My Questions and Answers (MQA): An Interview with Andreas Koch

[audiventory]

45 minutes ago, Jud said:

 

I think the Xivero fellow is hypothesizing (of course no one knows for sure except the MQA folks) that MQA is using an apodizing filter at the recording/encoding stage to band-limit (not completely effectively), in to avoid some aliasing/ringing at the DAC end, since we know the MQA filtering at the rendering/DAC end is not effective to remove either of those.

 

I suppose we could see how effective any MQA recording/encoding process was at band-limiting by looking at the analog result on playback of a variety of MQA-encoded files.  (I'm thinking of MQA 44.1kHz files in particular, like Beyonce's "Lemonade.")

 

Main issue of comparison lossy formats is "threshold of audibility". Before all disputes on same topics ended without results.

[Jud]

Just now, audiventory said:

 

Main issue of comparison lossy formats is "threshold of audibility". Before all disputes on same topics ended without results.

 

Yes, what I was suggesting would not determine whether what MQA was doing was audible, either bad or good.  But since I haven't read anything that has been confirmed about the filtering at the MQA recording/encoding stage, I wondered if just examining the analog frequency spectrum of MQA 44.1kHz files on playback would show reasonable band-limiting in the upper audible and ultrasonic frequency ranges; and if we might guess from this that an apodizing filter was used in the recording/encoding stage as the Xivero fellow speculates.

[Fokus]

These are recordings off the output of a Meridian Explorer2, done at 192kHz. The HF bump is the combined modulator noise of the E2 and my ADC (a PCM1804).

 

Both traces are from the first track of Lemonade.

 

Red is the standard Tidal file, which plays at 44.1kHz. You see that the treble is cut off

sharply at 21kHz (which is a bit unusual for a modern recording, due to the use of half-band

ADC AA filters or downsamplers the spectrum more often runs flat to 22kHz).

 

Green is the MQA Tidal file, decoded by the E2. Again the treble cuts off at 21kHz, and rises again

above 23kHz. The part of the spectrum between 23kHz and ~36kHz clearly is an image of the baseband between and 8 and 21kHz. Strictly speaking, this is added distortion, and quite similar to the output of a NOS DAC.

From the Stereophile Mytek measurements it is known that the MQA replay filter is rather lazy and does not cut in below 22kHz, so it is not apodising at all, only passing along a lot of imaging.

 

[mansr]

4 hours ago, Fokus said:

These are recordings off the output of a Meridian Explorer2, done at 192kHz. The HF bump is the combined modulator noise of the E2 and my ADC (a PCM1804).

 

Both traces are from the first track of Lemonade.

 

Red is the standard Tidal file, which plays at 44.1kHz. You see that the treble is cut off

sharply at 21kHz (which is a bit unusual for a modern recording, due to the use of half-band

ADC AA filters or downsamplers the spectrum more often runs flat to 22kHz).

 

Green is the MQA Tidal file, decoded by the E2. Again the treble cuts off at 21kHz, and rises again

above 23kHz. The part of the spectrum between 23kHz and ~36kHz clearly is an image of the baseband between and 8 and 21kHz. Strictly speaking, this is added distortion, and quite similar to the output of a NOS DAC.

From the Stereophile Mytek measurements it is known that the MQA replay filter is rather lazy and does not cut in below 22kHz, so it is not apodising at all, only passing along a lot of imaging.

 

 

This result indicates that the input to the MQA encoder was 44.1 kHz. The HF hump is actually shaped dither added by the MQA decoder. DAC and ADC modulator noise should at most reach -120 dB.

 

Could you send me a capture of 10 seconds or so from that track? I'd like to see exactly what filter configuration they're using here.

[Fokus]

1) Of course it is 44.1kHz. That is what Jud asked for, above.

 

2) Get the PCM1804 datasheet. It really is my ADC's modulator noise.

 

3) You can have two fragments:

 

samp1

 

samp2

[mansr]

43 minutes ago, Fokus said:

1) Of course it is 44.1kHz. That is what Jud asked for, above.

 

2) Get the PCM1804 datasheet. It really is my ADC's modulator noise.

 

That ADC does indeed have a lot of noise. None of mine have anything near those levels. If you had a better ADC, you'd notice the MQA decoder adding dither with roughly the same noise profile. A firmware bug in the E2 makes it stay in this mode until power cycled after playing an MQA file.

 

43 minutes ago, Fokus said:

 

3) You can have two fragments:

 

samp1

 

samp2

 

Sorry, apparently I wasn't clear. I wanted a digital capture of the undecoded MQA stream.

[Fokus]

Sorry, can't do that. These were live recordings from Tidal. I never stored anything in the digital domain.

And I cancelled my Tidal subscription last weekend.

I was already puzzled why you would want the analogue captures, except perhaps for viewing the spectrum in real time.

[audiventory]

5 hours ago, mansr said:

Green is the MQA Tidal file, decoded by the E2. Again the treble cuts off at 21kHz, and rises again

above 23kHz.

 

May be it is border between separatelly restored frequency parts of original signal and the cut is transient band between two filters?

[Fokus]

The 21kHz cut-off is in the original, standard, recording and has nothing to do with MQA.

 

The green graph is just the result of upsampling with this filter

 

Mytek Brooklyn 'MQA' filter for baseband content

 

There are no mysteries here. The takeaway is that replaying MQA 1x recordings with an MQA DAC generates a lot of imaging, whereas is you replay MQA 2x or 4x or 8x recordings with an MQA DAC at least the 24-48kHz band remains clean (while anything above that so far also seems the result of upsampling, more details later).

[PeterV]

It is sad to notice that there is so much hostility against this groundbraking innovation..

 

MQA is a game changer, since it is the only technology available which is capable to correct for both past and present time-smear errors in the whole audio chain.

 

I've found a very informative publication on the subject of temporal decay http://www.breem.nl/Artikel...

Below a summary, with some personal additions:

 

High-end audio systems often sound better with analog recordings than with digital ones. The temporal decay is one of the few points at which analog systems beat their digital counterparts and it is therefore a very important parameter.

 

Any audio system has the tendency to 'smear out' the the signal both in amplitude and time, which always leads to a degradation of the original sound.

Digital PCM recording requires a very steep low pass filter before digitization to alleviate aliasing. Such a filter introduces anomalies including ripple, resonance at the cutoff frequency, oscillation (ringing) phase shift and high frequency loss.

 

CD quality has a temporal blur of 5 ms, 24/192 1/10th of CD (500 microseconds), but MQA has found a method to reduce this to 10 microseconds throughout the entire digital encoding and decoding chain.

 

Nowadays, such anomalies can be overcome by using oversampling techniques or by using another format like DSD. But it is a fact that all digital recordings in the 80's and 90's intrinsically are degraded by these influences and even nowadays 24/96 recordings are, since filtering can severely degrade the temporal behaviour of audio systems.

 

So..let's focus much more to this beautiful aspect of the new algorithm.. MQA is capable to repair the temporal blur of PCM recordings to a very large extend. It is therefore capable to restore the original sound of the Recording and not just that what has been stored on the final Mastertape. MQA enhances the temporal decay of the recording significantly and this is audible to everyone.

 

Every serious audiophile should first listen to MQA and make up their mind based on what they hear, feel and experience instead of judging the technology beforehand, based on speculative publications.

 

MQA is for sure not a fraud or just a pseudo MP3 compression tecgnology. It is a paradigm shift and a new, disruptive technology.

 

This is the reason why not everyone is happy with its existence and fear is firing up rumours and aligations. That is a shame, since it will be much wiser to embrace this stunning innovation and make use of it.

[Norton]

2 hours ago, PeterV said:

Every serious audiophile should first listen to MQA and make up their mind based on what they hear, feel and experience instead of judging the technology beforehand, based on speculative publications.

 

 

 

That seems very sensible advice.  For the serious listener: if it sounds better then all other concerns are effectively irrelevant; if it doesn't sound better then it will fail.

 

This is after all an audiophile site, not a forum for the mass music marketing industry.  By the standards of the latter, much of the body of hardware, software and recordings currently enjoyed by CA members  would most likely also be dismissed as vapourware too (if using the somewhat idiosyncratic definition proposed by MQA critics on this site)

 

[Jud]

4 hours ago, PeterV said:

It is a paradigm shift and a new, disruptive technology.

 

Actually it isn't.  Everything done by MQA has been done before (sometimes better).  Certain decisions the developers made about what MQA does (e.g., lossy compression) are at least controversial from a sound quality point of view.  The stuff about time smear is to some extent marketing "puffery."  (That's what I referred to above about things MQA claims to do having been done before and sometimes better.)  And while I agree there has been a great deal of simple hostility, MQA's real technical foibles have also been thoroughly discussed.  Much of the hostility is due to the potential for DRM built into MQA rather than any "disruptive technology."

 

I have no fight with anyone who likes the sound of MQA.  But there are valid reasons for the people who object to it to do so.  Everyone should therefore be comfortable with his or her own evaluation and not feel the need to say someone else's personal evaluation is wrong.

[mansr]

5 hours ago, PeterV said:

High-end audio systems often sound better with analog recordings than with digital ones. The temporal decay is one of the few points at which analog systems beat their digital counterparts and it is therefore a very important parameter.

That statement cost you whatever credibility you might have had.

5 hours ago, PeterV said:

Any audio system has the tendency to 'smear out' the the signal both in amplitude and time, which always leads to a degradation of the original sound.

Digital PCM recording requires a very steep low pass filter before digitization to alleviate aliasing. Such a filter introduces anomalies including ripple, resonance at the cutoff frequency, oscillation (ringing) phase shift and high frequency loss.

Normal microphones have practically no response above 50 kHz or so. With a sufficiently high sample rate, you thus don't need much of an analogue filter at all. 192 kHz is plenty, and even 96 kHz is enough in most cases. Of course, modern ADCs are sigma-delta designs operating at several MHz followed by a digital low-pass filter. The problems you allege simply do not exist.

5 hours ago, PeterV said:

CD quality has a temporal blur of 5 ms, 24/192 1/10th of CD (500 microseconds), but MQA has found a method to reduce this to 10 microseconds throughout the entire digital encoding and decoding chain.

Oh, no. Not this again. CD has a temporal resolution of a few picoseconds. We discussed this at length not long ago. Here's a post from that discussion where I demonstrate timing accuracy of 44.1 kHz audio much better than what you suggest:

My scope doesn't have picosecond resolution, and even if it did, noise in the DAC would make it impossible to measure such small differences.

5 hours ago, PeterV said:

Nowadays, such anomalies can be overcome by using oversampling techniques or by using another format like DSD. But it is a fact that all digital recordings in the 80's and 90's intrinsically are degraded by these influences and even nowadays 24/96 recordings are, since filtering can severely degrade the temporal behaviour of audio systems.

 

So..let's focus much more to this beautiful aspect of the new algorithm.. MQA is capable to repair the temporal blur of PCM recordings to a very large extend. It is therefore capable to restore the original sound of the Recording and not just that what has been stored on the final Mastertape. MQA enhances the temporal decay of the recording significantly and this is audible to everyone.

 

Every serious audiophile should first listen to MQA and make up their mind based on what they hear, feel and experience instead of judging the technology beforehand, based on speculative publications.

 

MQA is for sure not a fraud or just a pseudo MP3 compression tecgnology. It is a paradigm shift and a new, disruptive technology.

 

This is the reason why not everyone is happy with its existence and fear is firing up rumours and aligations. That is a shame, since it will be much wiser to embrace this stunning innovation and make use of it.

How much are MQA paying you to post this?

[miguelito]

1 hour ago, mansr said:

That statement cost you whatever credibility you might have had.

MQA troll... Thanks for setting it straight.

[Samuel T Cogley]

8 hours ago, PeterV said:

This is the reason why not everyone is happy with its existence and fear is firing up rumours and aligations. That is a shame, since it will be much wiser to embrace this stunning innovation and make use of it.

 

Bob, is that you?  You'd do better over at one of the Enthusiast Network sites.  Your sock puppet kind of sticks out like a sore thumb here...

 

[labjr]

1 hour ago, miguelito said:

MQA troll... Thanks for setting it straight.

I think he's just an MQA fan who's a bit overzealous. He's also on a couple other forums. And apparently English is not his first language. 

 

But I don't know why anyone would disregard the opinion of Andreas Koch. He's a very brilliant man.

[Jud]

2 hours ago, mansr said:

That statement cost you whatever credibility you might have had.

Normal microphones have practically no response above 50 kHz or so. With a sufficiently high sample rate, you thus don't need much of an analogue filter at all. 192 kHz is plenty, and even 96 kHz is enough in most cases. Of course, modern ADCs are sigma-delta designs operating at several MHz followed by a digital low-pass filter. The problems you allege simply do not exist.

Oh, no. Not this again. CD has a temporal resolution of a few picoseconds. We discussed this at length not long ago. Here's a post from that discussion where I demonstrate timing accuracy of 44.1 kHz audio much better than what you suggest:

My scope doesn't have picosecond resolution, and even if it did, noise in the DAC would make it impossible to measure such small differences.

How much are MQA paying you to post this?

 

I don't understand the predilection to personally attack someone who read some MQA marketing, heard something he liked, and consequently believed the marketing.  

[mansr]

6 minutes ago, Jud said:

I don't understand the predilection to personally attack someone who read some MQA marketing, heard something he liked, and consequently believed the marketing.  

I see no problem with pointing out the blatantly wrong.

[Samuel T Cogley]

4 minutes ago, Jud said:

 

I don't understand the predilection to personally attack someone who read some MQA marketing, heard something he liked, and consequently believed the marketing.  

 

You're a very trusting, "glass half full" kinda guy.  Perhaps even a de facto "CA Ambassador". Based on the post count of that user, and the utterly over the top MQA sycophancy, there's a significant chance that post was done at the behest of MQA.  I suspect you won't agree with that assessment, but that doesn't mean it isn't a possibility.

[Jud]

6 minutes ago, mansr said:

I see no problem with pointing out the blatantly wrong.

 

That isn't all you did, as you're quite aware.  Hmm, and you're the one who just told someone *else* his credibility was suffering.

[mansr]

24 minutes ago, Jud said:

That isn't all you did, as you're quite aware.  Hmm, and you're the one who just told someone *else* his credibility was suffering.

When people repeat that old fallacy about the temporal resolution of sampled signals, it casts doubt on everything else they say. I see no irony in pointing this out.

 

Perhaps I accused him of being a paid shill too hastily. His post sure made him sound like one, but maybe he's merely unusually gullible.

[Samuel T Cogley]

BTW, "online reputation management" is a thing.  This isn't a paranoid fantasy.

[PeterV]

Well.. I am again surprised.., strange question 'Am I being paid by MQA..' would be nice, but no, not at all. I am not a troll either, but just a critical audiophile, who is not easily convinced.

 

You all notice that I have not posted much and indeed I am Dutch, so English is not my first language, apologies for this.

 

But with regard to MQA and what it brings in my audio chain at home and how the technology works, I am a bit of a truth digger and want to know and understand why it works so convincingly well in my system with so many sorts of music. I am convinced by what I hear and when listening to MQA compared with normal or HD flac files, MQA is simply better, more natural every time I hear it.

 

From most of your responses, no one seems to believe that temporal blur in AIR seems to be measurable and audible. All I see are graphs of measurements of DAC with or without MQA decoders. What MQA is doing is related to the impulse response in AIR as shown in Stereophile and AES meetings:

 

http://www.stereophile.com/content/mqa-questions-and-answers-some-real-world-comparisons#JTt3j2x2Xja08jVy.97

 

And what is also nice: anyone who does not like MP3 or MQA, just do not consume it and be happy with what you have.

 

 

 

[audiventory]

26 minutes ago, PeterV said:

And what is also nice: anyone who does not like MP3 or MQA, just do not consume it and be happy with what you have.

 

Me seems, some people beware, that part of audiophile records will in MQA only.

[labjr]

24 minutes ago, PeterV said:

Well.. I am again surprised.., strange question 'Am I being paid by MQA..' would be nice, but no, not at all. I am not a troll either, but just a critical audiophile, who is not easily convinced.

 

You all notice that I have not posted much and indeed I am Dutch, so English is not my first language, apologies for this.

 

But with regard to MQA and what it brings in my audio chain at home and how the technology works, I am a bit of a truth digger and want to know and understand why it works so convincingly well in my system with so many sorts of music. I am convinced by what I hear and when listening to MQA compared with normal or HD flac files, MQA is simply better, more natural every time I hear it.

 

From most of your responses, no one seems to believe that temporal blur in AIR seems to be measurable and audible. All I see are graphs of measurements of DAC with or without MQA decoders. What MQA is doing is related to the impulse response in AIR as shown in Stereophile and AES meetings:

 

http://www.stereophile.com/content/mqa-questions-and-answers-some-real-world-comparisons#JTt3j2x2Xja08jVy.97

 

And what is also nice: anyone who does not like MP3 or MQA, just do not consume it and be happy with what you have.

There are many A-D and D-A conversions and multiple converters used in some recordings.. It would seem impossible for deblurring to compensate for all of it. Everything looks good on paper. Especially when it's written by MQA people. The proof is in the listening. 

 

Stereophile Editor, John Atkinson has said he couldn't reliably tell the difference between MQA and non-MQA versions of his own recordings. Thus it would seem that well done recording and mastering is far more important than MQA. MQA has likely cherry-picked the material for their demos which will favor their process. 

 

24 minutes ago, PeterV said: