advertisement


MQA pt II

UliBru said:
Similar topic: aren't there several DACs out there where it is possible to switch the characteristic of the brickwall filter?
Click to expand...

The Sony CDP-X3000 had four switchable filters in 1995. Probably the first to do so.

DimitryZ said:
embracing Dolby Atmos, which is perceptually lossy and often plain bad when downmixed to two channels.
Click to expand...

Years ago in one of the first mega-MQA threads I wrote that we should not worry about MQA, but about what the industry would foist on us after MQA.

DimitryZ said:
For a truly elegant target tracking system, ...
Click to expand...

I don't do weaponry.
 
UliBru said:
A given all pass filter convolved with its time-reversed all-pass filter results in a Dirac pulse. Thus a removed pre-ringing can be recovered.

Similar topic: aren't there several DACs out there where it is possible to switch the characteristic of the brickwall filter?
Click to expand...

Yes. Having had a chance to delve into this I think it quite likely that we can find a filter that corrects the dispersion. The snag would be how to employ it, though. cf below...

Yes, many DACs have allowed us to switch and choose our reconstruction filter. But IIUC MQA *require* you to use *their* filter not any alternative because their filter is part of the way to 'ensure Master Quality'.

Since altering the data in the audio file or from CD loses MQA 'authentication' and thus stops the 'recovery' of the hidden HF you can't easily filter the input to the DAC. Thus to correct you may need to buy a good ADC, capture the output, and correct *that*! Not very practical for most people. But of interest from my POV at some future instant, so I may sometime give it a go.

BTW my tests seem to show that the MQA decoding does 'freewheel' for a short time. I added a non-MQA 'tail' to my 2L MQA test files for other calibration and alignment purposes. And it looks like the impulses I included get given the full MQA added-dispersion treatment. Whereas those in plain non-MQA files don't.
 
DimitryZ said:
All modern DACs have many variable filters that can be switched. Contrary to popular belief, many are audible, but not dramatically so.
Click to expand...

It would not surprise me if some find them audible when others don't. One consequence is often a change in the peak/mean ratio of waveforms. That then gets fed to non-linear loudspeakers...

Another consequence is that the 'short' filters tend to generate anharmonic distortion products that aren't in the source material. Also potentially audible in some cases.

So it may be a matter of "Would you like salt on that?"
 
People may find this of interest:
http://jcgl.orpheusweb.co.uk/temp/OnImpulse.png

I started with some 2L MQACD files. I added a non-MQA 'tail' to each of them. This starts with a burst of impulses (single non-zero values surrounded by zeros). The result remains 44k1/16 with embedded MQA encoding for the bulk, but not the attached tail. I then used sox to generate an 88k2/24 version. This loses the MQA authorisation, and spreads the impulses into standard sinc approximations.

The image shows the results as captured with an ADC from the output of the Meridian. It looks like the impulses on the MQA file get the full MQA dispersion. But the ones on the 88k2 file don't get that. You mainly see the sox-generated sinc shape with a possible hint of some added dispersion. But note that I can't clock-lock the ADC because the Explorer doesn't have a clock output! So the slight asymmetry of the sinc may be due to a slight phase offset between the clocks.

I mainly added the impulses to help me time align comparisons, but this result was quite interesting.
 
It looks in case of MQA decoding an all-pass filter gets applied and it shifts the phase at high frequencies. Thus pre-ringing is removed by
linear phase * excess phase = minimum phase.

Repeating the tests with modified bursts (including pre- or postringing) as tail would be quite interesting.
 
Yes. Initially I assumed that the MQA decoder would probably 'disengage' so quickly that it might not show on the impulse functions that come first in the 'tail' I added. When playing, the pretty coloured LED goes off pretty promptly, but it isn't obvious exactly when wrt the tail-start. I do plan to experiment with this. But thought it useful to point out as it means others who have an MQA DAC can also experiment. mansr said that the authetication bursts are spaced a given number of samples apart. So it may be that we can get at least that long for added test patterns to be processed. More research required. :)

It is useful for confirming what might be needed to un-do the dispersion, but this seems impossible to do for MQA files as per above. May be useful for comparison analysis though. The dispersion messes up doing a simple sample-diff, and dispersion correction might aid making that possible.

Given this I'll probably finish what I have for the current webpage I'm writing and leave the temporal side for a new page. Hmmm... does that mean this one is the 'spiritual' page, I wonder... 8-]
 
Jim Audiomisc said:
It would not surprise me if some find them audible when others don't. One consequence is often a change in the peak/mean ratio of waveforms. That then gets fed to non-linear loudspeakers...

Another consequence is that the 'short' filters tend to generate anharmonic distortion products that aren't in the source material. Also potentially audible in some cases.

So it may be a matter of "Would you like salt on that?"
Click to expand...
To the extent people use these filter choices at all, it typically involves listening and choosing one, then leaving it alone.

Seems like a normal part of the hobby.
 
DimitryZ said:
To the extent people use these filter choices at all, it typically involves listening and choosing one, then leaving it alone.

Seems like a normal part of the hobby.
Click to expand...

It has been. But MQA may mean you can't do it for MQA material because you can't alter the MQA files without risking losing the MQA being decoded. (1) Do any MQA DACs allow you to alter the replay filter when replaying MQA? The Explorer 2 doesn't seem to show any sign of this.

(1) Unless MQA release or allow someone under payment to do it. Do they?
 
Jim Audiomisc said:
It has been. But MQA may mean you can't do it for MQA material because you can't alter the MQA files without risking losing the MQA being decoded. (1) Do any MQA DACs allow you to alter the replay filter when replaying MQA? The Explorer 2 doesn't seem to show any sign of this.

(1) Unless MQA release or allow someone under payment to do it. Do they?
Click to expand...
Actually, it's absolutely under your control, if you use software unfolding - available with Tidal, Roon and others - and use an MQA-unaware DAC.

MQA's filter and upsampling instructions are ignored and you can use whatever filter that DAC allows.

From information on other forums, this appears to be a popular way to use MQA without paying for an MQA DAC.
 
Jim Audiomisc said:
Interesting. Can you point me at their documentation for that?
Click to expand...
This works automatically on IOS and Android platforms for my two MQA applications - Tidal and Nugs. In Windows, there are three settings for the MQA decoding in Tidal, one of which performs the software unfold, instead of leaving that to the DAC. This is described in Tidal's documentation. Roon and Audiorvana would have similar or more advanced settings, which should be described in their documentation.

This has been the case for 3-4 years for Tidal and 1-2 years for Roon.

As for comparison of different MQA decoding modes, this has been explored at length by Archimago, with an ADC setup, several years ago.

http://archimago.blogspot.com/2017/01/comparison-tidal-mqa-music-high.html?m=1

http://archimago.blogspot.com/2017/02/comparison-hardware-decoded-mqa-using.html?m=1

If you are going to take advantage of this long-available decoding mode on the Unix OS, Roon would seem to be the most direct path. Roon will also allow for DSP room correction to be applied within MQA file, while remaining fully MQA compliant - which some people find to be a worthwhile feature.

@mansr software decoder should perform the same function, giving you an unfolded file that you can play on any DAC with any filter choice. I think he provides documentation.
 
Last edited:
I'll have a look at the links. My main question wrt the Roon/Tidal software decoding is if it outputs a digital stream that has the artifacts and dispersion so that any later choice of DAC filter simply convolves on top of that. i.e. you can't get the output without the dispersion or artifacts which show up with, say, just using the Explorer 2.

The implication of MQA claiming that their output is "audibly the same as the 'master'" implies you get out a stream with the dispersion, etc. i.e. if they thought the dispersion has no audible effect, it seems odd that they add it!

It also makes me wonder if perhaps DAC makers can include a 'remove MQA dispersion' mode?
 
Jim Audiomisc said:
I'll have a look at the links. My main question wrt the Roon/Tidal software decoding is if it outputs a digital stream that has the artifacts and dispersion so that any later choice of DAC filter simply convolves on top of that. i.e. you can't get the output without the dispersion or artifacts which show up with, say, just using the Explorer 2.

The implication of MQA claiming that their output is "audibly the same as the 'master'" implies you get out a stream with the dispersion, etc. i.e. if they thought the dispersion has no audible effect, it seems odd that they add it!

It also makes me wonder if perhaps DAC makers can include a 'remove MQA dispersion' mode?
Click to expand...
I had assumed the MQA file was subject to the “deblurring”/ dispersion effect when the 88/96khz Fs version of the track was created prior to being folded up.
 
I had a quick look at the archmago pages, but although interesting they didn't seem to deal with what I was curious about.
I'm not clear on if some of the deblurring is on output, or all on input or what. Just that it keeps cropping up in the results.
From what MQA claim they do a source-end 'deblurr' which varies with the ADC they are told was used. But In the 2L examples I've looked at the shape is essentially the same in each case. So I'd probably need to do some non-2L examples to compare. I wonder if there are some highly clipped and level compressed 'MQA CDs' out there as the results might be really weird. 8-]

Haven't seen any sign that the roon/tidal software lets the user change it to being time symmetric or undo what was done at encoding, etc. However I gave up on the archmago and got on with the page I'm writing.

FWIW I've decided to split what I'm doing at present into two pages. That keeps down the page sizes and means I can put the first page (part) up sometime soon and focus on the second that deals with other aspects. So all being well the 'first half' page will appear today or tomorrow with some more results, etc. Give people something to mull over whist I look at other aspects inc the magic of time domain effects and corrections. But if software like the roon/tidal don't let the user do this then we end up needing some other route. I can go via an ADC for tests. Fine for experiment, but not something most people would do for their music collection. :) Hmmm... maybe we can work out an analogue filter that undoes the dispersion. Or find a speaker that does it. :)
 
Jim Audiomisc said:
I'll have a look at the links. My main question wrt the Roon/Tidal software decoding is if it outputs a digital stream that has the artifacts and dispersion so that any later choice of DAC filter simply convolves on top of that. i.e. you can't get the output without the dispersion or artifacts which show up with, say, just using the Explorer 2.

The implication of MQA claiming that their output is "audibly the same as the 'master'" implies you get out a stream with the dispersion, etc. i.e. if they thought the dispersion has no audible effect, it seems odd that they add it!

It also makes me wonder if perhaps DAC makers can include a 'remove MQA dispersion' mode?
Click to expand...
A number of them do, though typically at a higher price point than the one you are using.

These have a firmware toggle switch to enable/disable MQA.

Since in 99% of actual usage is streaming and the unfold can be performed in the streaming software, the choice of DAC filter and of the DAC itself ends up with the user. So if you don't like the "salt" of MQA filters, it's easy to avoid them.

My new DAC is Musician Pegasus, which is R2R and entirely MQA unaware. Software unfold in Tidal/IOS appears to work fine based on the frequency reported by the DAC and subjective sound quality is also fine - I don't think I can score well in a DBT against another, MQA enabled DAC. This suggests that MQA rendering step is relatively unimportant.

If your end goal in all of this is to come up with a full deMQAfication algorithm, that sounds like a tall order.
 
Last edited:
You must log in or register to reply here.


advertisement


Back
Top