Jump to content
IGNORED

MQA is Vaporware


Recommended Posts

3 hours ago, Fokus said:

This is the process for a 384k master:

 

-downsample to 192k with leaky MQA filter

-downsample to 96k with leaky MQA filter

-fold into 48k space using Quadrature Mirror Filter pair 1 (*)

=============================================

-unfold 48k to 96k using Quadrature Mirror Filter pair 2

-upsample to 192k using leaky filter

-upsample to 384k using leaky filter

-light the blue LED

The decoder may or may not upsample to the original rate of the master. It depends on the capabilities of the DAC chip and other parts. The blue light comes on regardless, of course.

Link to comment
3 hours ago, Fokus said:

(* As an aside: the output of QMF1 is what people without MQA decoding have to listen to.

QMF1 has to be optimised to allow a lossless split-join in the origami folding step. This is an

extremely limiting constraint. This means that QMF1 cannot likely be optimised, at the same time,

for optimal sound quality for non-MQA listening. This is mathematics.)

 

Hi,

What is the type and order of the Quadrature Mirror Filter ?

Thanks and regards,

Shadders.

Link to comment
2 minutes ago, mansr said:

The decoder may or may not upsample to the original rate of the master. It depends on the capabilities of the DAC chip and other parts. The blue light comes on regardless, of course.

 

Assuming sufficient capabilities, of course.

 

Link to comment
46 minutes ago, Shadders said:

What is the type and order of the Quadrature Mirror Filter ?

 

It is not a filter type, but rather a relationship between two filters, so that when combined they satisfy (free of error) a specific criterion. In this case two QMF pairs have to be used to split a 96k signal into two bands that each can be sampled at 48k, so that after recombination the original 96k signal is obtained again.

 

https://en.wikipedia.org/wiki/Quadrature_mirror_filter

 

It is my suspicion that the two origami folding QMFs do not reach their stop band exactly at 24kHz(*). In other words, both would alias badly when resampled to 48k. A direct consequence of this is that the baseband signal of a hi-res MQA file must be infected with aliasing. If and when the true nature of the encoding QMFs is revealed then this would constitute mathematical proof that the undecoded MQA signal is objectively inferior to CD-rate, contrary to MQA's claims.

 

 

(* Why? Well ... this universe was not created for the sake of audiophiles, otherwise playing LPs properly would not be an insolvable problem. It would be one hell of a coincidence if a set of filters would at the same time satisfy an audiophile's needs and the double QMF criterion.)

 

Link to comment
1 minute ago, Fokus said:

 

It is not a filter type, but rather a relationship between two filters, so that when combined they satisfy (free of error) a specific criterion. In this case two QMF pairs have to be used to split a 96k signal into two bands that each can be sampled at 48k, and which after recombination yield the original 96k signal.

 

https://en.wikipedia.org/wiki/Quadrature_mirror_filter

Hi Fokus,

Thanks. I do believe that there must be an order to the filter - to be able to split the two bands.

In addition, every filter has a type - Bessel, Butterworth, Chebychev etc. which provides a type of passband amplitude and roll off, with or without in band ripple etc.

Thanks and regards,

Shadders.

Link to comment
12 minutes ago, Shadders said:

I do believe that there must be an order to the filter -

 

Of course. But with the present public knowledge we cannot know that order. It would be immensely interesting to have access to an MQA encoder. My first test would consist of pink noise steeply filtered below, and then above, exactly 24kHz, in each case studying the undecoded output signal.

 

Link to comment
Just now, Fokus said:

 

Of course. But with the present public knowledge we cannot know that order. It would be immensely interesting to have access to an MQA encoder. My first test would consist of pink noise steeply filtered below, and then above, exactly 24kHz, in each case studying the undecoded output signal.

 

Hi Fokus,

Thanks.

What is interesting is that correcting dispersion (temporal blur) is the crux of MQA, yet they use extremely high order filters (QMF) which will cause significant dispersion, and this is not corrected.

Unless someone has evidence to the contrary ?

Regards,

Shadders.

Link to comment
4 hours ago, Fokus said:

This is the process for a 384k master:

 

-downsample to 192k with leaky MQA filter

-downsample to 96k with leaky MQA filter

-fold into 48k space using Quadrature Mirror Filter pair 1 (*)

 

Why does MQA encoding downsample twice, instead of downsampling directly from 384K to 96K in one step?
 

 

4 hours ago, Fokus said:

=============================================

-unfold 48k to 96k using Quadrature Mirror Filter pair 2

-upsample to 192k using leaky filter

-upsample to 384k using leaky filter

-light the blue LED

 

Same question for upsampling. Upsampling from 96K to 384K is much more efficient in one step, instead of wasting cpu cycles doing this in 2 steps.

I speak from experience, as I once rewrote our own DSP steps and reached an 8x speed gain without any negative sound impact. What MQA is doing seems like a waste of resources including power and cpu cycles.

Designer of the 432 EVO music server and Linux specialist

Discoverer of the independent open source sox based mqa playback method with optional one cycle postringing.

Link to comment
11 minutes ago, Shadders said:

 yet they use extremely high order filters (QMF) which will cause significant dispersion, and this is not corrected.

 

1) there is nothing that forces the QMFs to be extremely steep, so your assertion is based on no evidence.

 

2) due to the QMF property the recombination of the two bands is totally lossless. This is proven: passing a 96k original through MQA gives a pretty decent copy of that original.

 

 

Link to comment
Just now, mansr said:

Where have they said that dispersion is what they mean by temporal blur?

Hi mansr,

They have not. Temporal blur is dispersion. Temporal blur is a marketing term, which describes loosely what is happening, whereas dispersion is an engineering term, which many people are not aware of.

In addition to this - they have stated filters cause temporal blur, and besides the amplitude and phase effects, dispersion is the only engineering understanding that filters cause, which will delay different frequencies a different amount.

If you know that temporal blur is not dispersion, then what in engineering terms is temporal blur - as none of the many communications texts i have describe such a phenomena.

Thanks and regards,

Shadders.

Link to comment
2 minutes ago, Fokus said:

 

1) there is nothing that forces the QMFs to be extremely steep, so your assertion is based on no evidence.

 

2) due to the QMF property the recombination of the two bands is totally lossless. This is proven: passing a 96k original through MQA gives a pretty decent copy of that original.

 

Hi Fokus,

Correct - it is my assertion that it is steep to remove aliasing ?

As per your point 2., if the recombination is totally lossless, then dispersion is not an issue, hence MQA is not required.

I assume a "pretty decent copy" is a lossy reconstruction.

Regards,

Shadders.

Link to comment
6 minutes ago, FredericV said:

Why does MQA encoding downsample twice, instead of downsampling directly from 384K to 96K in one step?

 

That was just for the explanation. Maybe MQA uses sequential steps, maybe all in one go. It probably depends on whether they have developed their filter herd for /2 steps only, of for all possible source and target rate combinations. But this does not matter. We are discussing concepts, not implementations.

Link to comment
Just now, Shadders said:

, if the recombination is totally lossless, then dispersion is not an issue, hence MQA is not required.

 

No. MQA's fight with blur is at the top of the signal band, in our 96k example at 48kHz.

If the QMFs injected their own blur, for whatever reason, then this would affect the area

of 24kHz. But we know from spectral analysis that after unfolding the 24kHz area does not

seem to be damaged: 'pretty good'.

Link to comment
1 minute ago, Fokus said:

 

No. MQA's fight with blur is at the top of the signal band, in our 96k example at 48kHz.

If the QMFs injected their own blur, for whatever reason, then this would affect the area

of 24kHz. But we know from spectral analysis that after unfolding the 24kHz area does not

seem to be damaged: 'pretty good'.

Hi Fokus,

Dispersion appears across the entire audio band - it is just to what degree it occurs at which frequency.

Spectral analysis will not prove whether dispersion affects a specific band unless you have the before and after results - that is, the audio file before dispersion, and the audio file after dispersion (the damaging filtering as per MQA claims).

Regards,

Shadders.

Link to comment
8 hours ago, Lee Scoggins said:

 

Oh shut up.  We are plenty experienced to not be at the whim of a first impression.  I've heard valid MQA demos on several occasions.  And you don't even understand how process standards like authentication differ from DRM. 

 

The real problem here are armchair idiots like yourself who can't respect a differing opinion.  Go back to school and take a Debating 101 course and learn how to present some points backed up by some evidence before slamming the journalists who are working hard to explore audio.

 

7 minutes ago, rickca said:

The industry insider credentials and connections you reference in your Part-Time Audiophile article are irrelevant to debating the merits of MQA.  

 

Don't be so hard on poor Lee. Can't you see he's working hard here?  :)

 

 

Link to comment
6 minutes ago, mansr said:

"Temporal blur" doesn't need to be anything at all beyond a made-up marketing term.

Hi mansr,

OK - there can be two approaches - temporal blur is made up by MQA and does not exist, and everyone is being scammed that it does exist.

Alternatively, temporal blur is dispersion, and engineering analysis shows that this can never be corrected.

Probably better to cover both aspects to ensure that false claims by MQA can be countered.

Regards,

Shadders.

Link to comment
20 minutes ago, mansr said:

"Temporal blur" doesn't need to be anything at all beyond a made-up marketing term.

 

Come on.

 

Ever since Craven's first papers on the subject, back in 2006 or so, we surely all know they are on a crusade against the ringing seen in the impulse response of steep low-pass digital filters, especially linear phase filters.

 

The audiophile face of MQA is all about avoiding visible ringing by not using any steep filters in the signal path. Leaky filters are selected, allegedly on a per-case base, so that any aliasing falling below 20kHz is of the same order as the programme's innate noise.

 

 

This of course ignores the inconvenient fact that the aliasing can only be low when the signal content at the filter's transition is low too, in which case its ringing would be low as well.

 

When it comes to blur, MQA is a solution without a problem. When it comes to streaming hi-res, MQA is a solution without a problem.

 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...