Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Meridian Audio's new... sub-format called MQA. (Read 145494 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Meridian Audio's new... sub-format called MQA.

Reply #50
Stuart is proposing a theory. It could be proven or disproven in a rigorous psychoacoustic experiment, but to my knowledge no such experiment has been performed. His recent paper is not proof. It contains too many other variables.

In showing the impulse response of the filter, the paper is pretty much showing you what will happen at the leading edge of any wide bandwidth sharp transient. The waveform changes significantly, but all that horrible looking stuff is at 22kHz.

Cheers,
David.
Thanks. I do appreciate that there is no psychoacoustic evidence to support Stuart's theory, and of course that is arguably that.

However I think that there may be some benefit in trying to understand what his claim might even mean

It is my understanding that only energy in the transition band of a filter is distributed through the impulse response. If there isn't very much energy in that transition band then there won;t be very much to distribute. On that basis I find questionable the assumption that a shallow filter with a wide transition band (and short ringing)is preferable to a filter with a very narrow transition band and longer (but very low amplitude) ringing 

Returning though to my earlier hobby horse. It seems to me very likely that the entire concept of a sharp transient sonic event (occurring between the samples relative to 44.1khz  sampling) is very dubious. 
As I understand it air is not a very friendly medium to ultrasound. I have been struggling to find references but this paper here fig 2.2 suggests that air at 20 C and ordinary pressure and humidity
absorbs 100Khz sound at 2.2dB/m , 500kHz sound at 40dB/m and 1MHz sound at 160dB/m
 
http://www.ktu.lt/ultra/journal/pdf_50_1/5...ladisauskas.pdf
and also here which suggests 7-8dB/m for 200khz
http://www.ndt.net/article/ultragarsas/63-...-jakevicius.pdf

If this is correct then surely we can rule out any sonic event you are likely to hear at 5m* away from having any 500Khz content.
btw if you are at higher temperature or humidity it seems the absorption is much greater.
Unfortunately the article does not give figures for frequencies between 200khz to 500 Khz.

Ah no I have found this:
http://www.sengpielaudio.com/calculator-air.htm (does not work with google chrome) which enables one to calculate absorption for any frequency. (about 11.dB/m at 250Khz). 350 Khz seems a reasonable maximum at 5m

I am happy to be shot down in flames if I have misunderstood this, but it seems to me that this is sufficient to demonstrate that you simply can't have a dirac or sonic square wave- or even anything that would look a bit like say a sonic  40Khz square wave or it seems a rise time of much less than a microsecond if the event occurs 5m away. It seems to me that one ought to start, even when theorising about sound recording, with a possible sound event.   

It would be interesting to look at the effect of this air filter throughout the frequency range if we were to consider the inherent time smear in any sonic transient (assuming there were any point in considering inaudible frequencies)


* I'm assuming that we are interested in recording some sort of event one might watch, although I suppose some people might be interested in capturing the sound of a trumpet next to their ear. And it seems to me that the "our bodies have adapted to be able to localise the snapping of a twig" must refer to something a little way away or it would not be very useful.

Meridian Audio's new... sub-format called MQA.

Reply #51
Returning though to my earlier hobby horse. It seems to me very likely that the entire concept of a sharp transient sonic event (occurring between the samples relative to 44.1khz  sampling) is very dubious.
His theory doesn't require that. His sharp-ish filter takes a few samples to get 40dB down, and hundreds of samples to get 60dB down. If you Nyquist filter with something like a sinc filter, energy leaks out from any sharp rising edge a very long way. But only at 22kHz, and only if the rising edge has energy at 22kHz.

Hang on, you've said the same thing already, so you understand all this. But then you've gone on to do calculations that require an order of magnitude higher frequency because you've been distracted by the idea of needing sub-sample rise times for this to be significant. You don't.

Cheers,
David.


 

Meridian Audio's new... sub-format called MQA.

Reply #52
I am happy to be shot down in flames if I have misunderstood this, but it seems to me that this is sufficient to demonstrate that you simply can't have a dirac or sonic square wave- or even anything that would look a bit like say a sonic  40Khz square wave or it seems a rise time of much less than a microsecond if the event occurs 5m away. It seems to me that one ought to start, even when theorising about sound recording, with a possible sound event.
In 2009 I've done some microphone testing with electrical spark discharges. Since the spark isn't a perfect dirac the recorded impulse isn't just the IR of the microphone alone, but it did allow me to compare different microphones.
From AES paper 7065:
Quote
An acoustic test impulse which approximates the Dirac (t) distribution can be generated with a pistol shot or an electrical spark discharge. The former approach is poorly reproducible and provides single impulses only. Spark discharges can be reproduced periodically, and thus better captured. A spark discharge between two electrodes has the appearance of a heavily damped period close to the aperiodic boundary [2]. The short positive overpressure peak is followed by underpressure as the suddenly-expanded air flows back toward the center of ‘explosion.’ Consequently there is almost no acoustic wave propagation, since the real part of the acoustic radiation resistance approaches zero.
[/size]
These are the wav file and the waveform.
Spark Recording 24bit96kHz wav (20kB)


The spark was recorded at 24bit 96kHz from about 5 cm distance. I could redo it at 24/192 if anyone thinks that's useful.
The B&K4006 (now DPA4006) microphone is a "typical" high quality small diaphragm condenser microphone.
Note that one sample period in the graph is about 10 µs.

Meridian Audio's new... sub-format called MQA.

Reply #53
Returning though to my earlier hobby horse. It seems to me very likely that the entire concept of a sharp transient sonic event (occurring between the samples relative to 44.1khz  sampling) is very dubious.
His theory doesn't require that. His sharp-ish filter takes a few samples to get 40dB down, and hundreds of samples to get 60dB down. If you Nyquist filter with something like a sinc filter, energy leaks out from any sharp rising edge a very long way. But only at 22kHz, and only if the rising edge has energy at 22kHz.

Hang on, you've said the same thing already, so you understand all this. But then you've gone on to do calculations that require an order of magnitude higher frequency because you've been distracted by the idea of needing sub-sample rise times for this to be significant. You don't.

Cheers,
David.

Thanks, I did go a bit off on one. I think I was distracted by the concept (in the passage I quoted) of the inter sample dirac. His theory about the importance of ringing  may not require it, but he goes on about it when he taks about time resolution. I didn;t make that bit up.

I do get though that there will be pre-ringing if there is energy in the transition band of the rising edge, so we only need 22Khz energy to excite the ringign in 16/44

Staurt maintains that people can hear the effect of a conventional (normally linear phase, but he implies even minimum phase) filter at 48kHz.  But how much energy?

What confuses me is that he applies a significance test for aliasing by asking how big the product is relative to the noise level, but shouldn't we be asking the same question of the pre-ringing which might matter. I have never seen an analysis of a real world trasnient showing the amount of energy which is distributed at a given time relative to the noise floor- after all the MQA system depends in its coding on the point that there isn't actually any significant signal over 48Khz and the peak to floor level isn't very great in the octave below.

Meridian Audio's new... sub-format called MQA.

Reply #54
In 2009 I've done some microphone testing with electrical spark discharges. Since the spark isn't a perfect dirac the recorded impulse isn't just the IR of the microphone alone, but it did allow me to compare different microphones.
From AES paper 7065:
Quote
An acoustic test impulse which approximates the Dirac (t) distribution can be generated with a pistol shot or an electrical spark discharge. The former approach is poorly reproducible and provides single impulses only. Spark discharges can be reproduced periodically, and thus better captured. A spark discharge between two electrodes has the appearance of a heavily damped period close to the aperiodic boundary [2]. The short positive overpressure peak is followed by underpressure as the suddenly-expanded air flows back toward the center of ‘explosion.’ Consequently there is almost no acoustic wave propagation, since the real part of the acoustic radiation resistance approaches zero.
[/size]
These are the wav file and the waveform.


The spark was recorded at 24bit 96kHz from about 5 cm distance. I could redo it at 24/192 if anyone thinks that's useful.
The B&K4006 (now DPA4006) microphone is a "typical" high quality small diaphragm condenser microphone.
Note that one sample period in the graph is about 10 µs.
Wow thanks, that's the sort of thing I was looking for, although I'm not sure what the limiting factor is.

Would it be a fair summary that the impulse response is dominated by that of the microphone.? There seems to be maybe 70us of pre ringing visible. What is the frequency response of the mic?  Is is safe to assume that its stopband is much lower than 48Khz? I was guessing that the pre ringign was at about 30Khz. but would be grateful for fonrimation of how one measures this.

does the A/D use a linear phase filter?

I'm intrigued by what it would look like at 24/192 unless you think it's obvious that it wouldn't make a difference

Meridian Audio's new... sub-format called MQA.

Reply #55
DPA4006A

Frequency range, ± 2 dB: 10 Hz to 20 kHz

4006


Meridian Audio's new... sub-format called MQA.

Reply #56
DPA4006A

Frequency range, ± 2 dB: 10 Hz to 20 kHz

4006

Thanks- I did look at those figures, but I was wondering at what point it reaches what one might call a stop band. If for the sake of argument it reached -80 dB at 30 Khz, then the impulse response of a conventional A/D filter for 24/96 would be irrelevant (as would the air filtering effect too). If however the mic has a very slow roll off above 20Khz then the impulse response of the A/d filter at 24/96 might have an impact.

In order to see the effect that the A/D filter has at 24/96 it would be useful to see a recording of a mic which could definitely capture frequencies over 40Khz (ideally 100 khz). ideally with flattish response but i imagine that might be tricky

Assuming that the DPA mic was flattish to 22.05, It would be interesting to compare a 44/1 recording where one really would be looking at the response of the  filter to the spark

Meridian Audio's new... sub-format called MQA.

Reply #57
Assuming that the DPA mic was flattish to 22.05, It would be interesting to compare a 44/1 recording where one really would be looking at the response of the filter to the spark

Most (studio) recordings are made at 96 kHz (or higher) sampling rate and then downconverted to 44.1, so that's what I did here. The original spark signal was sample rate converted to 44.1kHz and back to 96kHz (with iZotope RX SRC). I've assembled a short stereo file with the 24/96 original in the left channel and the 44.1 version in the right channel, followed by the same but with L/R swapped. It allows to listen for differences (looped playback can be handy).
Hope this helps.

Spark sample 24/96 versus 24/44.1


Meridian Audio's new... sub-format called MQA.

Reply #58
Well, MQA has some DXD competition, though I'm not sure where it falls on the Meridian "Quality vs Convenience" graph.

From the Probates website


and



and of course:

Quote
For me, it is clear that there is an undeniable relationship between resolution and quality. Each format seems to have has its own expression and conversely, its own limitations. To reproduce classical music, for example, I have always preferred the quality of DSD.

Yet, it was not until I experienced DXD that I was no longer able to hear the signal as a digital reproduction. In contrast, DXD is calm and warm, with a deep, resonant and well defined stereo perspective. So much so that it evokes memories of the heady days of analog.

Peter Scheelke


cheers,

AJ
Loudspeaker manufacturer

Meridian Audio's new... sub-format called MQA.

Reply #59
I'm sure Nyquist will move up one octave again as soon as the technology allows it.


Meridian Audio's new... sub-format called MQA.

Reply #61
From the Probates website
. Then those amplitudes could be anything.


People have identified this possible problem in the past. I've only actually heard it with sound blaster cards and test signals. Even so, you can solve it during playback by using a reconstruction filter that cuts at about 20kHz, not 22kHz. You can solve it during recording by using an anti-alias filter that cuts "early" like that too. Looking at the spectrum of CDs, such a thing is quite common. Not all, but many CDs have almost no spectral content approaching Nyquist, which implies there was nothing left above Nyquist to alias.

Cheers,
David.

Meridian Audio's new... sub-format called MQA.

Reply #62
a most interesting firsthand account of Meridian's award-winning paper presentation at the last AES conference.  Apparently not everyone in the audience was convinced...

http://www.avsforum.com/forum/286-latest-i...ml#post29832714

Quote
I have read the paper and was also present when it was presented. In the question and answer period following the presentation, Analog Devices' Bob Adams (designer of the first successful IC asynchronous sample rate converter for AD) pointed out that the shapes of the filters used in the study were pathologically selective compared to normal commercial practice, having a much longer impulse response (by about a factor of 4x) than filters commonly used. So the take-away from this paper, at least for me, is that one can design an audible linear-phase digital filter whose transition region is above 20 kHz if you make the impulse response long enough.

As for the use of RFDF [sic] dither in the 16-bit tests, this just seems bizarre to me because it is so contrary to good engineering practice.

From the paper:

"The frequencies of the transition bands were 23500-24000 Hz and 21591-22050 Hz, corresponding to the
standard sample rates of 48 kHz and 44.1 kHz respectively."

In normal commercial practice, the transition region is allowed to start at 20 kHz.


(Actually maybe this belongs in a different thread, but that thread is closed.  New thread, mods?)

Meridian Audio's new... sub-format called MQA.

Reply #63
(Actually maybe this belongs in a different thread, but that thread is closed.  New thread, mods?)
I think it was well established in the closed thread that the authors of this paper (intentionally?) used inappropriate practices to push Meridian's agenda to sell expensive gear and their new format. Is there anything new to say to this?
It's only audiophile if it's inconvenient.

Meridian Audio's new... sub-format called MQA.

Reply #64
I'm slightly confused: is the great contention of Meridian that we can hear the ringing introduced by conventional A/D and D/A conversion? This thread hasn't quite clarified things for me, and I already bought two of their papers (and deeply regret doing so ). Many thanks for any advice.

Meridian Audio's new... sub-format called MQA.

Reply #65
(Actually maybe this belongs in a different thread, but that thread is closed.  New thread, mods?)
I think it was well established in the closed thread that the authors of this paper (intentionally?) used inappropriate practices to push Meridian's agenda to sell expensive gear and their new format. Is there anything new to say to this?



Given the sheer amount of noise on that thread compared to signal, perhaps adding your note above to the end of it would help the newcomer (or better, pin it to the start)?

Meridian Audio's new... sub-format called MQA.

Reply #66
Given the sheer amount of noise on that thread compared to signal, perhaps adding your note above to the end of it would help the newcomer (or better, pin it to the start)?
I agree, better yet would be to bisect the thread into signal and noise, which will be a humongous task for the holidays.
It's only audiophile if it's inconvenient.

Meridian Audio's new... sub-format called MQA.

Reply #67
I'm slightly confused: is the great contention of Meridian that we can hear the ringing introduced by conventional A/D and D/A conversion? This thread hasn't quite clarified things for me, and I already bought two of their papers (and deeply regret doing so ). Many thanks for any advice.


I also bought both papers and FWIW and based them and on other documents and ads of theirs, Meridian and Dolby are definitely of the belief that many people can hear the ringing introduced by conventional A/D and D/A conversion, and that tihs is a major reason for consumers being dissatisfied with the sound quality of their audio systems of all kinds except of course those blessed with Meridian's patented technology (which Dolby has licensed).

IME Meridian has always played this game, but with Dolby it is a clear case of Sic Transit Gloria.

Meridian Audio's new... sub-format called MQA.

Reply #68
I haven't seen any link reported between MQA and Dolby - have you read something I haven't?

Obviously Meridian licensed MLP to Dolby who sell it as Dolby True HD, but MQA is a different thing.

Cheers,
David.


Meridian Audio's new... sub-format called MQA.

Reply #70
He's referring to Advanced 96k Upsampling.


Thanks.

I was answering the question that was asked:

Is it the "...great contention of Meridian that we can hear the ringing introduced by conventional A/D and D/A conversion?"

And it is exactly that which is the contention of both Meridian and Dolby Labs

More specifically:

http://www.dolby.com/us/en/technologies/do...white-paper.pdf

They are describing "...applying an advanced Apodizing that masks pre-ringing..."

Given Dolby's apparent influence in the AES, this makes that  best paper award some kind of a slam dunk.  IMO it is just a case of one dirty hand buffing the dirt on another. :-(

Meridian Audio's new... sub-format called MQA.

Reply #71
Ah, I see - thank you.

Meridian Audio's new... sub-format called MQA.

Reply #72
Given the sheer amount of noise on that thread compared to signal, perhaps adding your note above to the end of it would help the newcomer (or better, pin it to the start)?
I agree, better yet would be to bisect the thread into signal and noise, which will be a humongous task for the holidays.


I would call that a Christmas miracle.

Meridian Audio's new... sub-format called MQA.

Reply #73
Just what I thought. Revolutionary developments in psychoacoustics my arse.

The funny thing is that, as I understand it, for 44.1ksps you can make a faint case for the potential audibility of said ringing: if you carefully pick your content, and then get some kids with exceptional HF hearing, it's not completely out of the realms of possibility that the ringing might be very barely possibly audible (recalling a post by JJ on SkepticForum).

Of course - and I'm preaching to the choir here, I realise - by the time you get to 88.2ksps ringing is unconditionally inaudible to whatever freak combination of hypothetical listeners and material you could devise. So Meridian's format helps with...what, exactly?

But seriously, is that it? I am bemused as to how people can take this seriously - does nobody in the audience laugh when they try to pull this crap? I'm very, very far from an expert on the subject, but as far as I can tell this is reasonably clear-cut BS. It's not even new BS: issue 16 of "The Audio Critic" (http://www.theaudiocritic.com/back_issues/The_Audio_Critic_16_r.pdf) records Wadia trying to shovel similar shit in 1991 (only their ingenious solution was simply rolling off earlier enough to screw with the audioband, but very slowly...).

Meridian Audio's new... sub-format called MQA.

Reply #74
if you carefully pick your content, and then get some kids with exceptional HF hearing, it's not completely out of the realms of possibility that the ringing might be very barely possibly audible


But how many of those kids would care?

Quote
But seriously, is that it? I am bemused as to how people can take this seriously - does nobody in the audience laugh when they try to pull this crap?


Well, if they take audiophile ethernet cables seriously...