• 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!

Question regarding original data where loudspeaker preference scores were given by trained listeners

gags11

Senior Member
Forum Donor
Joined
Feb 8, 2019
Messages
356
Likes
528
So I have had a chance to read some of discussions and glance over the papers regarding the original research Drs. Toole and Olive have done regarding loudspeaker preference scores and curves. I truly appreciate all the work that really went into this and it is highly commendable. Amir, has also presented the work in his own video showing how trained listeners are better in giving scores that correlate with the most flat curve.

I have a problem with this, and hope after I explain, some of you should also. It seems these trained listeners are very good in knowing what the flat FR loudspeaker should sound like. They are very well trained in knowing and identifying FR deviations. My problem is jumping from this to saying that the most flat FR is the most preferred. Based on the study they have performed, theses trained listeners were highly biased in trying to score and identify the most flat FR loudspeakers. I feel I have become biased this way also.

A true preference score should be done with non-trained listeners who will give their subjective preference without being biased toward identifying the most flat FR. You may need way more people in trying to come up with an experiment like this.

I may be wrong in this. Would appreciate input.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
This is addressed in the paper:

1637391175450.png


Yellowed highlight has been proven to be true in multiple studies.
 
OP
G

gags11

Senior Member
Forum Donor
Joined
Feb 8, 2019
Messages
356
Likes
528
Thank you Amir, appreciate the clarification. I see this paper claims that untrained listeners have the same preference, but we have not seen the actual data.

I remember though, you presented some data that showed untrained listeners preference scores did not really correlate, or there was a hint of correlation at best, with the flat FR and trained scores. My point being, a larger study may be necessary.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
Oh I see what you are saying. Trained listeners are indeed far more picky and provide a larger spread for speaker rating. As such, a model matching them will over predict listener preference for ordinary listeners. Is this what you mean?
 
OP
G

gags11

Senior Member
Forum Donor
Joined
Feb 8, 2019
Messages
356
Likes
528
Oh I see what you are saying. Trained listeners are indeed far more picky and provide a larger spread for speaker rating. As such, a model matching them will over predict listener preference for ordinary listeners. Is this what you mean?

Yes, overpredicting and generalization would be a good way to put it.

I would be interested in a study where a single speaker is tested by untrained listeners, and various tweaks to the FR are done at a DSP level. This would take away listener bias as well as other confounding factors in trying to identify what is the preferred FR for most people.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
The preferred frequency response as I noted has been extensively studied and matches what the study says.

With respect to scoring, it is very true that listeners for the most part like just about any speaker. :) So what a trained listener likes which garners low scores, doesn't represent the general public/general audiophile market.
 
Top Bottom