• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

MQA Deep Dive - I published music on tidal to test MQA

Status
Not open for further replies.

Don Hills

Addicted to Fun and Learning
Joined
Mar 1, 2016
Messages
708
Likes
464
Location
Wellington, New Zealand
GoldenOne posted somewhere the multiple error messages that he was sent when he queried why his original track wasn't processed. I can't find them. Can anyone please point me to them? I think there may be useful insight to be gained into the workings of the encoder by analysing the errors.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
Wait, what? What is the point of this site again? Because I sure as hell thought it was objective measurement of the quality of audio equipment to weed out snakeoil in the audio industry. Should I just go start reading $2000 power cable reviews if subjectivity is what matters, not objective measurements of marketing claims? Because that is the exact line that power cable selling scam artists use to justify their products
What do you mean "wait what?" You asked me about "Quality" in their name and I said that is a subjective thing with respect to lossy encoding. Lossy encoders have "state" which electronics does not do. What comes in a sample of audio depends on what comes before it. As such, you can't subject them to pathological test signals and expect them to work as well as any other. Audio gear does not have this so we we can test them with signals. A power cable doesn't know the difference between sine wave, noise or music. A lossy compressor hugely cares.

There is not one research paper on lossy encoding that uses test signals. All testing is done with music. To push the codec, "codec killer" test tracks are used but they are still not artificial test signals. My team developed such technology so you can't ask me to go along with a different metric of testing them against 100% of research and industry practice.

There is an entire forum dedicated to codecs called hydrogen audio. Here is one of their terms of service [bolding mine]: https://hydrogenaud.io/index.php?topic=3974

"TOS 8. All members that put forth a statement concerning subjective sound quality, must -- to the best of their ability -- provide objective support for their claims. Acceptable means of support are double blind listening tests (ABX or ABC/HR) demonstrating that the member can discern a difference perceptually, together with a test sample to allow others to reproduce their findings. Graphs, non-blind listening tests, waveform difference comparisons, and so on, are not acceptable means of providing support. "

I came within an inch of being banned due to posting measurements to prove my point!

Here is a random article from my article on codec testing from EBU (European standards setting organization): EBU Evaluations of Multichannel Audio Codecs

1618959340138.png


There is no test signals in there.

Here is another in peer reviewed journal of AES:

1618959622649.png



1618959553655.png



There is actually one paper on such controlled testing with MQA:
A Comparison of Clarity in MQA Encoded Files vs. Their
Unprocessed State as Performed by Three Groups – Expert
Listeners, Musicians, and Casual Listeners


Mariane Generale, Richard King, and Denis Martin
Graduate Program in Sound Recording, McGill University, Montréal, Canada
Centre for Interdisciplinary Research in Music Media and Technology (CIRMMT), McGill University, Montréal, Canada

Here is the content they used:

1618959760700.png


The study results were inconclusive by the way (and somewhat ill conceived by asking listeners to rate "clarity" of music).

So folks, don't argue with me on this stuff. :) I am explaining how things are done in this domain. Any test subscribing to how research is done should be with listening tests.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
DMS (from Abyss headphones) seems to imply that Amir has changed his tune on this topic because the MQA bashing negatively impacts the companies that (allegedly) pay him.
I have always held a neutral position on MQA. And gotten a lot of flak for it here and elsewhere. For some reason people want me to join them in this fight but I am not the right soldier. I have my hands full doing my reviews and defending them. Don't need another battle to fight.

As for DMS, I suggest he spends time learning about signal processing, measurements, etc. rather than making another comment without knowledge.
 

jensgk

Active Member
Forum Donor
Joined
Mar 21, 2020
Messages
256
Likes
564
Location
Denmark
What do you mean "wait what?" You asked me about "Quality" in their name and I said that is a subjective thing with respect to lossy encoding. Lossy encoders have "state" which electronics does not do. What comes in a sample of audio depends on what comes before it. As such, you can't subject them to pathological test signals and expect them to work as well as any other. Audio gear does not have this so we we can test them with signals. A power cable doesn't know the difference between sine wave, noise or music. A lossy compressor hugely cares.

There is not one research paper on lossy encoding that uses test signals. All testing is done with music. To push the codec, "codec killer" test tracks are used but they are still not artificial test signals. My team developed such technology so you can't ask me to go along with a different metric of testing them against 100% of research and industry practice.

So is it ok to test bluetooth equipment with test signales?
E.g. https://www.audiosciencereview.com/...th-receiver-and-bt-codec-quality-video.21084/
 

mkawa

Addicted to Fun and Learning
Joined
Sep 17, 2019
Messages
788
Likes
695
you mean for the 2 (3?) products he's tested, one via the wired connection (qudelix) and the Kali receiver that he literally notes in the second paragraph that because of lossy compression it's not really fair to test with standard signals?
 

Raindog123

Major Contributor
Joined
Oct 23, 2020
Messages
1,599
Likes
3,554
Location
Melbourne, FL, USA
I tried to explain the reason in previous posts: at higher frequencies music has much lower amplitudes (Meridian shows this in sets of statistic analysis of real music, although the same could be inferred from the amplitude patterns of harmonics of every instrument)...

I’ll try for the third time... Can you please point me/us at where Meridian shows this statistic analysis of real music?

I've posted this. Maybe my understanding is incorrect, I would appreciate help getting to the bottom of it... Thanks!
 
Last edited:

11Parsecs

Member
Joined
Apr 20, 2021
Messages
5
Likes
2

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
So is it ok to test bluetooth equipment with test signales?
E.g. https://www.audiosciencereview.com/...th-receiver-and-bt-codec-quality-video.21084/
I was wondering how long it was going to be before someone brought that up. :)

When I ran those tests, I expected all the codecs to perform the same (producing the sine wave perfectly), then forcing me to conduct listening tests. Alas, there were some basic issues that the measurements showed so I shared that as an easy way out. :) Do note that I used a pure sine wave that I said was fine to run. I did not throw noise, ultrasonics, etc. at the codecs that I knew would be problematic.

To expand, when I took over the codec development team at Microsoft, I ran a frequency response error which showed issues. The team found the source of all of them and fixed them. The overall quality did not change however. I just thought it was good hygiene that we pass a series of sine waves accurately as they are dead simple to encode. My team agreed and hence the fixes.
 

dmac6419

Major Contributor
Joined
Feb 16, 2019
Messages
1,246
Likes
770
Location
USofA
I liked local stations when I was on the road.

Didn't get the same canned stuff from two stations a thousand miles apart.
What are you saying? ,you have a problem with satellite radio.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
’ll try for the third time... Can you please point me/us at where Meridian shows this statistics analysis?
Some of the data is in the peer reviewed AES paper I cited earlier:

The audibility of typical digital audio Filters in a high-fidelity playback system

Helen M. Jackson, Michael D. Capp, and J. Robert Stuart1
1Meridian Audio Ltd, Huntingdon, Cambridgeshire, PE29 6YE, England
Correspondence should be addressed to Michael Capp ([email protected])

Winner of the AES 137th Convention Best Peer-Reviewed Paper Award

1618967091498.png
 

NTK

Major Contributor
Forum Donor
Joined
Aug 11, 2019
Messages
2,656
Likes
5,819
Location
US East
Of course. I have not said otherwise and neither did Keith Howard. As you write, the single high sample is "mapping" the coefficients of the filter. In my 2018 article I showed how three A/D converters with different antialiasing filters coped with a shaped analog impulse that had content up to 60kHz and rolled off above that frequency.

John Atkinson
Technical Editor, Stereophile
Then I am puzzled as to why Keith wrote "Although the input signal was only one sample wide, in the recorded version the impulse energy has been smeared over many samples due to the ringing behavior of the anti-alias filter."

A properly sampled impulse (i.e. anti-alias filtered) will NOT be 1 sample wide.

The anti-alias filter is upstream of the A/D conversion, and since it is an analog filter, it is causal and impossible to cause "pre-ringing" as shown in Fig 1. Attached is what an ideal impulse (and its digital samples) would look like after anti-alias filtered. The filter spec is elliptic, pass band frequency and ripple, 20 kHz & 0.001 dB, stop band frequency and attenuation, 24.1 kHz & 80 dB, to ensure that any contents above Nyquist aliased back into the audible band of 20 — 20k Hz are attenuated by 80 dB.

anti-alias.PNG

What Keith showed was the D/A reconstruction of a synthesized signal that did not represent real life (i.e. sampled in accordance with the requirements of the sampling theorem).
 
Last edited:

Grooved

Addicted to Fun and Learning
Joined
Feb 26, 2021
Messages
679
Likes
441
Thanks for the tip.

It was the first thing to jump out at me, how much louder Qobuz was out the gate.

https://www.sageaudio.com/blog/mast...form-loudness-and-normalization-explained.php

This link includes a video with lots of errors :
they said Youtube uses AAC while they have more used OGG 165 since some years.
For Tidal, they don't know the Low profile (it's AAC 96kbps), they are right on High (AAC 320), wrong on HiFi as they said FLAC 24/44.1 (it's only 16bit, and it starts to be 16bit MQA instead of FLAC), and partially wrong on Master saying it's all 24/96

On LFUS, they said it's at least -14LUFS for Tidal while I tested a track on Tidal and Qobuz to compare, and both were lower than that, with huge room on peak too.
These are the results on "Bayou" track from Thomas Strønen that someone talked about in the first pages
All files from Roon output, main differences are on Min. RMS

FLAC 16-44.1 Qobuz
FLAC 16-44.1 Qobuz.PNG


FLAC 16-44.1 Tidal (real FLAC for this one, not a 16bit MQA)
FLAC 16-44.1 Tidal.PNG


FLAC 24-48 folded Tidal
FLAC 24-48 folded Tidal.PNG


FLAC 24-96 Qobuz
FLAC 24-96 Qobuz.PNG


MQA 24-48 Tidal Unfolded via Roon to 24-96
MQA 24-48 Tidal Unfolded via Roon 24-96.PNG
 
Last edited:

Raindog123

Major Contributor
Joined
Oct 23, 2020
Messages
1,599
Likes
3,554
Location
Melbourne, FL, USA
Some of the data is in the peer reviewed AES paper I cited earlier:

The audibility of typical digital audio Filters in a high-fidelity playback system

Thanks Amir. Unfortunately, I do not have access to the AES library even through work, so still am trying to cheap-skate - sent a couple of emails to my buddies at various places. And if not, will bite the bullet and pay the $33 (2x)... :)

But now I am rather confused and puzzled, as the caption you just posted mentions a “test signal”, while @mieswall talks about “analysis of real music”. With the real music and its dynamism (thus instantaneous spectrum variance over its duration, and whether it matters) being the point of my earlier posting. Agin, I would like to understand whether/how a musical piece in its entirety can be analyzed to achieve this “frequency-selective dynamic range reduction” @mieswall keeps on talking about.
 

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
I have always held a neutral position on MQA. And gotten a lot of flak for it here and elsewhere. For some reason people want me to join them in this fight but I am not the right soldier. I have my hands full doing my reviews and defending them. Don't need another battle to fight.

As for DMS, I suggest he spends time learning about signal processing, measurements, etc. rather than making another comment without knowledge.

I just need clarification on whether my memory betrays me or not.

There was a place, some forum or something, where I recall you, and perhaps mansr being present long ago (think 2017?). I could have sworn you were tearing ass up there against MQA proponents. Now I can't remember, but I'd wager that Bob was there as well.

Did this not happen?

Also when you say neutral position. Neutral position of what? The soundness of the claims MQA purports about the product itself (like claiming they have access and provide the true masters of all the audio they offer to consumers, claiming they can preserve artist intent, claiming MQA is lossless, claiming it's the original sound as heard by the people during production present, claiming they have profiled every ADC and that somehow their encoder accounts for all of them for some sort of correction processing)?

Are those the things you remain neutral on, or is it simply the claim that MQA can potentially deliver on audibly transparent recorded music?

If it's the latter, then sure, count me there with you. If it's any of the former, I would question what makes you neutral considering I remember you going in and destroying people on that forum about most of these claims.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
Also when you say neutral position. Neutral position of what?
It means this:

1. I use MQA for free on Tidal. It was added to my subscription at no cost to me.

2. I have DACs and players that decode it. Definitely did not pay for it in my player (Roon).

3. I have no issue with someone/MQA trying to build a format and charge people to use it. MP3, AAC, WMA, Dolby, DTS, MPEG, etc. all have license fees (some with very high dollars). The few that were not likely infringed on someone's patent and left it to the implementer to deal with the lawsuit. Folks are paying probably $10 to $20 to license Blu-ray format to build a player and that hasn't cause riots in streets. And companies like Netflix and such are getting sued for millions if not billions of dollars over codec/DRM IP. It is just the way this field works.

4. I see no fear of MQA taking over the world and predicted that Apple and Amazon's of the world would never pay a fee to use it. Amazon has since come out with high-res service with no MQA proving that MQA had no leverage over them, or the the labels.

5. I consider Bob Stuart a professional colleague with major contributions to audio research (AES Fellow, highly reference journal papers, etc.). I would need really good reasons to throw arrows at him.

I realize some of you are motivated differently and that is fine. Just don't ask me to shed blood with you. I have a review to get out. :)
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
There was a place, some forum or something, where I recall you, and perhaps mansr being present long ago (think 2017?). I could have sworn you were tearing ass up there against MQA proponents.
Mansr and I were on the opposite sides of that argument in this forum. He was against, I was neutral but arguing against things that I thought were not correct. I never fought for or against MQA itself. My argument is with the details provided to say something is wrong.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,384
Location
Seattle Area
But now I am rather confused and puzzled, as the caption you just posted mentions a “test signal”, while @mieswall talks about “analysis of real music”.
By "signal" in the paper they mean music. You can see the spectrum of it and it definite not a test signal. It is a common terminology but admittedly, confusing in our context in this discussion.
 

mieswall

Member
Joined
Nov 10, 2019
Messages
63
Likes
106
I’ll try for the third time... Can you please point me/us at where Meridian shows this statistic analysis of real music?

I've posted this. Maybe my understanding is incorrect, I would appreciate help getting to the bottom of it... Thanks!
May search more detailed info other time. For now: https://www.aes.org/tmpFiles/elib/20210417/20456.pdf. Chapter 3.4 and Fig 7, Fig 10. Before your answer: these in particular are not statistics summaries, but you will get the idea.
 

dmac6419

Major Contributor
Joined
Feb 16, 2019
Messages
1,246
Likes
770
Location
USofA
It means this:

1. I use MQA for free on Tidal. It was added to my subscription at no cost to me.

2. I have DACs and players that decode it. Definitely did not pay for it in my player (Roon).

3. I have no issue with someone/MQA trying to build a format and charge people to use it. MP3, AAC, WMA, Dolby, DTS, MPEG, etc. all have license fees (some with very high dollars). The few that were not likely infringed on someone's patent and left it to the implementer to deal with the lawsuit. Folks are paying probably $10 to $20 to license Blu-ray format to build a player and that hasn't cause riots in streets. And companies like Netflix and such are getting sued for millions if not billions of dollars over codec/DRM IP. It is just the way this field works.

4. I see no fear of MQA taking over the world and predicted that Apple and Amazon's of the world would never pay a fee to use it. Amazon has since come out with high-res service with no MQA proving that MQA had no leverage over them, or the the labels.

5. I consider Bob Stuart a professional colleague with major contributions to audio research (AES Fellow, highly reference journal papers, etc.). I would need really good reasons to throw arrows at him.

I realize some of you are motivated differently and that is fine. Just don't ask me to shed blood with you. I have a review to get out. :)
But ain't you a God Amirm?,oh ok I thought so,P.S. don't answer that.
 
Status
Not open for further replies.
Top Bottom