• 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.

Audibility thresholds of amp and DAC measurements

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
In short - they can't discover which DUT is better, but only - which DUT is closer to reference.
So closer to the reference is not better performance ?
One could argue closer to the reference is not preffered by some but that is taste/preference not actual fidelity.
We need to establish what is meant with 'better'.
Does 'better' mean closer to the original signal waveform or does 'better' mean preferred by someone/a lot of people ?
We may be talking about different 'betters' and different 'fidelity'.

Audibility differs from person to person and from recording to recording.
You cannot pin a number on it.
I applaud the efforts to try and find a single 'fidelity' number but one cannot put taste in it.

When looking at the vast differences in tonal balance (linear distortion) of headphones for instance we can see that some folks like bass shy headphones yet others find very dark and bassy headphones (like Nighthawk etc.) to be neutral and have the most realistic headphone.

Of course this has been investigated and the result is known. However, this too is an 'average', a compromise. Based on science of course.
But it is known that trained listeners have a different preference than less trained people.

I really do want to see a metric that shows a good relation to perceived SQ that works for everyone.
 
Joined
Dec 7, 2019
Messages
207
Likes
84
Does 'better' mean closer to the original signal waveform or does 'better' mean preferred by someone/a lot of people ?
We may be talking about different 'betters' and different 'fidelity'.
... and that talking can be endless because it's about tastes. That is why for me the answer is clear - the better means closer to the original. The measurements of distortion audibility on this forum, my measurements of portable players and theoretical estimations of the thresholds (s-levels) clearly show that it is not hard to find the level of difference, which will be transparent for all. For portable players it is around -50dB (Hugo 2 has -34dB but probably lower/better because Babyface pro has comparable df levels in loopback). Really it is not a rocket science.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
Then at least we talk about the same 'better'.

The thing is that when we are nulling there will be differences because of phase differences and amplitude differences (freq. response deviations from truly flat).
These, however may not be that audible at the extremes of the audible range yet will give an imperfect null and thus a lower rating while it may not be sound degrading or even audible.
One needs to account for this (audibility of phase shifts for instance). The non linearities in the frequency response (not wiggles due to the filters used but sheer drop-off) could well be compensated for in the device with some tone control and not be audible.
 
Last edited:

Wombat

Major Contributor
Joined
Nov 5, 2017
Messages
6,037
Likes
4,949
Location
Australia
Take your pick: better

Such words need to be used with clear context in forum posts.
 
Joined
Dec 7, 2019
Messages
207
Likes
84
These, however may not be that audible at the extremes of the audible range yet will give an imperfect null.
One needs to account for this (audibility of phase shifts for instance). The non linearities in the frequency response (not wiggles due to the filters used but sheer drop-off) could well be compensated for in the device with some tone control and not be audible.
I think there is no need to research this - just eliminate all of them regardless of their nature/audibility. It is easier from engineering point of view. Or may be you think that it is too hard to design a circuit that will have at the output exactly similar waveform? (Linear pitch and phase shift of the output signal are not accounted already).
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
I think in (audio) engineering there are compromises that have to be and are made.
Some of which have less audible (sound degrading) consequences then others but all will show up in nulling.

The biggest culprit in sound quality changes of course are transducers.
That's why I think splitting certain aspects up in multiple measurements makes sense.
One needs to understand all the measurements and know their own audible limits for these aspects to estimate if this could mean poorer fidelity or not.
A null alone doesn't do that but needs to be analyzed (which is what you do).
I have no idea whet the analysis takes or does not take into account.
I do know when looking at various plots but that may be just me and does not account for personal taste of myself or others.

I applaud any attempt to measure stuff instead of assuming ears can do a better job.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
I think most of the forumers will agree that "better" means "closer to the original/reference"
I agree but also know there are a lot of readers and some forum members that think 'better' means sounds better to them.
This can cause confusion so I do think it is a good thing to explain the word 'better' when used.
Better sounding as in higher fidelity (closer to the original signal) or... better sounding to (fill in persons name) where personal preference and subjective 'test' methods are used.
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
772
Likes
1,288
Location
My dining room.
Quite the opposite - the essence. When the output of a DUT is corrupted - it is corrupted, or you need to insert some DSP between amplifier and headphones/speakers. For a listener this corruption is final/irreversible.

What is this all about? I said you must convolve the all passed signal with the reversed impulse response, truncated below the noise level. Apparently you agree when you say "insert DSP" while you make a sham disagreement. Do you even know what "convolve" means?

Yes it's fair game to put a DSP in the chain to undo it. WHICH IS WHAT I SAID TO DO. At this point I must conclude that you are a hostile troll who cares more about muddying the waters than solving any problem. EITHER you have no idea what I just said, or you're just trying to insult people who DO understand what's going on. And, let us note I referred ONLY to a linear operation. So no games about anything other than linear, sport.
 
Last edited:

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
772
Likes
1,288
Location
My dining room.
You may have a different view of the objective measurements, please, share it. We should have a common ground for the discussion of audibility of distortions and corresponding thresholds.
What part of compare error spectrum to masking spectrum, on a time-varying, signal=dependent basis was hard to understand>

There is no magic number beyond total error from all mechanism being below absolute threshold, and for that, you ***must*** know the spectrum of the error and the final acoustic system gain.

You are now trying to muddy the waters again and you're simply trying to fool people into the idea you're saying something reasonable
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
772
Likes
1,288
Location
My dining room.
I think most of the forumers will agree that "better" means "closer to the original/reference"
Which "original reference"? Original soundfield? PCM recording? LP? Tape? What **IS** your "original reference"? I suspect some of your difficulty is the assumption you can define an "original reference".
 
Joined
Dec 7, 2019
Messages
207
Likes
84
Yes it's fair game to put a DSP in the chain to undo it. WHICH IS WHAT I SAID TO DO.
Nobody will do this. If you have a DUT, you just use it, not tweak.
and for that, you ***must*** know the spectrum of the error and the final acoustic system gain.
My point is that below some low level of difference signal 'the spectrum of the error' doesn't matter. And "final acoustic system gain" along with noise floor of the listening environment define that low level (#177).
 
Joined
Dec 7, 2019
Messages
207
Likes
84
Which "original reference"? Original soundfield? PCM recording? LP? Tape? What **IS** your "original reference"? I suspect some of your difficulty is the assumption you can define an "original reference".
"Original reference" is the waveform produced by author of music piece in the form of file.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
Yes, that's what the audio business (above all cable and snake oil business) is all about.
There are some well designed affordable and super expensive (and everything in between) devices around that do value quality and are in the business for offering quality stuff but IMO are a small minority.

Look how some of the uber expensive horrors will measure.
I am quite certain most of these are rated very highly in sound quality but would score really bad when nulled.
This means it would (very likely) also score poorly on your software yet some folks will say it is the best sounding equipment they ever heard.
Bye bye correlation to perceived SQ unless the software can discriminate (or learn) what might sound good or not.
There would be no correlation to measurements of any kind.
 
Joined
Dec 7, 2019
Messages
207
Likes
84
I am quite certain most of these are rated very highly in sound quality but would score really bad when nulled.
I'm quite certain in the opposite, having measured dozens of portable players, codecs and audio interfaces. The correlation of time-adjusted null to SQ is remarkable:
- early df measurements of portable players - http://soundexpert.org/portable-players-beta
- portable players measured with different sound material - http://soundexpert.org/portable-players
- recent df measurements with different audio interface - http://soundexpert.org/articles/-/blogs/audio-quality-of-high-end-portable-players
- df measurements of Bluetooth codecs - http://soundexpert.org/articles/-/blogs/audio-quality-of-sbc-xq-bluetooth-audio-codec
- direct comparison of df levels with SQ in listening tests - https://www.audiosciencereview.com/...od-for-measuring-distortion.10282/post-285578

Df measurements fail only in cases when artifact signatures of DUTs are different (>1.5 - 2.0dB)
 
Last edited:

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
6,356
Likes
11,777
Location
The Neverlands
Have you tried tube amps yet ? or Pass Labs amplifiers versus say the AHB2 ?

Do the measurements correlate with blind level matched tests with a few participants ?
 
Joined
Dec 7, 2019
Messages
207
Likes
84
Have you tried tube amps yet ? or Pass Labs amplifiers versus say the AHB2 ?
No, I don't have possibilities. I have only simple recorder for measuring relatively bad DUTs.
Do the measurements correlate with blind level matched tests with a few participants ?
The last link in my post above is all what I have to the moment concerning listening tests.
 
Top Bottom