MQA is Vaporware

[John_Atkinson]

7 minutes ago, lucretius said:

3 hours ago, Rt66indierock said:

I'm giving reviewers until next year to get one or I'm viewing them as third class citizens in the audio world.

 

Surely, Stereophile can afford one of those.

 

Stereophile the magazine doesn't own any test equipment. Senior management at the company that owned Stereophile in 2004 decided to discontinue Stereophile publishing technical tests. I therefore decided at that time to invest personally in the necessary test equipment so that the magazine could keep publishing reviews accompanied by measurements. Since then I have spent more than $40k on test equipment. My measurement system is based on the Audio Precision SYS2722, but I am currently trying an Audio Precision 5-series analyzer, which is not inexpensive.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

10 hours ago, Jud said:

- In order to minimize aliasing and imaging from ultrasonics, a slow-rolloff filter must start cutting in the audible range. Do you regard a recording rolled off in the upper audible range to be a “typical high-quality music recording”?

 

When it comes to recording with slow-rolloff antialiasing filters, as the original sample rate is generally 2Fs or 4Fs, there is no significant top-octave rolloff in the audioband. For example, the "Listen" filter of Ayre's QA-9 A/D converter with a sample rate of 192kHz reaches –3dB at 70kHz but is flat in the top octave (–0.1dB at 20kHz).
 

With playback of CD-resolution recordings, a slow-rolloff reconstruction filter typically gives a rolloff reaching between 1dB and 3dB at 20kHz. See fig.8 at https://www.stereophile.com/content/mytek-hifi-brooklyn-da-processorheadphone-amplifier-measurements

for example, reproduced below. I doubt that is audibly significant. YMMV.

 

BTW, IIRC it was mentioned elsewhere in this thread that Ayre's Charley Hansen was not a fan of minimum-phase reconstruction filters. This is not correct, as can be seen from the impulse responses of his "Music" and "Listen" filters,  both minimum-phase, at https://www.stereophile.com/content/ayre-acoustics-qx-5-twenty-da-processor-measurements

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

45 minutes ago, Doug Schneider said:

1 hour ago, John_Atkinson said:

With playback of CD-resolution recordings, a slow-rolloff reconstruction filter typically gives a rolloff reaching between 1dB and 3dB at 20kHz. See fig.8 at https://www.stereophile.com/content/mytek-hifi-brooklyn-da-processorheadphone-amplifier-measurements for example, reproduced below. I doubt that is audibly significant. YMMV.

 

If you've ever played with tweeter rolloffs, 1-3dB is significant and clearly audible as you're usually talking about a fairly wide bandwidth in the top octave of the audioband.

 

Fairly wide bandwidth? Not really, Yes, if you are talking about the level of a tweeter, I have found, in a blind test, that I can detect a level difference of just 0.5dB. But that 0.5dB difference covered 2.5kHz-20kHz, ie, 3 octaves, which is a large "area under the curve." In the case of the example of the slow-rolloff reconstruction filter I gave,  the output is flat to 10kHz, -0.1dB at 13kHz, -0.86dB at 17kHz, and -2.4dB at 20kHz, ie, the area under the curve is very small. And that area is in a region where human hearing sensitivity is reduced compared with frequencies below 13kHz. I doubt that it will be audible.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

25 minutes ago, Jud said:

With such a filter at the recording end, an MQA DAC filter would alias and image, causing intermodulation distortion.

 

As I have written, both in this thread and in Stereophile but you must have missed, the probability of there being aliased image energy in the audioband with high-quality recordings of music having a typical spectrum, is very low.  Yes, with music having high amounts of top-octave energy, such as victims of the Loudness Wars that can have an almost-white spectrum, this probability is very much higher. But such recordings sound awful even with steep-rolloff, linear-phase anti-imaging filters.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

3 hours ago, mansr said:

If you look at your own graphs, you'll see that even the Ayre "listen" filter reaches about 90 dB attenuation at 40 kHz and stays there. In contrast, the MQA filters achieve at best 40 dB attenuation apart from a few narrow dips. Like this:

 

 I am note sure what filter that is, other than a simple moving-average type.  If you look at fig.2 in my measurements of the Mytek Liberty DAC (below) - which uses one of the MQA filters for all PCM data - see https://www.stereophile.com/content/mytek-liberty-da-processor-measurements - the filter has a slow rolloff off above the audioband, with very little suppression of the image at 25kHz of a 19.1kHz tone. However, there are no aliased images of this high-level tone in the audioband and the stop-band attenuation is consistent with frequency..

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

3 hours ago, crenca said:

5 hours ago, John_Atkinson said:

 

 I am note sure what filter that is, other than a simple moving-average type.  If you look at fig.2 in my measurements of the Mytek Liberty DAC (below) - which uses one of the MQA filters for all PCM data - see https://www.stereophile.com/content/mytek-liberty-da-processor-measurements - the filter has a slow rolloff off above the audioband, with very little suppression of the image at 25kHz of a 19.1kHz tone. However, there are no aliased images of this high-level tone in the audioband and the stop-band attenuation is consistent with frequency..

 

John Atkinson

Technical Editor, Stereophile

 

Read more  

 

Before I scrolled down to @mansrpost JA I saw this and said "this is not an MQA filter".

 

An error has been made...

 

No error. I as I wrote, this is the filter that a DAC that doesn't have any other reconstruction filters other than MQA applies to non-MQA, linear PCM data. The spectrum was taken from the Mytek processor's analog output and extends to 100kHz because 200kHz is the A/D converter's sample-rate limit of my current Audio Precision analyzer. Given that the analog stage of a D/A processor is rolling off above 100kHz, I don't see that as any kind of limitation in the measurement.

 

I do note that mansr's spectrum extended to >384kHz, which implies his analyzer supports an A/D sample rate of at least 768kHz. Perhaps he would share with us what analyzer he used. If, that is, he examined the analog output of a D.A processor and didn't simulate the spectrum with, for example, MatLab.

 

John Atkinson

Technical Editor, Stereophile

 

 

 

[John_Atkinson]

8 minutes ago, mansr said:

My graph was created using the Matlab 'freqz' function with the actual filter coefficients extracted from a DAC.

 

Thank you.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

4 minutes ago, crenca said:

MQA (as what - a software application I assume you mean) has a particular filter it applies to non MQA encoded PCM? 

 

The set of MQA's digital filters that a D/A processor uses to decode MQA-encoded files includes this filter that is intended to be used with conventional linear-PCM files if the processor doesn't have other filters. Like this Mytek, for example, or the Bel Canto Black and Aurender A10 I mentioned earlier in this thread.

 

The identical filter is used in every D/A processor Stereophile has reviewed that will decode MQA files, so no, it is not exclusive to Mytek.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

1 minute ago, crenca said:

So this is not the particular Mytek model that was applying one or more of the filter(s) that were created by MQA for MQA encoded files to standard PCM, unless the end user rebooted (or something...

 

No, this is the Mytek Liberty, which applies this MQA filter to all PCM files.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

5 hours ago, mansr said:

16 hours ago, Doug Schneider said:

By the same token, when I read that Naim Audio "downsamples" data in JA's technical measurements in this review, it gave me pause for thought: https://www.stereophile.com/content/naim-audio-uniti-nova-integrated-amplifier-media-player-measurements

 

But I didn't close the door on the thought. I got in touch with Naim Audio's director of engineering and asked them why they were downsampling data. He told me they weren't resampling anything -- that they design all their components to roll off output at about 27kHz (if memory serves me). He said that I should fine their preamps will do the same. Why not ask questions?

Read more  

I would have gone a step further and taken the thing apart to see what it really was doing.

 

Perhaps, but I described  my attitude to reviewing at this link – https://www.stereophile.com/content/2011-richard-c-heyser-memorial-lecture-where-did-negative-frequencies-go - writing “there was one experience that foreshadowed my career as an audio reviewer. For one of my bachelor's degree final exams, I was handed a black box with two terminals and had to spend an afternoon determining what it was. (If I recall correctly, it was a Zener diode in series with a resistor.) That experience is echoed every day in my endeavors to characterize the performance of the audio components reviewed in Stereophile - every product, be it speaker, amplifier, CD player, is fundamentally a black box with input and output terminals. All I have to do is ask the question ‘What does it do?’"

 

I must admit some surprise reading that I apparently stated as fact that the Naim Audio Uniti Nova "downsamples" data. I didn’t remember writing that and if you go to the link provided, you see that I actually wrote “I suspect that the Uniti Nova downsamples high-resolution data so that its DSP can be applied to those data.” I provide my reasoning for that conjecture in the review, primarily because, as I found, the Naim converts its analog input signals into digital with a sample rate of 48kHz to allow it to be processed with DSP.

 

Naim did respond to my implied question in their Manufacturer’s Comment in the same issue as the review, saying “JA also wrote that he said he suspects we downsample to get high sample rates into a DAC chip that is billed as 192kHz on the Burr-Brown website. But we don’t. The bottleneck is the DAC chip’s internal digital filter. We bypass the DAC’s internal filter and do all digital filtering in the SHARC DSP. The DSP will play natively up to 384kHz, and integer-oversamples that to 768kHz before sending it directly to the DAC element in the DAC chip.”

 

So whether a “bottleneck” limits sample rate or the Naim has an analog low pass filter that results in the output lying at -1dB at 20kHz and at -9dB at 29kHz, to be frank I regardless this as a distinction without a difference. While the Naim will accept digital data sampled at up to 384kHz, the question whether there is any audible benefit of the high sample rates  becomes moot when the files are auditioned with Naim’s Uniti Nova.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

13 hours ago, The Computer Audiophile said:

Any chance you can provide us a track of yours pre and post MQA processing? I figure that because you own the rights this shouldn’t be like pulling teeth. 

 

I don't own the rights. I was a hired gun for the recording projects.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

2 hours ago, Ishmael Slapowitz said:

Yet, as a hired gun, you felt perfectly free to send the master files to Bob Stuart. 

 

 

Yes, as this was for my private use.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

3 hours ago, The Computer Audiophile said:

How bout a little fair use snippet?

 

People can find my analyses of these recordings at www.stereophile.com.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

9 minutes ago, John Dyson said:

I do believe that properly cleared demonstrations should be done, especially with such unbelievable claims about MQA. 

 

As was noted many many postings earlier in this thread, Wilson Audio's Peter McGrath has performed comparisons of the MQA versions of some of his files and the original PCM versions at dealer events and audio shows.

 

Regarding my unwllingness to share the MQA versions of some of my own recordings, as I explained, these versions were made for my private use only.

 

John Atkinson

Technical Editor, Stereophile 

[John_Atkinson]

1 hour ago, The Computer Audiophile said:

A technology so impressive has never been so cloaked in secrecy and hidden from those willing to get onboard. 

 

There was absolutely nothing, Chris, to prevent you 4 or 5 years ago from asking Bob Stuart to encode some of your own recordings or those made by engineers with whom you were friendly. Now, of course, I doubt Stuart would be amenable to any such request from you. Which is why, I suppose, you are trolling me 🙂

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

1 hour ago, The Computer Audiophile said:

I'm not trolling you by asking reasonable questions.

 

I used the word "trolling" Chris because I have been asked this question before and I assume you knew what my answer would be.

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

25 minutes ago, labjr said:

Trolling?  It's not like he went to your forum.

 

As well as Chris's question posted on this forum, I was sent an email alert.

 

John Atkinson

Technical Editor, Stereophile

 

 

[John_Atkinson]

24 minutes ago, Thuaveta said:

Should one understand that the senior technical editor of Stereophile doesn't know that you can fix a simple annoyance like this...

 

It's not an annoyance. I was explaining how I learned of Chris's question, to which he has now acknowledged that he already knew the answer. Hence my use of the word "troll" (which I beginning to believe applies to you also - see below).

 

24 minutes ago, Thuaveta said:

to top it all, he's bitching about it ?

 

I am not bitching, that is your projection. (Perhaps English is not your first language?)

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

1 hour ago, crenca said:

I recall reading (can't recall where - probably SBAF) where JA had Border Patrol DAC in house that had an unusually high output impedance, so JA went ahead and measured it into an equally unusually high impedance.

 

 

I measured the BorderPatrol DAC into 100k ohms, which is what I do with all DACs and preamplifiers.This is not an "unusually high impedance" and is standard for audio analyzers. I do mention when a product has problems driving low impedances. If it doesn't I also measure into the punishing 600 ohm load. If the product is borderline, I measure how it behaves into increasingly low impedances starting at 10k ohms, which is about the lowest input impedance a DAC might see.

 

Quote

This of course had the effect of improving the DAC's measurements in the test, typical use case be damned.

 

As you will see if you read the review, I did not try to play down the BorderPatrol DAC's measured performance. In fact, the manufacturer objected strenuously to what I wrote, in particular my final summing up where I concluded "The BorderPatrol Digital to Analogue Converter SE's measured performance is dominated by its use of the underperforming TDA1543 DAC chip."

See www.stereophile.com/content/borderpatrol-digital-analogue-converter-se-measurements

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

On 3/8/2021 at 8:42 PM, Ishmael Slapowitz said:

Either Atkinson has no integrity, or he is obsessed with giving MQA as many mentions as he possibly can before he rides off into the sunset.  Pathetic, and sad. I would not buy a box of paperclips based on his recommendation.

 

https://www.stereophile.com/content/grimm-audio-mu1-music-streamer

 

Thank you for reading my review so thoroughly, Mr, Slapowitz. However, I am afraid I don't know enough about different kinds of  paperclips to be able to make a recommendation.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

22 hours ago, Currawong said:

Very often Stereophile writers were handing over music directly to the MQA group for processing, and, in cases where it was analysed, receiving it back without it having gone through the origami compression, and thus still showing completely intact content when a spectrum was posted. If I'm wrong, @John_Atkinson can correct me.

That's not correct. The MQA-encoded files were all 24/44.1k.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

7 hours ago, JoeWhip said:

John, do you not think that the differences you heard could be created in settings in playback software?

I don't think so. Also, peak levels were matched.

 

When a manufacturer visited a while back (pre-pandemic) we were talking about MQA so I asked him if he would mind taking part in a single-blind listening test. He agreed, so with peak levels matched I played him the original 24/88.2k files and the MQA-encoded 24/44.1k versions unfolded to 88.2k with Roon. I didn't identify what he was listening to until afterwards. And as I was sitting behind the manufacturer and to the side, he couldn't see my face and my body language would not influence the results.

 

I played A-B, B-A, etc, so the usual fact that the second time the file is played will be preferred was not a factor.

 

He consistently preferred the unfolded MQA files, and described the difference in the same terms that I have written about in the magazine. See

 

John Atkinson

Technical Editor, Stereophile

 

[John_Atkinson]

12 hours ago, JoeWhip said:

Not sure you understood my point JA. It is that I can use filters and settings in software like Sox etc. to end up with the same sonic signature on a given mqa file. Hence, we do not need Mqa. 

 

As I was comparing a hi-rez file with an MQA version of the same file, use of plugins would have been a confusing variable.

 

And I have no idea if the audible improvement in sound quality that I reported finding with MQA could be duplicated with plugins.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

6 hours ago, SoundAndMotion said:

Slamming Jbara for his behavior in your RMAF talk, and expected behavior in the session on streaming is one thing; slamming all of AES is inappropriate, IMHO.

 

Agree. And one of the participants, Vicki Melchior, is one of the most respected DSP experts around.

 

John Atkinson

Technical Editor, Stereophile

[John_Atkinson]

1 hour ago, Thuaveta said:

 

your conclusion is that @The Computer Audiophile is unfairly calling the panel an infomercial, because there's one person on it that is clearly extraordinarily competent and likely still has a shred of integrity left.

Please don't  put words in my mough. With all due respect, I think you are arguing with the voices in your head, Thuaveta.

 

John Atkinson

Techncial Editor, Stereophile