• 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.

Dagogo interview of Earl Geddes

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
16,836
Likes
13,190
Location
Seattle Area
#2
It is a very good read, especially the earlier section. I have known Earl for many years now. He has a very strong math background which he puts to good use. On acoustics and speaker design though, he has very strong opinions, some of which is distorted by wanting to sell his speakers. For example he used to put down JBL products because they relied on EQ. He would fight strongly against use of EQ only to later say it was OK to use.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
4,993
Likes
3,566
#3
Well I'd expect someone with his background and experience to have strong opinions.

Sounds like he needs a manufacturing guru to work with him and have his speakers produced in larger quantities.
 

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,408
Likes
2,234
Location
Monument, CO
#5
Nice article! Loved this line from it, which echos something I have often said (but did not know in what great company I was in!):
Again, this situation <ed: Golden Ears> is analogous to religion and its relationship to the Theory of Evolution. Many people will completely discount any and all scientific facts if they contradict a closely held personal belief.
And this:
I have always thought that if someone’s measurements do not “tell the whole story” then they are the wrong measurements. Technology has simply come too far to believe that “there are things that we cannot measure.” I have also never believed that all that matters is “how it sounds,” because this is such an unstable and personal opinion. Sound quality opinions can and will differ from person to person, system to system and most importantly even within the same person on different days (as I said before, I have personally witnessed this in well regarded “reviewers”).
He also says the same thing I and others have said many times hear about the rising "THD" levels at low levels:
“THD + noise” at low levels, as usually shown, could be all noise
Definitely a great read.

Note Dr. Toole did much (most?) of his research at the NRC of Canada, not at Harman, if that is your "Harman clique" (not sure why the dig at Harman, at least they are trying).
 

DDF

Member
Joined
Dec 31, 2018
Messages
42
Likes
23
Location
Ottawa
#7
EG wrote "Technology has simply come too far to believe that “there are things that we cannot measure.”
Ah, if that were only true:
Therein they show lack of correlation between controlled subjective testing and outcomes from THD and IM tests. The purpose of the study (Harmon) was to try and advance predictability of preference from trained listeners through measurements of distortion. One could argue perhaps they used the wrong metrics of measure (some advocate for weighting high order non masked components, or perhaps DMOS should be used instead of preference) but you would think leaders in the field wouldn't study this if the answer was already out there.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
16,836
Likes
13,190
Location
Seattle Area
#9
EG wrote "Technology has simply come too far to believe that “there are things that we cannot measure.”
Ah, if that were only true:
Therein they show lack of correlation between controlled subjective testing and outcomes from THD and IM tests. The purpose of the study (Harmon) was to try and advance predictability of preference from trained listeners through measurements of distortion. One could argue perhaps they used the wrong metrics of measure (some advocate for weighting high order non masked components, or perhaps DMOS should be used instead of preference) but you would think leaders in the field wouldn't study this if the answer was already out there.
Earl (and his wife) published the same years and years ago. See for example:
Geddes E., Lee L.: Auditory Perception of Nonlinear Distortion - Theory, Presented
at the 115th Convention of the Aud. Eng. Soc., Paper 5890 (2003 Dec.)

What he is saying in there is not that THD and IMD are the measurements but that if we wanted to measure some characteristics of a system, we can do it. Harman for example has shown that weighted frequency response of the speaker in anechoic chamber can be a strong predictor of listener preference. This was said to be impossible "as everyone hears differently." Well, turns out when it comes to sound reproduction in rooms, most of us hear the same and hence the reason the anechoic chamber data can predict their preference.
 

DDF

Member
Joined
Dec 31, 2018
Messages
42
Likes
23
Location
Ottawa
#10
Earl (and his wife) published the same years and years ago. See for example:
Geddes E., Lee L.: Auditory Perception of Nonlinear Distortion - Theory, Presented
at the 115th Convention of the Aud. Eng. Soc., Paper 5890 (2003 Dec.)

What he is saying in there is not that THD and IMD are the measurements but that if we wanted to measure some characteristics of a system, we can do it. Harman for example has shown that weighted frequency response of the speaker in anechoic chamber can be a strong predictor of listener preference. This was said to be impossible "as everyone hears differently." Well, turns out when it comes to sound reproduction in rooms, most of us hear the same and hence the reason the anechoic chamber data can predict their preference.
The GedLee metric, very familiar, I studied and critiqued it when I used to work in the field. I'm also familiar with most if not all the NRC/Harmon studies (and worked for years in an anechoic chamber) so I agree with the utility of Dr Toole's work.

I respectfully disagree however about EG's statement intent. "Can" (some day) and "know how" (now) are not the same and I think it makes no logical sense to assume he means the first. That would make his statement empty like saying "we can fly to Jupiter on a commercial flight". I think Earl is saying there is nothing we can't measure now. The fact that we learn how to measure things over time, or that there are some measures which give strong correlation to some subjective factors (i.e. referencing some of the NRC work) do not make his statement true.

I remain ever appreciative to have learned a lesson early in my career from the late great Richard C Heyser's take on the subject: we try to measure non linear phenomenon using linear measurement & analysis techniques and then wonder why we struggle?
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
2,469
Likes
1,569
#11
EG wrote "Technology has simply come too far to believe that “there are things that we cannot measure.”
Ah, if that were only true:
Therein they show lack of correlation between controlled subjective testing and outcomes from THD and IM tests. The purpose of the study (Harmon) was to try and advance predictability of preference from trained listeners through measurements of distortion. One could argue perhaps they used the wrong metrics of measure (some advocate for weighting high order non masked components, or perhaps DMOS should be used instead of preference) but you would think leaders in the field wouldn't study this if the answer was already out there.
They did indeed simply use the wrong metrics. The problem is not that what is heard cannot be measured, it's that what is measured is not weighted correctly.

There has been a lot of excellent attempts to develop and test more sophisticated distortion metrics than THD/IMD.

This paper tests the correlation between these new metrics and subjective experience.

In any case, there's nothing here that can be heard but not measured, just some persistent misunderstandings about which measurements correlate to subjective experience.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
4,993
Likes
3,566
#12
I thought it was interesting that he said speaker distortion is pretty much a non-issue. If I understood, he was saying it is usually 2nd and 3rd order which we don't much hear unless it is very large.
 
Joined
Jan 26, 2019
Messages
79
Likes
78
#13
Nice article! Loved this line from it, which echos something I have often said (but did not know in what great company I was in!):


And this:


He also says the same thing I and others have said many times hear about the rising "THD" levels at low levels:


Definitely a great read.

Note Dr. Toole did much (most?) of his research at the NRC of Canada, not at Harman, if that is your "Harman clique" (not sure why the dig at Harman, at least they are trying).
Oh no, I highly respect Harman's work. Just that Toole and Olive seem to get a disproportionate amount of coverage and its good to see someone else's research discussed.

Well I'd expect someone with his background and experience to have strong opinions.

Sounds like he needs a manufacturing guru to work with him and have his speakers produced in larger quantities.
This is frustrating. Very clever design, state-of-the-art horizontal polars, presumably excellent dynamics and basically no real substitute for that oblate spheroid + foam waveguide. What's the point of such a sophisticated speaker if next to nobody can have it - and not for want of trying? Though I believe there was a joint venture years past that failed because of the 2008 crisis, and talks of a licensee a few years ago
 

DDF

Member
Joined
Dec 31, 2018
Messages
42
Likes
23
Location
Ottawa
#14
They did indeed simply use the wrong metrics. The problem is not that what is heard cannot be measured, it's that what is measured is not weighted correctly.

There has been a lot of excellent attempts to develop and test more sophisticated distortion metrics than THD/IMD.

This paper tests the correlation between these new metrics and subjective experience.

In any case, there's nothing here that can be heard but not measured, just some persistent misunderstandings about which measurements correlate to subjective experience.
Looks like an excellent paper, thanks for sharing, I started reading. Correlations were high for the test conditions but I would argue it doesn't test time variant distortion transfer functions (such as in transducers) nor important secondary effects like group delay which change the signal envelope and change threshols (proven by Zwicker IIRC). Its a step forward but leaves much to explore. Still, very cool.

While an interesting study and potentially an important stride in being better able to predict audibility of distortion under limited scenarios, it by no means is supporting evidence that we have all the necessary tests in our arsenal to predict audibility or preference. I think that's an indefensible position and what I believe Earl was saying (hopefully I'm misinterpretting his intent). How would you even prove such a proposal? I've been involved with enough professionally conducted DBT to know such a thought is folly.
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
2,469
Likes
1,569
#15
Looks like an excellent paper, thanks for sharing, I started reading. Correlations were high for the test conditions but I would argue it doesn't test time variant distortion transfer functions (such as in transducers) nor important secondary effects like group delay which change the signal envelope and change threshols (proven by Zwicker IIRC). Its a step forward but leaves much to explore. Still, very cool.

While an interesting study and potentially an important stride in being better able to predict audibility of distortion under limited scenarios, it by no means is supporting evidence that we have all the necessary tests in our arsenal to predict audibility or preference. I think that's an indefensible position and what I believe Earl was saying (hopefully I'm misinterpretting his intent). How would you even prove such a proposal? I've been involved with enough professionally conducted DBT to know such a thought is folly.
Well, the study wasn't looking at group delay; it was only interested in nonlinear distortion. There have been plenty of other studies where nonlinear distortion was kept constant and group delay audibility was tested. If the authors of this study had tried to test for both simultaneously, they would have introduced more than one variable which would have invalidated the results of the study.

I think what you say about "time variant distortion transfer functions (such as in transducers)" is interesting. Could you flesh out what you mean by this or give a more concrete example to illustrate? If you mean what I think you mean, I believe such a thing could be (for example) measured in a specific speaker system, then modelled using DSP and tested for (in isolation from other variables).
 

Ron Texas

Addicted to Fun and Learning
Joined
Jun 10, 2018
Messages
822
Likes
395
Location
Equidistant From Everywhere
#16

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,408
Likes
2,234
Location
Monument, CO
#17
Top Bottom