Article: MQA: A Review of controversies, concerns, and cautions

[Archimago]

31 minutes ago, Fokus said:

 

Not linear phase. Presumably an additional all-pass network that reduces the non-linear phase component in its response somewhat. I'll see if I can find that old group delay plot.

 

 

Edit: got it.

 

Look in this document at Figure 3 and the text below it.

 

http://www.audio-focus.com/Townsend/pdf/Why_supertweeters.pdf

 

 

(Apparently with thanks to Peter Baxandall...)

 

Thanks Fokus! Nice look back at the performance of these devices...

[Fokus]

23 minutes ago, Don Hills said:

 

Just to complicate things, there was also an aftermarket filter set for the Sonys that was quite popular. The brand name escapes me at the moment...

 

Apogee Electronics.

 

 

And then there is also the Nakamichi version of the PCM-F1.

 

 

[Fokus]

26 minutes ago, Archimago said:

 

Thanks Fokus! Nice look back at the performance of these devices...

 

You may like this comparison of early PCM adapters:

 

 

http://www.theimann.com/Analog/A77/A77vsPCM/

[miguelito]

6 hours ago, Fokus said:

 

Where? By deeply understanding, since 2014, what MQA is about.

 

There are two approaches to deblurring in this story.

 

1) Start at 192k or higher, and downsample to 96k with a narrow-impulse, leaky filter that injects aliasing (but hopefully not too much) and that (also hopefully) puts a null at the original ADC's filter transition frequency. It is totally clear that this is exactly what MQA are doing today for such originals.

 

2) Start at 96k or lower. In this case there is no downsampling possible, so they will have to revert to eq-style solutions. This can be additional narrow-impulse low-pass filtering for a 96k original (i.e. apodising in Meridian-speak). For 44.1k and 48k not even this is possible, so all that could be done is some phase/group-delay equalisation, perhaps also level equalisation, at the treble end.

What you're saying makes sense. There's clearly a downsampling to 96 or 88 of higher sample rate source files, which likely happens with a filter of similarly slow rolloff, then a folding into the 48 (or 44) package. 

 

Since we are talking about this aspect, I will also mention that it seems that the "target rate" embedded in MQA data is solely used to determine a particular upsampling filter - in all likelihood the "mirror" of the downsampling filter used as you say - rather than truly having the DAC run at that sample rate. 

 

Specifically, I don't think you can make an ESS DAC run it's output at any other than the max sampling rate (352/384). So when in those cases you're saying the target rate is 192, the only thing this means is a particular choice of upsampling filter assigned to that 192 target rate. It doesn't mean the DAC is running at 192.

 

One more thing: It does make sense to use an upsampling filter from 88/96 that is a "mirror" of the downsampling filter to 88/96 used at encoding time. This is really unique to MQA. But why downsample/upsample in the first place? I much rather have the true source!!!

[FredericV]

8 hours ago, Em2016 said:

 

Like this?

 

"This “slow roll-off” filter reduces the time smear by a factor of ~20x compared to conventional digital filters. The net result is a much more musically natural sound, as the ear-brain is very sensitive to time-related distortions. This filter provides an outstanding compromise between frequency response and transient response, and for ten years was the mainstay of Ayre’s digital audio filters."

 

https://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf

 

 

 

These filters can be implemented using SoX.

 

They also leak a lot, here's the spectrum from a 16 bit MQA file, using MQA's filter vs SoX minimum phase and one cycle of postringing:
 

 

All the content above 20 Khz = fake.

[Archimago]

10 hours ago, Em2016 said:

 

Like this?

 

"This “slow roll-off” filter reduces the time smear by a factor of ~20x compared to conventional digital filters. The net result is a much more musically natural sound, as the ear-brain is very sensitive to time-related distortions. This filter provides an outstanding compromise between frequency response and transient response, and for ten years was the mainstay of Ayre’s digital audio filters."

 

https://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf

 

 

 

Yeah, the Ayre "Listen" minimum phase and slow roll-off filter was not something I liked with my PonoPlayer. I think if Charles Hansen were alive today we would be having a good discussion about this technically today...

[firedog]

10 hours ago, Archimago said:

 

Looks like the same arguments he had a month ago before the previous thread on MQA got taken down after 50 pages or so at Steve Hoffman's. Of course, back then @The Computer Audiophile wasn't on his "hit list" as I recall .

 

@Lee Scoggins, I see you're not banned from here, so feel free to let us know the experiment results with those A/B files... Hmmm, did you ever publish a "part 2" to your MQA report? I thought you were going to do some research beyond being impressed by the "thousands of albums are coming and it sounds good to me"...

From what he writes there, looks like he won't be back here. Basically he made (mild) ad hominem attacks on your work without actually challenging any of the content. And he said people at CA were anti and closed minded. Again, no actual argument, just characterization. 

 

I think it's clear to him he doesn't have the knowledge or ability to answer the arguments made here.

He'd rather just keep repeating "what he's been told" - that fit's his pro MQA, pro monopolistic, pro "big corporations are good for you" POV. 

[JoeWhip]

Just read Mr. Scoggins postings over at Hoffman Forums. Just another example of the old tactic of attacking the person and not the data. Put another way, always attack, never defend. Sound familiar? I see this constantly from the MQA side. 

[WiWavelength]

On 3/15/2018 at 10:05 AM, miguelito said:

Where did you get that from?? So you're saying that deblurring cannot be done on any recording done to redbook format such as Dire Straights's "Brothers in Arms"???

 

12 hours ago, Fokus said:

2) Start at 96k or lower. In this case there is no downsampling possible, so they will have to revert to eq-style solutions. This can be additional narrow-impulse low-pass filtering for a 96k original (i.e. apodising in Meridian-speak). For 44.1k and 48k not even this is possible, so all that could be done is some phase/group-delay equalisation, perhaps also level equalisation, at the treble end.

 

For "deblurring" 44.1/48 kHz masters, I think that we could be overlooking another possibility: MQA encoding applies upsampling to 88.2/96 kHz.

 

Remember that the Tidal desktop app with MQA software decoder enabled always outputs a first "unfold," even for 44.1/48 kHz masters.  And the Tidal MQA decoder is not defined as a "renderer," suggesting that this double rate output may not be upsampling on the fly but actual decoding of whatever was encoded.

 

Furthermore, there is precedence for this type of upsampling before encoding.  While this got very limited use and never took off on Blu-ray a few years ago, Dolby TrueHD with Advanced 96k Upsampling took a typical 48 kHz soundtrack and upsampled to 96 kHz with an apodizing digital filter prior to encoding on Blu-ray.

 

https://www.stereophile.com/content/dolby174-upsamples-apodizes-truehd

 

Last but not least, a major reason for Dolby implementing this upsampling and apodizing approach was Rhonda Wilson.  Rhonda had moved over to Dolby Labs from, of course, Meridian.

 

Per usual, I take no stance on the efficacy or benefits of this process.  I just am interested in understanding its internal functions.  And I do believe that MQA "deblurring" 44.1/48 kHz with an upsampling and apodizing digital filter at the time of encoding is a real possibility.

 

AJ

[Archimago]

3 hours ago, WiWavelength said:

 

 

For "deblurring" 44.1/48 kHz masters, I think that we could be overlooking another possibility: MQA encoding applies upsampling to 88.2/96 kHz.

 

Remember that the Tidal desktop app with MQA software decoder enabled always outputs a first "unfold," even for 44.1/48 kHz masters.  And the Tidal MQA decoder is not defined as a "renderer," suggesting that this double rate output may not be upsampling on the fly but actual decoding of whatever was encoded.

 

Furthermore, there is precedence for this type of upsampling before encoding.  While this got very limited use and never took off on Blu-ray a few years ago, Dolby TrueHD with Advanced 96k Upsampling took a typical 48 kHz soundtrack and upsampled to 96 kHz with an apodizing digital filter prior to encoding on Blu-ray.

 

https://www.stereophile.com/content/dolby174-upsamples-apodizes-truehd

 

Last but not least, a major reason for Dolby implementing this upsampling and apodizing approach was Rhonda Wilson.  Rhonda had moved over to Dolby Labs from, of course, Meridian.

 

Per usual, I take no stance on the efficacy or benefits of this process.  I just am interested in understanding its internal functions.  And I do believe that MQA "deblurring" 44.1/48 kHz with an upsampling and apodizing digital filter at the time of encoding is a real possibility.

 

AJ

 

Looks like Meridian has been throwing darts every which way to monetize their digital filter. Hoping of course that one of those darts will stick and bring in some revenue eventually!

 

Remarkable how JVS always hears how good these Meridian filters are... I suspect if the term had been invented to be used as such, "deblurring" would have been part of the presentation :-).

 

[crenca]

3 hours ago, WiWavelength said:

 

 

For "deblurring" 44.1/48 kHz masters, I think that we could be overlooking another possibility: MQA encoding applies upsampling to 88.2/96 kHz.

 

Remember that the Tidal desktop app with MQA software decoder enabled always outputs a first "unfold," even for 44.1/48 kHz masters.  And the Tidal MQA decoder is not defined as a "renderer," suggesting that this double rate output may not be upsampling on the fly but actual decoding of whatever was encoded.

 

Furthermore, there is precedence for this type of upsampling before encoding.  While this got very limited use and never took off on Blu-ray a few years ago, Dolby TrueHD with Advanced 96k Upsampling took a typical 48 kHz soundtrack and upsampled to 96 kHz with an apodizing digital filter prior to encoding on Blu-ray.

 

https://www.stereophile.com/content/dolby174-upsamples-apodizes-truehd

 

Last but not least, a major reason for Dolby implementing this upsampling and apodizing approach was Rhonda Wilson.  Rhonda had moved over to Dolby Labs from, of course, Meridian.

 

Per usual, I take no stance on the efficacy or benefits of this process.  I just am interested in understanding its internal functions.  And I do believe that MQA "deblurring" 44.1/48 kHz with an upsampling and apodizing digital filter at the time of encoding is a real possibility.

 

AJ

 

Did you see the moderator just closed the Roon thread?  Is the moderator overstepping his bounds or is Roon afraid of an honest discussion about MQA?

[mansr]

3 hours ago, WiWavelength said:

Remember that the Tidal desktop app with MQA software decoder enabled always outputs a first "unfold," even for 44.1/48 kHz masters.  And the Tidal MQA decoder is not defined as a "renderer," suggesting that this double rate output may not be upsampling on the fly but actual decoding of whatever was encoded.

The core decoder includes the ability to upsample 44.1/48 sources to double rate.

[WiWavelength]

15 minutes ago, crenca said:

 

Did you see the moderator just closed the Roon thread?  Is the moderator overstepping his bounds or is Roon afraid of an honest discussion about MQA?

 

That is just the modus operandi in the Roon Community forums: enforced politeness at the expense of logical criticism.

 

I am a Roon lifetime subscriber, and while I like the software, I could do without many of the people, both among the users and on the Roon team itself.

 

AJ

[crenca]

1 hour ago, WiWavelength said:

 

That is just the modus operandi in the Roon Community forums: enforced politeness at the expense of logical criticism.

 

I am a Roon lifetime subscriber, and while I like the software, I could do without many of the people, both among the users and on the Roon team itself.

 

AJ

 

The moderator did explain himself here:

 

https://community.roonlabs.com/t/moderation-phrases-that-draw-an-emotional-response/40083

 

I asked him outright if Roon is trying to sidestep the debate about MQA.  Of course they are, real debate and disagreement is uncomfortable, and possibly bad for business.  Their problem is that their product naturally puts them right in the middle of it.  They will want to make it about me (the first reply already does) but lets see if they will allow MQA it's fair hearing...

 

I am also a lifetime subscriber and can hardly say enough good (and very little bad) about Roon.  However, they appear to be sheltering a few pro-MQA folks from the hard questions...

[WiWavelength]

1 hour ago, mansr said:

The core decoder includes the ability to upsample 44.1/48 sources to double rate.

 

Yes, the Tidal app decoder double rate output with 44.1/48 kHz masters very well may not be relevant here.  Just a thought.

 

But if the widely reported MQA claims that even undecoded files benefit from "deblurring" are to be believed, then that must entail some time domain pre processing at the encoding stage.  And as Meridian prior art, an apodizing upsampling digital filter would seem a likely candidate.

 

AJ

[Archimago]

1 hour ago, WiWavelength said:

 

Yes, the Tidal app decoder double rate output with 44.1/48 kHz masters very well may not be relevant here.  Just a thought.

 

But if the widely reported MQA claims that even undecoded files benefit from "deblurring" are to be believed, then that must entail some time domain pre processing at the encoding stage.  And as Meridian prior art, an apodizing upsampling digital filter would seem a likely candidate.

 

AJ

 

Again, it comes back to the speculation of what "deblurring" is. If "deblurring" is what makes an undecoded 24/44 or 24/48 MQA file "sound better" as they suggest, then there must be some processing being done during the encoding that affects those upper unencoded bits that they believe improves time-domain performance apart from the min phase/slow roll filter.

 

As I said yesterday about the "Daubert standard", until there is some actual science-based evidence about this, I refuse to believe "deblurring" (at least regarding the time-domain claim) is even a "thing" this scheme is able to improve/fix.

[OldBigEars]

I’m looking forward immensely to Roon providing MQA streaming. Great,  distinct upgrade in sound quality over red book for no cost.  What not to love, eh? 

[Archimago]

27 minutes ago, OldBigEars said:

I’m looking forward immensely to Roon providing MQA streaming. Great,  distinct upgrade in sound quality over red book for no cost.  What not to love, eh? 

 

Well, Roon is excellent software.

 

Given the close Tidal integration, I would want it implemented as well to take advantage of all that Tidal has to offer!

[astromo]

2 hours ago, Archimago said:

 

Again, it comes back to the speculation of what "deblurring" is. If "deblurring" is what makes an undecoded 24/44 or 24/48 MQA file "sound better" as they suggest, then there must be some processing being done during the encoding that affects those upper unencoded bits that they believe improves time-domain performance apart from the min phase/slow roll filter.

 

As I said yesterday about the "Daubert standard", until there is some actual science-based evidence about this, I refuse to believe "deblurring" (at least regarding the time-domain claim) is even a "thing" this scheme is able to improve/fix.

I really enjoyed reading about the "Daubert Standard". I'm not from the U.S., so it was news to me. Reads like very good legal sense.

 

Keep nailing the point of "deblurring" and the lack of scientific evidence to demonstrate that MQA is doing anything of value. It's always been a key issue for me.

[guymrob]

I’ve a quote from KR March 2018 Stereophile that can probably sum up what most of us think:

 

“Today, I’m less enthusiastic about MQA than I was. High-resolution streaming and downloads are now readily available …. MQA requires the purchase of compatible equipment, and holds the potential to eventually control all signal processing, such as room EQ. I don’t see a need for it, therefore, and I hope it doesn’t force the elimination of high-resolution, non-MQA downloads.” — Kalman Rubinson. “Multichannel MQA”, Stereophile, March 2018

[astromo]

3 hours ago, guymrob said:

I’ve a quote from KR March 2018 Stereophile that can probably sum up what most of us think:

 

“Today, I’m less enthusiastic about MQA than I was. High-resolution streaming and downloads are now readily available …. MQA requires the purchase of compatible equipment, and holds the potential to eventually control all signal processing, such as room EQ. I don’t see a need for it, therefore, and I hope it doesn’t force the elimination of high-resolution, non-MQA downloads.” — Kalman Rubinson. “Multichannel MQA”, Stereophile, March 2018

"Beep, beep, beep ..."

 

 - don't you love the sound of a reverse beeper? 

[Norton]

12 hours ago, WiWavelength said:

the Tidal desktop app with MQA software decoder enabled always outputs a first "unfold," even for 44.1/48 kHz masters

 

This is one thing I don't get.  Where the master is 48 kHz,  what does  the initial decode do and how is this materially different from just playing the 48kHz undecoded file?

[Fokus]

14 hours ago, WiWavelength said:

For "deblurring" 44.1/48 kHz masters, I think that we could be overlooking another possibility: MQA encoding applies upsampling to 88.2/96 kHz.

 

That might well be, although it would be a pretty strange approach.

 

When I was fooling around with Tidal and an Explorer2 I looked into 48k recordings such as Lemonade, but I did not investigate the digital-domain output of the Tidel core decoder. At least I have no recollection of this, and no such files remain on my server. I only have the raw MQA stream and a recordering from the analogue output of the Explorer.

 

But even if they did so, i.e. taking a 48kHz master, upsampling it to 96kHz, and then folding it into an MQA file, it would not do anything about the original 48k ADC ringing. This ringing cannot be removed, because this would mean cutting it out with a shallow filter starting far below 20kHz, and losing the music's treble in the process. Only the original pre-ringing can be converted into post-ringing, by means of a minimum phase filter cutting slightly below 24kHz (traditional Meridian CD player approach) or by manipulating phase above 20kHz (MQA's funny patent). Neither of these approaches require upsampling, they work perfectly fine in a 48kHz space.

 

The only reason I see for adding upsampling to 96kHz during encoding would be to give listeners without an MQA DAC broadly the same post-decoding signal as listeners with an MQA DAC, for 1x rate files. (Meaning leaky, with a lot of imaging.)

 

So in short: such upsampling might well be part of 1x rate MQA, but it does not address the deblurring issue.

 

 

[botrytis]

I agree with Archimago. Until we know what 'deblurring' means to MQA (aka Stuart), we can guess until the cows come home. 

[Archimago]

6 hours ago, Norton said:

 

This is one thing I don't get.  Where the master is 48 kHz,  what does  the initial decode do and how is this materially different from just playing the 48kHz undecoded file?

 

Hi Norton,

With MQA decoding, they could potentially recover a little more of the original bit depth.

 

As shown above with that Buno Mars sample in the article, we also see the "leaky" filter at work.