utilizing the undecoded, still safely folded material by default.

This is a breakthrough for the audiophile consumer, don't you think?

 

If my understanding is close to correct, the effect on the consumer is that it seems to give the 'very beneficial' 🙂 opportunity to pay more indirect license fees to the MQA license holders.  We need MORE research like MQA!!! 🙂

 

 

 

[John Dyson]

6 minutes ago, 4est said:

Please don't ask me to listen to Abba, pretty please?

NO!!!!   Abba is 'test material'.  Do you know how nuts it makes a person listening over and over again to obliterate modulation distortion (ABBA by default is full of it.)  ABBA is *test material* only nowadays.

 

We can do Olivia (pretty good quality), Carpenters (limited by the material), Linda Ronstadt (limited by the Aphex 'distorter'),  Simon & Garfunkel (not an ideal example of super high quality -- but cleaner than most available copies nowadays), 'The Cars' (surprisingly clean/clear/undistorted), a bit of Eric Clapner :-), some Roberta Flack (not 100% perfect), Bread (really nice and mellow), etc....   I can even do Taylor Swift (Shake it Off), but the DolbyA seems to have been used for effect -- not noise reduction.  Can even do Petula Clark (Downtown, Don't sleep in the Subway -- underpass) -- but highly processed stuff from the 1960s is not so good :-).

 

John

 

[John Dyson]

2 minutes ago, daverich4 said:

From the September 2019 issue of Stereophile...

 

”MQA files will still unfold to their full resolution (although it must be acknowledged, without going into detail, that above 96kHZ, an unfolded MQA file is not equivalent to a PCM file of the same specification).”

 

Is that because of this thread? Could be. At least it’s the desired result. 

Mp3 and opus can do 24 bits or so -- why aren't they as good as lossless PCM?  MQA is in the same boat as MP3/OPUS, simply use a different data reduction and re-interpolation scheme (I'd guess -- that is what compression does.)

 

Lots of bits with error -- that is called dither -- more or less effective.  Probably 'less' effective than true dither because of biases.

 

John

[John Dyson]

9 hours ago, Ishmael Slapowitz said:

Sorry he had a perfect opportunity to qualify the "24/192" unfold claim. He did not. 

Every time that I read 'unfold', I cringe.   'Unfolding' is imperfect.  (I am not disagreeing with the previous comment -- just that 'unfolding' sounds too nice.)

 

Unfolding is first IMPERFECT, secondly -- UNFOLDING would not be necessary unless there was the ugly 'FOLDING', which by the way that it is described -- is a loss of data with hints.   Plucking coefficents from a DCT/MDCT based on stats/levels is also that kind of lossage.

The "fold -> unfold" is lossy -- similar to "dct -> pluck out coefficents -> idct" being lossy.

 

Some of the math might be more 'natural' or 'built-in' to the system with folding/unfolding -- maybe, maybe-not, but that doesn't make it audibly good.   If the result is incomplete after 'unfolding', then the 'noise' had better be statistically close to dither, or it can produce problems.

 

John

[John Dyson]

9 hours ago, lucretius said:

 

Yes.  But I wondering why 2L even bothers to MQA 24/352.8 files.

I am starting to believe that there is a visceral dishonesty in the audio world from the beginnings of 'digitial'.   Imagine all of the commonly available 192k/24bit material that is so poorly mastered (or not mastered), yet is being sold as implied premium product.  The situation is kind of sick all of the way around.   All of the upsampling/extreme excess-precision stuff (I mean, truly upsampled or in-some-cases unneeded high bitrate to sell, not for DSP convenience) is deception, pure and simple.  There is SOME true  high quality 192k/24bit material for sure, but my expectations for POP material are very, very low now.

 

I am wondering if this dishonesty is a simple and cynical extension from the music distributors world of deception (or maybe im some cases, ignorance.)   I am ALL for the best quality being distributed, but frankly poorly mastered material being sold as 192k/24bit is NOT 'gilding a lily', but rather the more gross 'polishing a t*rd.'

 

Again -- I know that the main discussion is about MQA -- but this very short diversion is about the systemic marketing dishonesty about bitrate.   There is also GOOD 192k/24bit out there,  but there is a lot of specsmanship about 192k/24bit (or higher) also.  I am hazarding a guess that the 'specsmanship' is more prominent than true high quality material -- MQA only makes the situation worse.

 

If your equipment plays pop material realtime better with raw 192k/24bit PCM, then in some cases, go ahead and up-convert your own 'best' material from 44.1k,48k or 96k or from flac/other lossless compression.   However, paying even a $5 premium for someone else to do the upconversion for you -- well, in some cases, the audiophile can do it better than what is too-often sold.

 

Apparently,  MQA nonsense about providing better the 96k (when even the 96k is in question) is a bit embarassing for all involved.  It is also likely that the audiophile doing his own dithering and upconversion is just as good or better than what anyone else can do -- esp when the target is human hearing.  Anything other than true pseudo-random noise or actual signal in the lower precision and/or higher frequencies is tantamount to distortion.

 

John

 

[John Dyson]

1 hour ago, mansr said:

That is trivially incorrect since there is no such thing as filter ringing. A steep filter does, however, necessarily have some ripples.

 

The ripples occur at the corner frequency of the filter with an intensity proportional to the signal content at that frequency. Whatever spectral content music contains above 20 kHz is always accompanied by much higher levels at lower frequencies. Even if you could hear the 20 kHz "ringing" by itself, it would be masked by lower frequency sounds whenever it actually occurs.

 

Since "ringing" isn't a problem, nothing needs to be done in order to "deal with" it.

 

See also https://troll-audio.com/articles/filter-ringing/.

I 'upvoted' your comment -- but only one VERY MINOR modification -- I agree that the FIR filters used for LPF dont 'ring'.   it isn't that all filters don't ring (IIR or analog filters can ring -- store energy and continue with usually decayed oscillation.)   I didn't read the URL in the message, but the 'Gibbs effect' which is a residual from a series truncation is what people sometimes mistake for 'ringing' on waveform displays.  That 'Gibbs effect' doesn't result from energy storage in the same way that true ringing does -- Gibbs effect is simply a residual from a truncation of a series of sine functions (sometimes comprising something like a square wave) that leaves a left-over of 'undulations' that look a little like sine waves.  That 'truncation' of 'sine functions' is basically a constant delay low pass filter.

 

There can also be a difference in sound between a minimum phase vs. constant-delay-vs.-frequency filter, but it is a matter of relative time of arrival of the various signal components, not 'ringing' at all.

 

I suspect that the combination of the 'sine-like' residual signals (Gibbs effect), and the slight difference in sound character of filters with different kinds of time delay behaviors can trick non-technical people into believing that they are hearing the effects of 'ringing.'

 

John