MQA is Vaporware

[T.S. Gnu]

On 12/1/2021 at 8:01 AM, The Computer Audiophile said:

McNugget are perceptually chicken.

Indeed. They are <50% chicken, with a lot of filler. Analogies, anyone? Perhaps there is a lesson to be learned from the food processing industry. 
 

I proposing renaming MQA to Beyond Hi-Res. Much like Beyond Meat, which is not meat, Beyond Hi-Res is not Hi-Res, and doesn’t contain anything bearing semblance to Hi-Res. Despite the lack of meat, people have a preference for Beyond Meat. So too for Beyond Hi-Res. After all, as the Austins and Atkinsons say, it’s all about taste and preference. 
 

Another suitable candidate could be the Impossible Bitstream.* They could also tout their environment-friendly cred because it uses half the energy since only half the resources would be used in transferring the…oh heck, they already do that. ( ͡° ͜ʖ ͡°) 

 

*And, just in time, all hail the new Impossible Nuggets now available at Burger King. A lot of people like ‘em (as I too might), but none of them push the narrative that there’s chicken in them thar nuggets.

[The Computer Audiophile]

17 minutes ago, T.S. Gnu said:

Indeed. They are <50% chicken, with a lot of filler. Analogies, anyone? Perhaps there is a lesson to be learned from the food processing industry. 
 

I proposing renaming MQA to Beyond Hi-Res. Much like Beyond Meat, which is not meat, Beyond Hi-Res is not Hi-Res, and doesn’t contain anything bearing semblance to Hi-Res. Despite the lack of meat, people have a preference for Beyond Meat. So too for Beyond Hi-Res. After all, as the Austins and Atkinsons say, it’s all about taste and preference. 
 

Another suitable candidate could be the Impossible Bitstream.* They could also tout their environment-friendly cred because it uses half the energy since only half the resources would be used in transferring the…oh heck, they already do that. ( ͡° ͜ʖ ͡°) 

 

*And, just in time, all hail the new Impossible Nuggets now available at Burger King. A lot of people like ‘em (as I too might), but none of them push the narrative that there’s chicken in them thar nuggets.

You’re on a roll tonight :~)

[GregWormald]

3 hours ago, T.S. Gnu said:

Beyond Hi-Res is not Hi-Res

 Exactly. "Beyond" does not indicate a direction. If you are going down then even 128K mp3 is "beyond hi-res".

 

It's easy to lie when you know how.

[yahooboy]

20 hours ago, garrardguy60 said:

 

Strictly speaking, redbook CD goes up to 22.05 kHz, as per Nyquist. By definition, the Nyquist frequency is 22.05 kHz, because that is half the sample rate is 44.1 kHz.

Thanks I am aware of this, my question was toward FredericV's talking about MQA containing a full octave above redbook (which would be 40 kHz)

Since the measurements that show the MQA gap, seems to show that the max frequency handled by MQA is actually 24 kHz.

I was trying to find out whether or not I have/had misunderstood this.

[Fokus]

On 12/1/2021 at 4:40 PM, yahooboy said:

You say an octave above redbook. Redbook goes to 20 kHz, an octave above that is at 40 kHz. But ... it seems that Actual frequency response only goes to approx. 24 kHz.

Or have I gotten this wrong? 

 

Yes.

 

You must make distinction between at the level of the source material's sample rate. MQA handles CD-rate, 2x-rate, and 4x/8x-rate differently.

 

In the case of 2x-rate MQA passes the signal more or less unmolested, i.e. with a bandwidth extending to 44kHz or 48kHz.

[bogi]

One example of product price difference with/without  MQA:

$340 SMSL SU-9n
$400 SMSL SU-9

 

Interestingly the non MQA version has lower THD+N.
https://shenzhenaudio.com/pages/search-results-page?q=SMSL+SU-9

[yahooboy]

18 hours ago, Fokus said:

 

MQA passes the signal more or less unmolested

So You're a standup comedian.

 

Original master 

MQA version

 

Yup, pretty much unmolested.........

[Stereo]

Lots of fools giving comments to the posted topic there.

[John Dyson]

19 minutes ago, Stereo said:

Lots of fools giving comments to the posted topic there.

I agree with the article, and it makes sense on some important points.

 

The most important use for sample rates above 44.1k (actually need 96k for dynamics processing), and higher bit-depth is for professional purposes.   Professional processing is sometimes nonlinear and also does dynamics processing .  These processing schemes create sidebands or harmonics -- both need wider than 22kHz to be properly contained.

 

For other reasons, bit-depth of greater than 16bits are also helpful in professional applications.

 

When people say '16bits is good enough', that is for signal transport, e.g. the playout audio file.  The equipment being used for playback should internally have higher precision when needed.

 

It is very easy to find places where 16bits isn't good enough, but if done properly, not needed for consumer recordings.  The confusion about 16bits not being enough for consumer purposes might happen because of mixing the idea of 'transport' like a CD, vs processing like internal to a CD player.  There is often a need for greater than 16bits precision while processing a 16bit input signal, but this greater precision isn't really needed for audio files used for consumer playout only.   Unless the audio files are being used during significant signal processing processes, and the audio files are being used for listening only, then space/time for greater than CD specs is pretty much wasted.

 

 

 

 

 

 

 

 

 

 

[Currawong]

The benefit of high-res was that it reduced the often poor filtering built into many newer sigma-delta-based DACs. Since they seem to have improved quite a bit, the benefits of greater than 24/48 or 24/96 is null IMO, especially as many ADCs just generate noise above 48 kHz. Dan Lavry pointed this out I believe over a decade ago.

[Norton]

 I have some sympathy with the notion that hi res is more important in recording than distribution, but in that respect this article  is  as much a refutation of  the need for any HiRes streaming/downloads, SACDs or upsampling/transcoding for consumer replay, as it is a specific criticism of MQA.
 

But the whole notion of lossy vs lossless, regardless of who advocates it,  is largely just meaningless marketing BS anyway.  Lossless in relation to what?  
 

All “lossless“ indicates is that the format/codec distributed to the consumer is lossless in relation to a file that precedes  it in the production/ distribution chain, it doesn’t mean that that preceding file is itself  lossless in relation to the original recording or even master. I’d suggest that for the likely majority of  CDs or 16/44 streams/ downloads and a good proportion of SACDs or high res file distribution, the consumer is not being given the “Crown Jewels” of a lossless copy of the original master and never has, regardless of distribution format or codec. 

 

 

[Cebolla]

9 hours ago, Stereo said:

Lots of fools giving comments to the posted topic there.

 

The majority of tracks on the TIDAL HiFi quality connection are MQA anyway - either the original MQA-CD tracks or mangled from 24bit MQA ones!

[Fokus]

On 12/5/2021 at 6:30 AM, yahooboy said:

So You're a standup comedian.

Yup, pretty much unmolested.........

 

Remove the shaped ultrasonic noise, then look again.

 

The point is that MQA, when given a 2x or more source, has a mechanism for passing the part between 24kHz and 48kHz, whereas anything above 48kHz is thrown away(*).

 

(* Actually it is allowed to alias into the 0-48kHz band, which is worse than just throwing it away.)

[yahooboy]

3 hours ago, Fokus said:

 

Remove the shaped ultrasonic noise, then look again.

 

The point is that MQA, when given a 2x or more source, has a mechanism for passing the part between 24kHz and 48kHz, whereas anything above 48kHz is thrown away(*).

 

(* Actually it is allowed to alias into the 0-48kHz band, which is worse than just throwing it away.)

Sorry but I have to disagree, The MQA gap shows up due to the mirroring of the signal below 24kHz. Hence the signals above 24 kHz is not an original recorded signal. But of course, this might be the mechanism You refer to. 

 

Above image loaned from archimago's Musings http://archimago.blogspot.com

Note how the signal is mirrored (although lowered) in a way I have never seen in any recording

[Cebolla]

On 12/1/2021 at 3:40 PM, yahooboy said:

A question, You say an octave above redbook. Redbook goes to 20 kHz, an octave above that is at 40 kHz. But if one looks at the measurements where the dreaded "MQA gap" is present it seems that Actual frequency response only goes to approx. 24 kHz.

Or have I gotten this wrong? 

 

Wasn't the "MQA gap" found just for MQA files where original sample rate wasn't above 48kHz? So wouldn't expect the frequency response to go above 24kHz in that case, anyway. There should be no such ~22kHz gap for the above 48kHz original sample rate 24bit MQA files.

 

The gap & (fake) high frequency response is caused by the software MQA Core decoder's leaky upsampling filter - see:

https://archimago.blogspot.com/2017/06/measurements-audioquest-dragonfly-black_24.html

Recently, I came across this interesting YouTube post entitled "Tidal Masters MQA - Why the Frequency Spectrum Gap?":

 

 

Well, in light of what we're discussing here, you probably guessed that the cause for this "gap" is because of the MQA filter. Let's reproduce what the YouTuber showed...

 

Take Beyoncé's song "Love Drought" (off Lemonade) as used in the video (16/44 is fine); here's a convenient spot 40 seconds into the track:

 

We see that the 44kHz track rolls off starting at 20kHz.

 

Now, let's upsample with a filter setting that is modeled after MQA... Here's an iZotope RX 5 setting that approximates the 24/96 MQA filter from the Dragonfly:
 

 

Now let's have a look at what happened to the 44kHz track upsampled to 96kHz (it could of course just as well have been 88kHz, doesn't matter). Have a peek at the same spot as above, 40 seconds into the song:

 

 

Et voilà! There's your "gap". What you're seeing is a "mirror image" beyond 22.05kHz (Nyquist for 44.1kHz samplerate) due to imaging from the use of the MQA-like digital filter that has poor anti-imaging properties. These are of course false frequencies that were not part of the original song. Because this track has that significant roll-off of frequencies around 20kHz, it's easy to see this U-shaped anomaly.

 

Remember back in the day when HDtracks sold some fake high resolution tracks that were just upsampled 44/48kHz material as 88/96kHz? Well, this is an example of what happens when 44/48kHz audio is "unfolded" into 88/96kHz by the MQA Core decoder in TIDAL using their "leaky" filter of choice allowing the ultrasonic distortions to seep through. It's an example of "fake hi-res" MQA unfolding.

 

We have to really question what it means to be "Master Quality Authenticated" when this kind of stuff happens. Clearly that aliased image is NOT part of what Beyoncé and the engineers created, "heard" or likely intended to be present in the recording! I find it rather amusing that Lemonade in MQA should even be promoted as something worthy of "Master Quality".

[yahooboy]

6 hours ago, Cebolla said:

 

Wasn't the "MQA gap" found just for MQA files where original sample rate wasn't above 48kHz? So wouldn't expect the frequency response to go above 24kHz in that case, anyway. There should be no such ~22kHz gap for the above 48kHz original sample rate 24bit MQA files.

 

 

I have seen it show up in even in "96kHz" and above. It seems to me that this is the way that MQA convinces people that their Hires is actually Hires.

 

I've also seen this from Måns Rullgård:

 

Note that the frequencies above the gap are a mirror image of those below it with a slight downward slant. This is because the weak upsampling filter used by MQA lets these images through rather than suppress them as it ought to. Nothing above that gap actually existed in the recording. It is all fake. The 2L recording originally has a high sample rate, and MQA preserves it reasonably well up to 44.1 kHz. If you did a playback and (analogue) capture through a certified MQA DAC at 176.4 kHz (or higher), you'd see similar mirroring in the 44.1-88.2 kHz range.

[Cebolla]

5 hours ago, yahooboy said:

I have seen it show up in even in "96kHz" and above. It seems to me that this is the way that MQA convinces people that their Hires is actually Hires.

 

I suspect that you've misinterpreted something. Do you have any links to those that you have seen?

 

 

5 hours ago, yahooboy said:

I've also seen this from Måns Rullgård:

 

That's just Mans's response posted in the "Why the gap?" YouTube video page and it doesn't actually back up what you are saying with regards the ~22kHz gap showing up even in 96kHz and above original sample rate MQA, so not sure why you've quoted it.

 

 

Quote

Note that the frequencies above the gap are a mirror image of those below it with a slight downward slant. This is because the weak upsampling filter used by MQA lets these images through rather than suppress them as it ought to. Nothing above that gap actually existed in the recording. It is all fake.

Here Mans is referring specifically the Beyonce track in the video which has the ~22kHz gap, but its original sample rate is not above 48kHz (the original sample rate is actually 44.1kHz):

 

 

 

Quote

The 2L recording originally has a high sample rate, and MQA preserves it reasonably well up to 44.1 kHz. If you did a playback and (analogue) capture through a certified MQA DAC at 176.4 kHz (or higher), you'd see similar mirroring in the 44.1-88.2 kHz range.

This is referring specifically to the 2L track in the video which has an original sample rate above 48kHz and notably does not have the ~22kHz gap:

 

Mans is saying that you will see mirroring with the 2L track similar to the Beyonce track if you were to capture the playback through an MQA DAC. However, it will be in the 44.1-88.2kHz range, ie, not 22-44.1kHz. Here, the MQA upsampling filter is applied in the DAC by the MQA renderer, not the MQA Core decoder.

 

In fact, this 2L track is the type of MQA file that @Focusis also referring to and ironically, Mans actually says something similar to what @Focusmentioned & you first took issue with - "MQA preserves reasonably well up to 44.1kHz"! 😀

[Fokus]

Above is correct.

 

And the MQA-gap relates only to 1x source material.

[yahooboy]

On 12/7/2021 at 8:14 AM, Cebolla said:

 

I suspect that you've misinterpreted something. Do you have any links to those that you have seen?

 

 

 

That's just Mans's response posted in the "Why the gap?" YouTube video page and it doesn't actually back up what you are saying with regards the ~22kHz gap showing up even in 96kHz and above original sample rate MQA, so not sure why you've quoted it.

 

 

Here Mans is referring specifically the Beyonce track in the video which has the ~22kHz gap, but its original sample rate is not above 48kHz (the original sample rate is actually 44.1kHz):

 

 

 

This is referring specifically to the 2L track in the video which has an original sample rate above 48kHz and notably does not have the ~22kHz gap:

 

 

Mans is saying that you will see mirroring with the 2L track similar to the Beyonce track if you were to capture the playback through an MQA DAC. However, it will be in the 44.1-88.2kHz range, ie, not 22-44.1kHz. Here, the MQA upsampling filter is applied in the DAC by the MQA renderer, not the MQA Core decoder.

 

In fact, this 2L track is the type of MQA file that @Focusis also referring to and ironically, Mans actually says something similar to what @Focusmentioned & you first took issue with - "MQA preserves reasonably well up to 44.1kHz"! 😀

Let Me try to clear up a few things here. I might express myself a little clumsy, since English is not my first language, nor my third.

 

What I wrote to Focus was regarding his statement: MQA passes the signal more or less unmolested

As ought to be clear from the embedded pics of the original file uploaded by Goldensound and the file that MQA has molested via it's process. I think we can agree that those two files are not even close. Hence my objection to the unmolested part. It seems it wasn't as clear, as I thought

 

Regarding the "gap" or rather the mirroring that causes the "gap". I am aware that the "gap" moves with the sampling rate, the phenomenon shows up if the original recording does not contain signals all the way up to the max frequency. For instance the pic You posted has a hard cut at 20 kHz which leaves a bit from 20 to 22,5 kHz without signal. Here below it's a 44,1 kHz sample rate (family), the numbers would be a little different if the recording had been made at a 48 kHz sampling rate (family). However the mirroring would still be there.  

 

 

 

I hope this clears up what I was (trying to) write. From what I read we are saying the same thing.

[Archimago]

45 minutes ago, yahooboy said:

...

 

Regarding the "gap" or rather the mirroring that causes the "gap". I am aware that the "gap" moves with the sampling rate, the phenomenon shows up if the original recording does not contain signals all the way up to the max frequency. For instance the pic You posted has a hard cut at 20 kHz which leaves a bit from 20 to 22,5 kHz without signal. Here below it's a 44,1 kHz sample rate (family), the numbers would be a little different if the recording had been made at a 48 kHz sampling rate (family). However the mirroring would still be there.  

 

 

 

I hope this clears up what I was (trying to) write. From what I read we are saying the same thing.

 

Exactly. The "gap" is just a combination of the 20kHz cut-off and Nyquist at 22.05kHz using the 44.1kHz samplerate while upsampling with a poor filter that has a very slow roll-off and stopband attenuation (which is what mQa uses). This will shift and change depending on whether there's such an obvious cut-off and where Nyquist resides depending on the samplerate (ie. 24kHz with 48kHz samplerate, 48kHz with 96kHz samplerate). These slow roll-off filters typically are low tap-length, weak, and when you see them in an impulse response, they look "nice" with little ringing. mQa capitalized on this picture of the impulse response as if that's "good" and silly folks like Stereophile perpetuated that impression to the audiophiles as if this was meaningful in the "time domain".

 

The problem of course is that this allows the "mirror image" artifacts to come through which is what that right half of the picture represents. That space of 2kHz between 20-22.05kHz is magnified with the mirror 22.05-24kHz, leaving that striking anomaly due to the "imagine". A strong (proper) "brick wall" filter with good attenuation should be clean and not have all that ultrasonic stuff above 22.05kHz to show up, at least not to the magnitude we're seeing here.

[danadam]

47 minutes ago, yahooboy said:

I am aware that the "gap" moves with the sampling rate

Ah... so when Focus said

"MQA, when given a 2x or more source, has a mechanism for passing the part between 24kHz and 48kHz"

then you actually agreed with that, and "Sorry but I have to disagree" was only a typo? 😉🙂

[Archimago]

1 hour ago, Archimago said:

 

Exactly. The "gap" is just a combination of the 20kHz cut-off and Nyquist at 22.05kHz using the 44.1kHz samplerate while upsampling with a poor filter that has a very slow roll-off and stopband attenuation (which is what mQa uses). This will shift and change depending on whether there's such an obvious cut-off and where Nyquist resides depending on the samplerate (ie. 24kHz with 48kHz samplerate, 48kHz with 96kHz samplerate). These slow roll-off filters typically are low tap-length, weak, and when you see them in an impulse response, they look "nice" with little ringing. mQa capitalized on this picture of the impulse response as if that's "good" and silly folks like Stereophile perpetuated that impression to the audiophiles as if this was meaningful in the "time domain".

 

The problem of course is that this allows the "mirror image" artifacts to come through which is what that right half of the picture represents. That space of 2kHz between 20-22.05kHz is magnified with the mirror 22.05-24kHz, leaving that striking anomaly due to the "imagine". A strong (proper) "brick wall" filter with good attenuation should be clean and not have all that ultrasonic stuff above 22.05kHz to show up, at least not to the magnitude we're seeing here.

 

Opps that "imagine" should have been "image".

 

Anyone who wants to actually see the coefficients of these filters mQa use, check out Mans' file here: https://code.videolan.org/mansr/mqa/-/blob/master/render-filters.txt

(192kHz mQa lowpass FIR filters can be replicated with <50 taps - nothing all that complicated which is understandable since this has to run on devices like little Dragonfly DACs.)

[yahooboy]

14 hours ago, danadam said:

Ah... so when Focus said

"MQA, when given a 2x or more source, has a mechanism for passing the part between 24kHz and 48kHz"

then you actually agreed with that, and "Sorry but I have to disagree" was only a typo? 😉🙂

Nope just disagreed with the ultrasonic bit.

Which is the bit I initially commented on.

 

   On 12/6/2021 at 3:25 PM,  Fokus said: 

 

Remove the shaped ultrasonic noise, then look again.

[Fokus]

What are you trying to say anyway?

 

You mistakenly thought MQA cannot pass on signal in the 24-48kHz band.

 

I corrected this. It can and it does, given true hi-res source material.

 

Then you tried to refute this with (and I had to hunt the source of the images down) ... Goldensound's test signal as converted to MQA and not even decoded.

 

Of course that one looks, and is, mangled. That was predictable.

 

But this has nothing to do with the original contention.

[FredericV]

20 minutes ago, Fokus said:

You mistakenly thought MQA cannot pass on signal in the 24-48kHz band.

I corrected this. It can and it does, given true hi-res source material.

It indeed can, but it's not enough for their end-to-end claims.

There are real world audio signals above 48kHz as shown from Bob's own research, yet they keep using the misleading encoding triangle which cannot contain all of these signals. mQa is band limited to the audible spectrum and the first octave above the audible spectrum. Everything above that is fake using upsampling.