"We need to move on and use distortion measurements that are perceptually meaningful."

[solderdude]

I will just repeat what I wrote earlier and am quite sure this is what Dr. Olive is talking about.

The biggest problem with using the word 'distortion' is that there are many kinds of distortion and the threshold of audibility differs with test program material (music, tones), type of distortion and listening level (SPL) and training of the listener(s).
There is no single number that can describe all these variables. That makes distortion measurements pointless in regard to perception.

There simply is no single number that can capture signal fidelity. The closest thing could be something like PK metric but that should also have AI to analyze the null with perceptional algo's.
Then again ... that, at least for amplifiers, might need to give a range of numbers it falls in with different loads.

With DACs you might need a range of numbers for 0dBFS and say -1dBFS program material so again... not a single number.

Then one would need to understand how you personally can relate to those numbers, in other words below which number you might hear audible degradation related to signal fidelity (excluding frequency response).

So ... one can dream on with wishing there will even be a single number that actually says someting.

Case in point: Some folks prefer tube amps that substantially modify the original signal and that effect should also be included in that single number.
What sounds nice to person A may not be nice to person B and may or may not measure well. How can you fit that in a single meaningful number applicable to all ?

 

[192kbps]

I will just repeat what I wrote earlier and am quite sure this is what Dr. Olive is talking about.



There simply is no single number that can capture signal fidelity. The closest thing could be something like PK metric but that should also have AI to analyze the null with perceptional algo's.
Then again ... that, at least for amplifiers, might need to give a range of numbers it falls in with different loads.

With DACs you might need a range of numbers for 0dBFS and say -1dBFS program material so again... not a single number.

Then one would need to understand how you personally can relate to those numbers, in other words below which number you might hear audible degradation related to signal fidelity (excluding frequency response).

So ... one can dream on with wishing there will even be a single number that actually says someting.

Case in point: Some folks prefer tube amps that substantially modify the original signal and that effect should also be included in that single number.
What sounds nice to person A may not be nice to person B and may or may not measure well. How can you fit that in a single meaningful number applicable to all ?

I read what was in the link Dr. Sean Olive posted, maybe the Dr. is talking about this?

What Loudspeaker Specifications Are Relevant to Sound Quality?

This past week I attended the International Loudspeaker Manufacturer’s Association (ALMA) Winter Symposium in Las Vegas where the theme wa...

seanolive.blogspot.com

 

[solderdude]

Linear distortions are not the same as non-linear distortions.
So one has to be careful with the term 'distortion' which is what you can see in the quote above (point 3)

That's why I stated one should exclude frequency response from a 'distortion' number. Frequency response is by far more sound determining than non-linear distortion.

So you need at least a few numbers that say something about tone (accuracy and character) and non-linear distortions.
For transducers it will be range it falls in depending on SPL as well so again no single number. It simply is impossible as it is too complex.

 

[computer-audiophile]

I like seeing audio measurements, and generally they indicate 'engineering excellence' but they are not metrics of sound quality.

That is aptly said. I see it that way too.

 

[somebodyelse]

From my end embedded tweets show up with that linked part of the discussion... not just links.

I only see links, and following them just takes me to a login page. I'm with @restorer-john on cross-posting.

 

[192kbps]

I only see links, and following them just takes me to a login page. I'm with @restorer-john on cross-posting.

Sorry, it may be a problem with different mobile phones or computers or mobile phones. I will re-edit and attach the screenshots.

 

[Sokel]

https://twitter.com/i/web/status/1708948631233941885

I really can't understand, help me.

Is that book talks about electronics or speakers?
Cause the "high fidelity" row states 5% H2 and 2.5% H3 as good for trained listeners which is at the 20db SINAD region and higher.

One page is not a good indicator,he can mean that for the 20Hz region of a loudspeaker for example.

 

[JSmith]

Sorry, it may be a problem with different mobile phones or computers or mobile phones. I will re-edit and attach the screenshots.

It's nothing you have done... it's the difference between using a PC and using a phone.

I'm with @restorer-john on cross-posting.

All good, you're welcome to your opinion of course. However there are no rules regarding this... myself, others and even Amir have done this in the past with posts and conversations from other forums, sites and YT comments. X/Twitter is no different IMO... I don't see the issue.


JSmith

 

[voodooless]

Sorry, it may be a problem with different mobile phones or computers or mobile phones. I will re-edit and attach the screenshots.

I guess people should stop using Netscape Navigator

 

[Sokel]

Also interesting is that I have the feeling that a sizable part of the audience that gulps the “must have SOTA” electronics mantra, are probably former high-end subjectivists, that now want to dabble in this latest ASR fueled fad.

I can confirm that they buy these new cheap DACs by the tenths (their price is a joke comparing with what they used to pay) but amps and speakers not so much.
That's a castle that stands strong still.
Specially the big speakers (and I mean big) .

 

[Talisman]

None of this should be surprising to anyone familiar with measurements and what certain measurements mean in the real world.
Sinad taken alone, beyond a modest value of 75/80db, I no longer care.
It makes much more sense at that point just the SNR value, for any problems with hiss audible from the listening position, but if it is not audible at the listening position, even SNR becomes irrelevant. Anyone who replaces a sinad 90 amplifier with a sinad 110 believing they have made an upgrade and hearing "a much cleaner sound" is no different from someone who pays hundreds of euros for a cable that does the same job as the 20 euro cable.
Especially for amplifiers, power output, ability to handle low impedance loads, and flat, impedance-invariant frequency response are infinitely more important factors.
But everyone has their gods to worship it seems.

 

[Leporello]

The quote from 1938: "is not a measure of the degree of distastefulness to the listener".

Does it mean that:

1. The distortion does not audibly alter the sound
or
2. The distortion audibly alters the sound but not enough to be objectionable

These are very different propositions.

If 1, how reliably could this be determined in 1938?

 

[Sokel]

If 1, how reliably could this be determined in 1938?

I wondered about this too.
It seems like Massas are holding a long tradition to this stuff at a strict science and high engineering level:

Whitepapers - Using Ultrasonic Technology - Massa

Read recent whitepapers by Donald Massa and Dawn Massa about using ultrasonic technology in challenging situations and environments.

www.massa.com

 

[restorer-john]

The quote from 1938: "is not a measure of the degree of distastefulness to the listener".

Does it mean that:

1. The distortion does not audibly alter the sound
or
2. The distortion audibly alters the sound but not enough to be objectionable

These are very different propositions.

If 1, how reliably could this be determined in 1938?

It's a quote from an historic text which conveniently suits the narrative being prosecuted by the person concerned. Take it with a grain of salt.

 

[restorer-john]

Which is most likely what Sean Olive is thinking about.

Well, he should make that crystal clear shouldn't he?

 

[PierreV]

"Floyd Toole made the observation that there is more useful performance information on the side of a tire (see tire below) compared to what’s currently found on most loudspeaker spec sheets"

I make the observation that my very life has more often been dependent on my tires than on my loudspeakers.

 

[restorer-john]

"Floyd Toole made the observation that there is more useful performance information on the side of a tire (see tire below) compared to what’s currently found on most loudspeaker spec sheets"

I make the observation that my very life has more often been dependent on my tires than on my loudspeakers.

What's next, the nutritional panel on a cereal box has more information than the specs printed on the box of a pair of JBL headphones? Oops...

 

[wynpalmer]

Virtins Pro includes the GedLee metric as part of its test suite. I use it, and a 21 tone 20-40kHz c. 1/f weighted stimulus with the peak set to the RMS "nominal" output level plus 3dB to test phono preamplifiers. The tones are spaced to produce non overlapping intermodulation products and the goal is for the products to not be visible above the noise floor, or to be below -110dBc, integrated.
The "goal" for the Gedlee metric as used in these designs, is for the result expressed as a % to be essentially identical to the THD expressed as a % over the operating range of the amplifier, and to be below 0.0003% or so which in practical terms means suppressing high order non linearities.
This effectively satisfies the "ancient texts" requirements for a harmonic order weighted THD, while actually focusing on the overall transfer function non-linearity of the "black box" rather than on any particular implementation.

 

[Moderate Dionysianism]

We had a discussion about this which touched on color perception. There are any number of ways to measure color which are irrelevant to humans but useful for engineers. Sherwin Williams could provide this graph for a Cadmium Red Hue:

Is this useful for a paint engineer? Absolutely. Is it useful for a consumer or someone evaluating paint colors? No

This is a false analogy that keeps coming back. This would be like evaluating the aesthetics of music based on a spectrogram, but no one is proposing that. The discussion here is about evaluating the reproduction gear, not the content it reproduces. You should've shown a measurement of an OLED screen for instance, not the spectrum of a hue. So yes, this graph is useless for me as a consumer when shopping for wall paint, but so would be the spectrum of a song when purchasing music. OTOH, a lab report showing how a screen 'distorts' the intended hue (=how a piece of audio gear distorts sound) is perfectly useful.

 

[617]

Virtins Pro includes the GedLee metric as part of its test suite. I use it, and a 21 tone 20-40kHz c. 1/f weighted stimulus with the peak set to the RMS "nominal" output level plus 3dB to test phono preamplifiers. The tones are spaced to produce non overlapping intermodulation products and the goal is for the products to not be visible above the noise floor, or to be below -110dBc, integrated.
The "goal" for the Gedlee metric as used in these designs, is for the result expressed as a % to be essentially identical to the THD expressed as a % over the operating range of the amplifier, and to be below 0.0003% or so which in practical terms means suppressing high order non linearities.
This effectively satisfies the "ancient texts" requirements for a harmonic order weighted THD, while actually focusing on the overall transfer function non-linearity of the "black box" rather than on any particular implementation.

Do you find the Gm moves your designs in a different direction compared to using THD?

The Gm metric is interesting because in theory it could lead to different design optimization strategies in transducers. In electronics, distortion can be minimized to such a degree that something with a good Gm score would probably also have extremely low, sub-sensible THD, but imagine if you were designing something where THD was an actual factor - a bluetooth speaker perhaps. I'm very curious if Gm would actually lead to different design decisions.

Geddes' own approach to speaker design is (was) somewhat akin to the approach we have in contemporary electronics - simply overbuild it to such an insane degree that nonlinear distortion is minimized. His approach was to build domestic speakers with pretty high end PA drivers, and they do indeed sound very clean at normal and loud levels. The crossover points strain the drivers more than they would in a PA setting, but this isn't a problem at home listening levels.

The only optimization I've seen in Geddes' speakers that corresponds to the Gm metric in particular is the sensitivity to delay based distortions, which he optimizes with very smooth waveguide transitions, large round-overs and of course the foam plug which attenuates the longer path lengths of horn-mouth artifacts compared to the shorter path lengths of the nice sounds. The foam plug is one of the cleverest things I've ever seen in a loudspeaker.

I do wonder how Gm could apply to headphones, that seems like an area where it could have a lot of benefit.