[bigshot]

It's arguable that time recognition extends beyond what is achievable by normal hearing bandwidth and losing bit depth can have a similar theoretical effect.

Are we talking about listening to music on home audio equipment here, or are we talking about brainwaves? If we're talking about listening to music, the normal perceptual thresholds are the best I can detect and I'm quite sure that's true of all other human beings too. Bats and dogs might feel differently, but they can buy their own stereos.

 

[Strangelove424]

Now I or you are confused. I was talking about a possible domestic ADC. So the pubic can play. What is the issue there?

It's always good when the pubic can play. Couldn't resist. Anyway...

I originally asked if there were software encoders on the market, not ADCs. Software encoders allow for digital conversion from PCM, not just analogue. The purpose of which would be to make comparisons of PCM files rendered to both MQA and FLAC. Not MQA's sample files, or people on the internet decoding and re-encoding using their own strange methods (including archimago's unfortunately). The structure of MQA has made this impossible. Or disregarding comparison, what if I was simply not in a position to record direct to MQA, but wanted to include MQA as one of many delivery codecs? And, again, don't tell me that MQA is an analogue to analogue only process that needs an ADC in order to be authorized, because the existence of MQA versions of digital-only masters on Tidal completely disproves that.

 

[pinnahertz]

Actually I think they can do this without altering the phase, and it is this that is new.

Wow, that would be new! Altering time response without altering phase. Too bad the two are inseparable. But if you have a link to where they make that claim, I'd be happy to read it.

Say you want a low pass filter. One pole with give you 90 degrees phase shift, and 45 at the 3dB point. You cannot fix that.

Incorrect. That phase response, and many others, can be compensated for with a properly designed all-pass filter. An all-pass filter is a filter with flat amplitude response but non-flat phase response. It can be constructed with multiple poles which, if properly aligned, can compensate for the phase response of another filter. It's not a new idea, it's been done for many years. And that all could be done in the analog domain, so even easier and more precise in the digital domain.

An aside: all-pass filters have found many applications in audio. Two major ones would be the encoding process for quadraphonic 4-2-4 systems, and the encoding process for Dolby Stereo/ProLogic, also a 4-2-4 matrix, though neither of those applications involve correcting filter phase response.

If you want 2 pole you get double that.

Also incorrect. The specific phase response of a high-order filter depends on its alignment, and it changes radically from something like a Butterworth vs Elliptical, for example. Both topologies can have the same cutoff frequency and same number of poles, but entirely different amplitude and phase response.

Unless you do one of the poles in the digital domain, and you do it effectively backwards in the time domain (which you can do if you have the file).
Then the phase in the forward playback direction is reversed. Add this to the other filter and you get cancellation, zero phase shift, and a 2 pole 12dB per octave low pass. Magic. It is FILTFILT in Mathcad

Obviously MQA version is a bit more complex, as they are compensating for a bunch of different ADCs and DACs.

Remember two things about that compensation. First, it's compensating for an unknown group of cascaded filters (not the ADC or DAC, but the associated filters), and second, that compensation has no confirmed audible effect.

Everyone seems obsessed with fixing the frequency response, which is fine, but all the f-ing about with EQ screws up the phase response as you say above, especially in the LF.

No, if you "fix" HF frequency response properly you don't mess up phase response, especially if the "fix" is complimentary. If not, the phase response changes in the HF band, not LF. It's pretty hard to change phase response in LF without doing something pretty radical like a filter.

 

[pinnahertz]

You know not to whom you speak

It's a Forum. We aren't using our real names or filling in details in our profiles. It works both ways.

so don't be so sure about what I do or know.

Apologies if I've judged you. I only have what you post on which to base any assumptions. You're not giving me much. Perhaps if we actually knew each other we wouldn't be quite so argumentative. Oh well, so goes the forum.

What their process does to anti aliasing filters would have the effect I'm speaking of in the overall signal processing. They're likely lining up peaks which would save bits and allow them to claim better timing or correction. We're not really in disagreement here but like you, I don't think their claim can be accomplished and is just some 'optimizing'. My opinion is that it affects more than just their claim.

I get your point. The way I see it is you can't correct for what you don't know, and each filter has a certain time characteristic. Cascading them doesn't cause those characteristics to add linearly unless the filters are identical and there's no absolute guarantee of that in any recording. So the correction needs to encompass everything, which they really don't have, don't know, and therefore can't do. I think assuming you're correcting the first ADC filter does nothing because of all the processing and filtering that goes on down stream.

Whether you're mixing or mastering, you are still dithering and rendering files.

No, not necessarily dithering or rendering until the final mixdown. There's no need to dither when the internal DAW processing is 64bit FP math.

 

[jagwap]

Wow, that would be new! Altering time response without altering phase. Too bad the two are inseparable. But if you have a link to where they make that claim, I'd be happy to read it.

I didn't say that. I was referring to "for a general idea only, you can think of using only an analog EQ to low pass. when you mess with FR you also change phase, and vice versa."

Incorrect. That phase response, and many others, can be compensated for with a properly designed all-pass filter.

Only up to a point. At some point the correction higher out of band the correction has to end.

An all-pass filter is a filter with flat amplitude response but non-flat phase response. It can be constructed with multiple poles which, if properly aligned, can compensate for the phase response of another filter. It's not a new idea, it's been done for many years. And that all could be done in the analog domain, so even easier and more precise in the digital domain.

An aside: all-pass filters have found many applications in audio. Two major ones would be the encoding process for quadraphonic 4-2-4 systems, and the encoding process for Dolby Stereo/ProLogic, also a 4-2-4 matrix, though neither of those applications involve correcting filter phase response.
Also incorrect. The specific phase response of a high-order filter depends on its alignment, and it changes radically from something like a Butterworth vs Elliptical, for example. Both topologies can have the same cutoff frequency and same number of poles, but entirely different amplitude and phase response.

OK, but I was trying to keep it simple. More poles, more phase shift, unless this less conventional approach. There are less experienced people here and if I say "2 cascaded butterworth filters to make part of a Linkwitz-Riley crossover" it would fit the pendantic, but lose many.

Remember two things about that compensation. First, it's compensating for an unknown group of cascaded filters (not the ADC or DAC, but the associated filters), and second, that compensation has no confirmed audible effect.

But you can correct the known items. If those are within your closed system they add less errors over all. If ADC and DAC are known (and their associated filters), then this much can be compensated, if there is anything to compensate.

No, if you "fix" HF frequency response properly you don't mess up phase response, especially if the "fix" is complimentary. If not, the phase response changes in the HF band, not LF. It's pretty hard to change phase response in LF without doing something pretty radical like a filter.[/QUOTE]

I didn't say an HF filter changes the LF. I was saying EQing the LF screws up the phase and timing a lot more than the HF. Look at a non steady state waveform through a sub crossover some time. Distortion doesn't cover the damage. Corruption is a better description.

The technique I outlined is Very complementary as you can have zero phase shift DC to light. All pass cannot do that.

 

[pinnahertz]

I didn't say that. I was referring to "for a general idea only, you can think of using only an analog EQ to low pass. when you mess with FR you also change phase, and vice versa."

Did I misunderstand this?

Actually I think they can do this without altering the phase, and it is this that is new.

Only up to a point. At some point the correction higher out of band the correction has to end.

You said it was impossible before. Now the above sentence grammar is too ragged to decode. One more try?

OK, but I was trying to keep it simple. More poles, more phase shift, unless this less conventional approach. There are less experienced people here and if I say "2 cascaded butterworth filters to make part of a Linkwitz-Riley crossover" it would fit the pendantic, but lose many.

I'm ok with simple right up to the point where the statement is just plain wrong. You can call it pedantic if you like, I'm fine with that. But if it's wrong I'm going to object.

But you can correct the known items. If those are within your closed system they add less errors over all. If ADC and DAC are known (and their associated filters), then this much can be compensated, if there is anything to compensate.

Sure. As I said, they don't have that for the entire chain, and very very fortunate if they have even one filter known. The whole thing is rediculous.

I didn't say an HF filter changes the LF.

Did I misunderstand this?

Everyone seems obsessed with fixing the frequency response, which is fine, but all the f-ing about with EQ screws up the phase response as you say above, especially in the LF.

I was saying EQing the LF screws up the phase and timing a lot more than the HF. Look at a non steady state waveform through a sub crossover some time. Distortion doesn't cover the damage. Corruption is a better description.

I agree EQing a sub has some phase issues, but the actual amount of phase shift in a anti-aliasing LPF is massive by comparison. What does LF EQ have to do with MQA anyway?

The technique I outlined is Very complementary as you can have zero phase shift DC to light. All pass cannot do that.

You're going in circles, and the vertigo is killing us. First you say you can't correct a single pole filter's phase shift, then you outline a fully complimentary correction with zero phase shift DC to light (which is impossible of course). Which is it? Can we pin down some actual, real, practical things here?

 

[jagwap]

Did I misunderstand this?

You said it was impossible before. Now the above sentence grammar is too ragged to decode. One more try?
I'm ok with simple right up to the point where the statement is just plain wrong. You can call it pedantic if you like, I'm fine with that. But if it's wrong I'm going to object.
Sure. As I said, they don't have that for the entire chain, and very very fortunate if they have even one filter known. The whole thing is rediculous.
Did I misunderstand this?

I agree EQing a sub has some phase issues, but the actual amount of phase shift in a anti-aliasing LPF is massive by comparison. What does LF EQ have to do with MQA anyway?

You're going in circles, and the vertigo is killing us. First you say you can't correct a single pole filter's phase shift, then you outline a fully complimentary correction with zero phase shift DC to light (which is impossible of course). Which is it? Can we pin down some actual, real, practical things here?

I don't have time for the smaller points, but conventional filtering cannot compensate phase. Complementary filtering in the reverse time domain can pre compensate it exaclty if you want an even order filter. I have explained this a few times. It is my guess as to one of the things they are doing as they have this ability as they can work on the completed file. Sorry if you don't get it.

 

[pinnahertz]

I don't have time for the smaller points, but conventional filtering cannot compensate phase. Complementary filtering in the reverse time domain can pre compensate it exaclty if you want an even order filter. I have explained this a few times. It is my guess as to one of the things they are doing as they have this ability as they can work on the completed file. Sorry if you don't get it.

If by "smaller points" you me "accuracy", I get that. I'll just continue to put it right. Or, you could take the time to get it right the first time, your call. Yes, I do get it, as to what MQA is doing, or at least attempting. Yes, you can pretty much perfectly correct for one or more well characterized filters. If they know the filter, have the data, sure they can do that. And I get that you can post process a file and pre compensate. That isn't the problem I'm having here.

The problems are (in no particular order):
1. Posts with blatantly wrong statements (which then get reversed in later posts). I quote the post, then the poster says he never said that (though it was a direct quote). Then we end up agreeing, and I'm labeled "pedantic".
2. MQA's claim of being able to reproduce the original performance
3. MQA's claim that they can correct for "blurring", when the causes are multiple and unknown
4. that any of this makes any audible difference.
5. Yes, all-pass filtering can correct for phase. There have been examples for many, many years. Here's one: FM Stereo requires the use of a filter nearly identical to an anti-aliasing filter in a digital system. In FM Stereo that filter keeps audio out of the 19kHz pilot signal. Those filters were analog. FM modulation processing involves deliberate audio clipping, but the clipped waveform caused those filters to ring and overshooting the target 100% modulation limit, partially (largely) as a result of their poor phase response. One manufacturer remedied the situation by phase compensating those filters with complimentary phase all-pass filters. The resulting filter had far less phase distortion, and rang far less, and permitted higher modulation levels. That technique became quite popular. Early attempts at ADC anti-aliasing filters were also analog of course. During the early days of digital audio many blamed those filters for the harsh sound of digital, so several companies developed phase-compensated filters as an after-market upgrade. That compensation was also done with all-pass networks. The results were mixed, mostly because the harshness being complained about wasn't related directly to the phase response of the filters.

Just a few instances from history where analog filters were phase corrected using all-pass filters. There are more, of course. No, that kind of phase comp isn't perfect, and certainly if you were in the digital domain and had "time" you could produce a complimentary phase comp filter. That's not the argument here. Of course it can be done. And of course MQA is probably doing that. Of course, it doesn't matter because the results remain unproven as to their audibility.

 

[castleofargh]

Actually I think they can do this without altering the phase, and it is this that is new. Say you want a low pass filter. One pole with give you 90 degrees phase shift, and 45 at the 3dB point. You cannot fix that. If you want 2 pole you get double that. Unless you do one of the poles in the digital domain, and you do it effectively backwards in the time domain (which you can do if you have the file). Then the phase in the forward playback direction is reversed. Add this to the other filter and you get cancellation, zero phase shift, and a 2 pole 12dB per octave low pass. Magic. It is FILTFILT in Mathcad

Obviously MQA version is a bit more complex, as they are compensating for a bunch of different ADCs and DACs.

"we go all in for the time domain!" is inaccurate, because while they are trying to do better there than others have before, apodizing tended to damage the HF through slow extra roll-off in the pass band, so they can fix that too. I think it is time the time domain got some attention. Everyone seems obsessed with fixing the frequency response, which is fine, but all the f-ing about with EQ screws up the phase response as you say above, especially in the LF.

of course I talked about analog EQ just to give the most basic example of what is really just the amplitude and phase relation of any sine and how we can ultimately, kind of transfer the data from one to the other to make time look better while keeping the same absolute resolution. I wasn't inferring in any way that MQA's modus operandi was that limited. it's just that I've argued about EQ with @goodvibes and knew he would get that example.
for the all situation, we'd have to consider aliasing/where to achieve practical band limiting/sample rate, FR/bit depth, ringing/phase, as related components of what gives the final resolution(am I missing something?). so there is a lot to play with, and indeed the digital+analog filters open a lot of doors. a chance to really control the entire chain could be ideal for that, but would really limit post processing to make the album. that's why I don't see that part happening as a standard of anything.
everybody so far went with trying to stick as close as possible to Nyquist-Shannon's less than ten commandments ^_^. and by doing so, ringing outside the audible range was almost always the one to drink at least at low sample rates. MQA and a few other guys have other priorities and say that ringing no matter where is the worst thing that happened to music since Lil B. so which variable is going to pay instead? amplitude is a certainty with less bits. FR is highly probable, the real question being when does the roll off start and does it matter? aliasing? well that will depend on FR. it's nothing new, what changed is the anti-ringing obsession for a minority of people.

all that is splitting air about true fidelity of course. because it's all happening at high frequency and sample rate. on MQA the lowest is 96khz, I'll be the first one to say that I don't care and really don't find any of it really wrong. I'm very fine with gentle roll off in DACs when playing high sample rate music, or whatever more advanced filter they decided to implement. so I'm personally fine with pretty much anything including MQA in the sense that I do not believe it will change audio for me in any way. that's about the only support I can offer to MQA. "it will do nothing for me". every other aspects make me dislike MQA and wish it was already gone.

 

[jagwap]

of course I talked about analog EQ just to give the most basic example of what is really just the amplitude and phase relation of any sine and how we can ultimately, kind of transfer the data from one to the other to make time look better while keeping the same absolute resolution. I wasn't inferring in any way that MQA's modus operandi was that limited. it's just that I've argued about EQ with @goodvibes and knew he would get that example.
for the all situation, we'd have to consider aliasing/where to achieve practical band limiting/sample rate, FR/bit depth, ringing/phase, as related components of what gives the final resolution(am I missing something?). so there is a lot to play with, and indeed the digital+analog filters open a lot of doors. a chance to really control the entire chain could be ideal for that, but would really limit post processing to make the album. that's why I don't see that part happening as a standard of anything.
everybody so far went with trying to stick as close as possible to Nyquist-Shannon's less than ten commandments ^_^. and by doing so, ringing outside the audible range was almost always the one to drink at least at low sample rates. MQA and a few other guys have other priorities and say that ringing no matter where is the worst thing that happened to music since Lil B. so which variable is going to pay instead? amplitude is a certainty with less bits. FR is highly probable, the real question being when does the roll off start and does it matter? aliasing? well that will depend on FR. it's nothing new, what changed is the anti-ringing obsession for a minority of people.

all that is splitting air about true fidelity of course. because it's all happening at high frequency and sample rate. on MQA the lowest is 96khz, I'll be the first one to say that I don't care and really don't find any of it really wrong. I'm very fine with gentle roll off in DACs when playing high sample rate music, or whatever more advanced filter they decided to implement. so I'm personally fine with pretty much anything including MQA in the sense that I do not believe it will change audio for me in any way. that's about the only support I can offer to MQA. "it will do nothing for me". every other aspects make me dislike MQA and wish it was already gone.

Agreed. I think as everyone has said there is a limit how far back in the recording chain this can go for recording after multichannel digital arrived. However as these claimed effect are generally cumulative (I underline generally so the pedants notice the word and maybe for the first time don't latch on to any possible loop hole to get their teeth into) improving part of the chain is a probably positive thing. (It's like wading through treacle trying to make a simple point here).

MQA is scalable to the original sampling rate, so if it was originally 48kHz it stays that way. No upsampling as I understand it, and there are plenty of albums out there from before96kHz. Then apodizing would take quite a bit of the top end (sure 96kHz would be far more innocuous). I've seen 2-3dB at 20kHz, but these are usually the other companies doing apodizing, not so much Meridian. Thats's quite a loss in HF energy, and a sizeable phase shift. So if MQA can do apodizing without that, they will. I was outlining a way they could be done (allpass wouldn't work).

 

[jagwap]

If by "smaller points" you me "accuracy", I get that. I'll just continue to put it right. Or, you could take the time to get it right the first time, your call. Yes, I do get it, as to what MQA is doing, or at least attempting. Yes, you can pretty much perfectly correct for one or more well characterized filters. If they know the filter, have the data, sure they can do that. And I get that you can post process a file and pre compensate. That isn't the problem I'm having here.

The problems are (in no particular order):
1. Posts with blatantly wrong statements (which then get reversed in later posts). I quote the post, then the poster says he never said that (though it was a direct quote). Then we end up agreeing, and I'm labeled "pedantic".

You say reversed, I say clarified. However, are you trying to deny you can be pedantic?

2. MQA's claim of being able to reproduce the original performance

Agreed, that's nonsense, and that has been discussed infinitum.

3. MQA's claim that they can correct for "blurring", when the causes are multiple and unknown

They do not say ALL blurring. So what if they are correcting for some blurring? The blurring they can correct. Does that fit?

4. that any of this makes any audible difference.

Of course, we will see, one day, when hell freezes over I expect.

5. Yes, all-pass filtering can correct for phase. There have been examples for many, many years. Here's one: FM Stereo requires the use of a filter nearly identical to an anti-aliasing filter in a digital system. In FM Stereo that filter keeps audio out of the 19kHz pilot signal. Those filters were analog. FM modulation processing involves deliberate audio clipping, but the clipped waveform caused those filters to ring and overshooting the target 100% modulation limit, partially (largely) as a result of their poor phase response. One manufacturer remedied the situation by phase compensating those filters with complimentary phase all-pass filters. The resulting filter had far less phase distortion, and rang far less, and permitted higher modulation levels. That technique became quite popular. Early attempts at ADC anti-aliasing filters were also analog of course. During the early days of digital audio many blamed those filters for the harsh sound of digital, so several companies developed phase-compensated filters as an after-market upgrade. That compensation was also done with all-pass networks. The results were mixed, mostly because the harshness being complained about wasn't related directly to the phase response of the filters.

Just a few instances from history where analog filters were phase corrected using all-pass filters. There are more, of course. No, that kind of phase comp isn't perfect, and certainly if you were in the digital domain and had "time" you could produce a complimentary phase comp filter. That's not the argument here. Of course it can be done. And of course MQA is probably doing that. Of course, it doesn't matter because the results remain unproven as to their audibility.

And we are back to the beginning. Me trying to discuss a novel technique, and you trying to talk about the one you understand.

 

[bigshot]

I can't see how any of the stuff this is supposedly correcting is audible. It seems to me like neatly arranging pebbles on Mount Everest then standing back a couple hundred miles away and admiring how nice it looks.

 

[pinnahertz]

You say reversed, I say clarified. However, are you trying to deny you can be pedantic?

Tomato - tomahto. I've already embraced "pedantic", though I look at it as a demand for correctness and precision. This is a Science forum.

They do not say ALL blurring. So what if they are correcting for some blurring? The blurring they can correct. Does that fit?

Their marketing implies all blurring, no distinction at all. So they can correct some of the time response. So what? Marketing implies a day/night difference. It's not there. "Blurring" is a term MQA coined with strong negative connotation. All comparisons now are sighted. Marketing is focused on these things and making them powerful. Yet we lack an actual scientific and controlled test. Nobody knows, but then, if it's "day/night", why don't we have the proof?

Of course, we will see, one day, when hell freezes over I expect.

I doubt it would take that long, but the real questions everyone should ask is, "Why do we have the proof already?" It should have been forefront in marketing, not totally absent.

And we are back to the beginning. Me trying to discuss a novel technique, and you trying to talk about the one you understand.

Yup, except I do understand the novel technique, you don't seem to understand the one I've talked about, denying it works, yet it has already been used effectively.
Look, jagwap, none of this really matters. MQA isn't going to change audio as we know it until MQA is proven beyond question to be an audible improvement anyone can hear. It's already been around long enough for that to have happened, but it hasn't. It doesn't matter what technique they use, or how many filters they can correct for. It's either clearly audible, vaguely audible, or not audible. I've said before, several times, if it's a clearly audible improvement I'll get on board and support it. All we have is overwhelming doubt, zero proof, and marketing hype.

 

[bigshot]

Jagwap, asking certain types of people not to be pedantic is like asking water to be not so wet. Technically minded people revel in details and minutia, even if their focus on crossing every t and dotting every i ends up making their overall point as clear as mud. This isn't exclusive to sound science. The same sort of pedantry is just as common in audiophool circles. They're the ones talking about jitter and ultrasonic frequencies and levels of distortion so low it's hard to even measure it. But of course to them, that tiny fly speck is a huge deal. Pedantic people are useful at pointing out facts and quantifying details. But it's up to a different sort of person to translate their facts and figures into something practical and applicable.

The fact of the matter is that we're talking about listening to music in our homes with human ears. That sets limits to what degree of pedantry is realistically called for. I think it's pretty clear that the sorts of things MQA is pointing to as advantages fall into the category of irrelevant pedantic details. There's no reason to believe any of it is audible. In fact, I think the whole point of MQA is to raise vague, yet pedantic details that pedantic people can grab onto and argue about. Normal people hear this hue and cry and figure the truth lies halfway between two opposing arguments- they figure that maybe MQA actually does help a little. Maybe they'll buy a DAC that does MQA and convince themselves they've made the right choice. Ka-ching! Another sale!

That's the same strategy used by the jitter-bugs. They kept throwing up theories about brain scans detecting tiny fractions of time, but when the time came to actually test for audibility, the whole argument fell apart. There's a very good reason why MQA makes it hard to do simple A/B comparisons with plain vanilla redbook, or even high bitrate lossy. Keeping the pedants busy pointing at fly specks is good for business. It sells lots of fancy equipment to the puzzled people in the middle.

Maybe MQA does have some magic mojo that they aren't telling us about. It's perfectly possible. But until I can actually hear it with my own ears and test its effectiveness myself, I'm not going to invest a penny into it. That is how audiophools chase their wallet down rabbit holes, not the way a person applies principles of science to make their home stereo system sound better.

 

[pinnahertz]

This was just posted by Mark Waldrep of Real HD-Audio:
http://www.realhd-audio.com/?p=6046