• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

MQA vs. Non-MQA, Empirical Measurements

solderdude

Master Contributor
Joined
Jul 21, 2018
Messages
9,803
Likes
21,208
Location
The Neverlands
I'm not sold on the idea of a lossy format, I still prefer reel-reel if everything is properly maintained and recorded.

Reel-to-reel is a lossy format. Just do 5 generations and you'll clearly hear it as well.
Compare what goes in and comes out and you'll be horrified of the differences. Those differences are losses. That doesn't mean the format cannot sound pleasant.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
2,038
Likes
2,122
Location
Pacific Northwest
Can you quantify what would constitute an improvement over CD quality?
Two things come to mind:

1. 24 bit, since while 16-bit is sufficient for most music, it is not perceptually transparent. Some very wide dynamic range sounds can bump into the limitations where dither noise becomes audible. 24 bit would be digitally transparent as it drops dither below the noise of even the best analog signal paths.

2. Higher sampling frequency, to make proper transparent filters easier to implement in real-time on hardware having limited processing power (e.g. typical DAC chips). With 44.1 kHz, the filter transition band is so narrow (20k to 22,050) that most DAC chips cheat and double the width, stretching past Nyquist to 24,100. You can see this in many of the measurements Amir makes here. If we used even a slightly higher sampling frequency, say 48 kHz, this kludge would be unnecessary.

These two improvements are fairly modest, easy to implement, and should comprise a perceptually transparent standard that is easy to implement while still being practical to use. Indeed, isn't this already the standard for DVD? So it's already a standard, we should simply use it for audio too.
 
OP
Tokyo_John

Tokyo_John

Active Member
Forum Donor
Joined
Mar 6, 2021
Messages
146
Likes
175
Can you quantify what would constitute an improvement over CD quality?

We can surely measure differences. Whether the difference constitutes an "improvement" requires an objective basis for comparison, such as playback from the original studio master. This can only be done in cases where such a master is available...but yes, it is certainly quantifiable.

Even in the absence of a studio master, we can detect the presence of certain characteristics that are more or less desirable, particularly if there is a large cohort for comparison. For example, bleed over of high frequency noise-shaping down into audible frequencies can be measured and detected. Evidence of diffusion, interpolation artifacts, etc., can also be detected. Any claimed effect involving any particular playback chain for a particular format can be measured and compared.
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
2,049
Likes
2,955
Location
Netherlands
We can surely measure differences. Whether the difference constitutes an "improvement" requires an objective basis for comparison, such as playback from the original studio master. This can only be done in cases where such a master is available...but yes, it is certainly quantifiable.

Okay, so your reference is the original unaltered (and hopefully) high-res studio master. And see how close to it both get? What to do with the spectrum above 22 kHz? Does MQA automatically win that one because there is something?

Even in the absence of a studio master, we can detect the presence of certain characteristics that are more or less desirable, particularly if there is a large cohort for comparison. For example, bleed over of high frequency noise-shaping down into audible frequencies can be measured and detected. Evidence of diffusion, interpolation artifacts, etc., can also be detected. Any claimed effect involving any particular playback chain for a particular format can be measured and compared.

I think most of this is already covered by multiple people. What do you think you can add or do differently?

If we used even a slightly higher sampling frequency, say 48 kHz, this kludge would be unnecessary.

Indeed, isn't this already the standard for DVD? So it's already a standard, we should simply use it for audio too.
I’m totally with you there. 24/48 flac would probably more than enough to get full transparency. It would also invalidate the usefulness of MQA.. but I realize that that is not part of the discussion right now ;). Oh, and shame on AKM! Not that MQA’s interpolation filters are free from aliasing :facepalm:
 
OP
Tokyo_John

Tokyo_John

Active Member
Forum Donor
Joined
Mar 6, 2021
Messages
146
Likes
175
Okay, so your reference is the original unaltered (and hopefully) high-res studio master. And see how close to it both get? What to do with the spectrum above 22 kHz? Does MQA automatically win that one because there is something?

Do you have another objective for comparison, besides the original master from which everything else is an (imperfect) copy?

Unless it comes out of my speaker, I don’t care about >22 kHz. I think my roll-off filters take care of most of that, but I can measure it to see if it is the case.

I think most of this is already covered by multiple people. What do you think you can add or do differently?

If you know of good information about people actually recording what comes out of their speakers, I’d certainly enjoy seeing it. So far the only thing I see are hand-wavy arguments about what people “think” something does, like this...

I’m totally with you there. 24/48 flac would probably more than enough to get full transparency. ...

...totally a speculation. Let’s see the data.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,004
Likes
130,463
Location
Seattle Area
...totally a speculation. Let’s see the data.
Ironically, the most definitive paper on this was published by Bob Stuart in the Journal of AES (peer reviewed):

Coding for High-Resolution Audio Systems*
J. ROBERT STUART, AES Fellow

On the need to go beyond 16 bits:

1620878638317.png


"This analysis of the dynamic-range capability of the 16-
bit 44.1-kHz channel makes it very clear that it cannot be
considered transparent. Even in the absence of quantization
distortion introduced by defective processing, the
benign noise introduced by quantization and dither is
audible from modest acoustic gains (around 100 dB SPL)."

1620878883888.png


As you see, 20 bits/120 dB is a bit more than what we need and hence you always hearing from me that -115 dB is the threshold we should aspire as far as full transparency.

His arguments for sample rate is softer and requires buying into a bit more ultrasonics being audible (at very high amplitudes) and room for noise shaping:

1620879450433.png
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
2,049
Likes
2,955
Location
Netherlands
View attachment 129481

"This analysis of the dynamic-range capability of the 16-
bit 44.1-kHz channel makes it very clear that it cannot be
considered transparent. Even in the absence of quantization
distortion introduced by defective processing, the
benign noise introduced by quantization and dither is
audible from modest acoustic gains (around 100 dB SPL)."

Actually all we need is better dither. TPDF is not noise shaped. Others can do better:

1620882144599.gif

And as we know, MQA relies on this as well.
 
Last edited:

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
2,049
Likes
2,955
Location
Netherlands
Do you have another objective for comparison, besides the original master from which everything else is an (imperfect) copy?

No, I think that is a good way, if you can confirm all of your versions actually come from that control. If not, all bets are off.

If you know of good information about people actually recording what comes out of their speakers, I’d certainly enjoy seeing it. So far the only thing I see are hand-wavy arguments about what people “think” something does, like this...
There is much more out there than just hand waving. If anyone is doing that, it’s mostly MQA itself, but that is not supposed to be the discussion. Nobody compared what comes out of a speaker though. In that way, your idea is definitely novel and interesting. But it also has quite a lot of variables that might influence the outcome. You’ll need to verify a few things first. For instance: how good does your amp handle ultrasonics? Some get unstable or have bad IMD products in the audible band. Also good to measure the frequency response of the speakers to use as a reference. There are probably more things to consider.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
35,004
Likes
130,463
Location
Seattle Area
Actually all we need is better dither. TPDF is not noise shaped. Others can do better:
Dither is not under your control. Who knows what dither content producer used. They may use nothing, rectangular, TPDF or noise shaped. I suspect many think dither=noise and hence bad. The concept of shaped dither is even more foreign to people so likelihood of it being used in content you play is very low.

This is why I like to get the high-res master. I don't want to leave it to chance as to what dither, if any, is used in the content I consume.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
2,038
Likes
2,122
Location
Pacific Northwest
...This is why I like to get the high-res master. I don't want to leave it to chance as to what dither, if any, is used in the content I consume.
I agree. And yet another reason to get the digital master is because I don't want to leave it to chance what resampling method, if any, they used when preparing the content for CD, streaming, or whatever other playback method one uses. Some resampling methods are transparent, but others are not, and who knows what they used.
 

Gradius

Senior Member
Joined
Aug 17, 2019
Messages
384
Likes
238
Location
Iquique, Chile
The true is, we don't need anything beyond 96kHz @ 24-bit. DSD is around 88.2kHz @ 24-bit (into PCM comparable conversion).

And FLAC as a true lossless (and open source) format, why we would need more formats? We just DON'T !

Besides, FLAC can support 1 to 8 channels! Isn't limited to stereo only.
 
OP
Tokyo_John

Tokyo_John

Active Member
Forum Donor
Joined
Mar 6, 2021
Messages
146
Likes
175
I still haven't found a few minutes to work on this, but I'm still going to do it. I'm convinced more than ever that the sound that comes out of my speakers is the only thing that matters. I don't care what MQA does above 20kHz, or what digital anomalies exist in a file, if it doesn't change the pressure waves generated in my room.

Noise, microphone anomalies, etc., I already discussed previously. In the sciences I have seen differences between data exploited to produce very sensitive and useful measurements, even when the raw data have noise, limited dynamic range, etc.. Interferometry is a great example.

More to come....
 

MaxBuck

Senior Member
Forum Donor
Joined
May 22, 2021
Messages
496
Likes
546
Location
SoCal, Baby!
I love music. And the MQA tracks I've listened to on my test-drive of Tidal have been superb. Much preferable (admittedly not blind, not by a trained listener) to the same tracks on Tidal and Qobuz "CD-quality." Testing was primarily Sting's The Last Ship, in the event anyone cares.
 
Top Bottom