Alternative perspective on the Harman loudspeaker preference curve.

[olieb]

Did you get an answer? Maybe I missed it.It is of course crucial.

Having 11 participants decide which is the best type of ice cream, you cannot generalize any results on what people in general think about that.

I did not (as far as I know). But that is probably my foult. I cannot access AES papers and did not read Toole's book.
But I found Sean Olive's blog and it is obvious that there has been extensive research into the Harman curves.
An example: http://seanolive.blogspot.com/2008/12/loudspeaker-preferences-of-trained.html

There are few published studies to support the notion that trained listeners have different loudspeaker preferences than untrained listeners. To study this question, the author conducted a large study (see reference 1) that compared the loudspeaker preferences of 300+ untrained and trained listeners. Over the course of 18 months, an identical controlled, double-blind listening test was repeated with different groups of trained and untrained listeners who rated 4 different loudspeakers on an 11-point preference scale using 4 different music programs. Loudspeaker positional effects were controlled via an automated speaker shuffler that moves each loudspeaker into the exact same position.

There can always be more research and better studies but so far there seems to be a very sound fundament.

 

[youngho]

Just to keep beating the not quite dead horse, please do not call it a target curve. It is not a target curve. It is an average of the preference of 11 people listening to some decent speakers at Harman. No one concluded or suggested that this should be a target for people tuning their systems.

I think this is a matter of semantics and viewpoints. The Olive et al AES paper was called “Listener Preferences for In-Room Loudspeaker and Headphone Target Responses." Toole's caption for the figure you linked in the first post is "Subjectively preferred steady-state room curve targets in a typical domestic listening room." Toole writes in that paper, "Over the years a few investigators have attempted to identify advantageous room curve targets for small rooms...More data would be enlightening, but this amount is sufficient to indicate that a single target curve is not likely to satisfy all listeners..." Welti and Voecks seem to go along with the description "target response curve" https://hometheaterhifi.com/technic...n-interview-with-todd-welti-and-kevin-voecks/. I am well aware of the limitations, as I have clearly outlined a number of my concerns above, but like it or not, this curve seems to treated as a target by many both here on ASR and elsewhere.

One point of interest to me is that in headphone target response curves (https://www.researchgate.net/public...or_different_headphone_target_response_curves), the hypothesis was that "A headphone’s target response should ideally approximate an accurate loudspeaker in a reference listening room" (https://www.listeninc.com/wp/media/Perception_and_-Measurement_of_Headphones_Sean_Olive.pdf) but the most preferred curve was Target 2 or RR1, "Similar to ["target response is based on the in-room response of the loudspeaker system ("a 7- channel loudspeaker system (JBL LSR 6332 [19]) with four subwoofers (JBL 1500) located in the four corners of the room") in the Harman Reference Listening Room calibrated with the standard in-room target curve"] but using the modified in-room target curve, which has less bass and treble" with Target 1 or RR "frequently described as sounding too bright." Obviously, the listeners did not have the ability to control the bass and treble separately like in https://www.aes.org/e-lib/browse.cfm?elib=17042, so perhaps more treble bothered the listeners than less bass, even though bass is an important part of listener preference, but I wonder whether confirmation bias led Olive et al to largely return in https://www.aes.org/e-lib/browse.cfm?elib=17042 to their original hypothesis, while this suggests to me that more damping (or increasing directivity...or fall-off with distance...) at higher frequencies may be preferred for a majority of listeners, even with "accurate" loudspeakers, with the implication being that this can should be done through room furnishings and treatments.

Absolutely. A total of 11 participants is hardly sufficient for a reliable statement.
So the trend/result may be correct or not so much. Are there follow up studies?

Not exactly. There was a follow-up study for headphones with a much larger sample suggesting somewhat more bass boost and less treble cut--see slide 59 at https://www.listeninc.com/wp/media/Perception_and_-Measurement_of_Headphones_Sean_Olive.pdf, though there seems to be a small error in the first bar graph, which should say +6.44 bass instead of +6.6. However, you can see that the study population was clearly more heavily male- and untrained-skewed, which could easily explain the difference. One could possibly extrapolate back to https://www.aes.org/e-lib/browse.cfm?elib=17042 accordingly.

Young-Ho

 

[sigbergaudio]

I think this is a matter of semantics and viewpoints. The Olive et al AES paper was called “Listener Preferences for In-Room Loudspeaker and Headphone Target Responses." Toole's caption for the figure you linked in the first post is "Subjectively preferred steady-state room curve targets in a typical domestic listening room." Toole writes in that paper, "Over the years a few investigators have attempted to identify advantageous room curve targets for small rooms...More data would be enlightening, but this amount is sufficient to indicate that a single target curve is not likely to satisfy all listeners..." Welti and Voecks seem to go along with the description "target response curve" https://hometheaterhifi.com/technic...n-interview-with-todd-welti-and-kevin-voecks/. I am well aware of the limitations, as I have clearly outlined a number of my concerns above, but like it or not, this curve seems to treated as a target by many both here on ASR and elsewhere.

There is a big difference between
"A speaker that naturally presents this in-room curve is likely to sound good"

and
"If I EQ my in-room response to look like this (no matter the original response), it is likely to sound good."

What many are doing / assuming is the latter. This is not what Harman is saying, and it is likely to result in a less than optimal outcome.

 

[youngho]

There is a big difference between
"A speaker that naturally presents this in-room curve is likely to sound good"

and
"If I EQ my in-room response to look like this (no matter the original response), it is likely to sound good."

What many are doing / assuming is the latter. This is not what Harman is saying, and it is likely to result in a less than optimal outcome.

Oh, my concern was different, more on the preference generalization side (than the correction side), like what you see here and here, such that "Every age, "pronoun", and ethnicity prefers the Harman Curve voicing" or that "100% of listeners prefer the Harman Curve [for headphones] to within 1 dB above 300 Hz."

 

[sigbergaudio]

Oh, my concern was different, more on the preference generalization side (than the correction side), like what you see here and here, such that "Every age, "pronoun", and ethnicity prefers the Harman Curve voicing" or that "100% of listeners prefer the Harman Curve [for headphones] to within 1 dB above 300 Hz."

Right. Yes, I agree there is significant variance in individual preference.

What I think is perhaps somewhat unknown is the reason for this. Is it a "learned" preference based on what one are used to in their own system, is music genre preference a parameter, etc etc.

 

[NTK]

There is a big difference between
"A speaker that naturally presents this in-room curve is likely to sound good"

and
"If I EQ my in-room response to look like this (no matter the original response), it is likely to sound good."

What many are doing / assuming is the latter. This is not what Harman is saying, and it is likely to result in a less than optimal outcome.

And therein lies the inherenet contradiction of the Olive preference scores: PIR (predicted in-room response) contributes to 38% of the score. If you neglect LFX (low frequency extension), it is more than half.

Dr Toole has emphasized that EQ the PIR to a certain target does not guarantee good results. But obviously one can EQ to optimize the required characteristics for the PIR curve (small narrow band deviation [NBD] and high smoothness [SM]) to maximize the score. No wonder Harman doesn't use the score internally. They know it'll probably lead you in the wrong direction.

 

[thewas]

We should keep in mind though that the PIR is a theoretical result based on anechoic measurents and constant/ideal room assumptions and therefore equalising to it is something different than equalising a loudspeaker in a room based on listening position measurements.

 

[JoachimStrobel]

This is due to the gradual and ongoing boosting of LF in all aspects of live/amplified/replayed music of any type in the last 30 years. In fact, I'd say it was digital and the ability to record any amount of LF without causing problems that created the problem.

The availability of vast amounts of clean power at LF and drivers that can absorb and reproduce it has resulted in an "arm's race" in the low frequencies.

When you are in a small club/venue with a live drummer only a few feet away and they feel the need to augment an acoustic with an electronic is where the problem lies IMO. It sounds stupidly disproportionate in situations like that or in 'live' situations where you are near the stage and the drum kit is just an excuse for LF electronic impulses. Drums don't sound like drums anymore on recordings- many never existed in the first place and the ones that did, have been amplified to ridiculous levels.

Most systems I hear where the unfortunate owners think they are "fully dialled in" sound putrid. Like the obsession with subwoofers to 'reproduce' sounds that never existed in the first place, except in the minds of the mixing guys (not engineers). Drums are mid bass, not seismic.

There is more to that. Once you listened to a poor bass heavy sound at home, people believe that the 200-500 hz resonance is a good bass response. Sadly they do not hear that in a concert or with a properly corrected home system. So they crank up the bass. And so it goes…

 

[Yos]

Absolutely. A total of 11 participants is hardly sufficient for a reliable statement.
So the trend/result may be correct or not so much. Are there follow up studies?
I overlayed your work to make it more easy to compare.
In the range 200-500 it seems to be a bit off, as the curves should be more or less parallel there.

View attachment 335718

EDIT: I updated the graph with the improved results from @sigbergaudio below.
View attachment 335751

"To establish the Harman curve, the team conducted a series of blind tests with 283 participants from four different countries across 11 test locations."

 

[muad]

"To establish the Harman curve, the team conducted a series of blind tests with 283 participants from four different countries across 11 test locations."

The "harman curve" is a headphone target.

The curve these guys are discussing were the preferred in room speaker response in a specific study. It used a very linear speaker with good dispersion to get

 

[sigbergaudio]

Even if you shift the curves to meet at 2 kHz they should keep their form and in the range 200-500 the curves are on top of each other.

Not sure exactly what you are getting at. We discussed something similar on page one, did you read through that?

 

[kemmler3D]

No wonder Harman doesn't use the score internally.

Well, yeah. A single number isn't really informative when you actually have access to measuring equipment and the actual speakers themselves. At most I would guess they'd set a pref. score target for the project, the way game studios set a metacritic score target. But it's not something you can actually intelligently use in the design process.

 

[olieb]

"To establish the Harman curve, the team conducted a series of blind tests with 283 participants from four different countries across 11 test locations."

Even if you shift the curves to meet at 2 kHz they should keep their form and in the range 200-500 the curves are on top of each other.

It seems a bit like we are at the start of round 2, where most aspects will be repeated.
Before that happens, I would like to introduce some slightly new things. (Maybe I fail.)

The graph from the very first post shows the result of a study with 11 participants, 7 °trained" and 4 "untrained". That makes the statistics somewhat shaky. Maybe there is abundant other data to support this but nothing has come up here so far. So I take this at face value.
The average of four preference curves of untrained listeners seems to me to be not reliable enough, I would discard it. At best it can give some hints at the difference between the groups. I would discard the curve for the full group too, as it is an average of two obviously differing subgroups.

So I am left with two curves (I kept the normalization at 2 kHz):

• "Trained Preference" (black), the averaged (and smoothed?) preferences (FR at listening position) of seven trained listeners adjusted in a good room with a well engineered speaker. The curve is quite close to a -1dB/oct tilted flat curve.

As I understand, Toole made clear that the trained "preference" curve is not supposed to be a (general) "target" curve. The reasons are not so clear to me. Is it because the curve not general enough (only one speaker in one room)? Other reasons?

• "Predicted FR" (red), the expected FR of the well engineered speaker in the room without adjustment (flat direct sound). This curve does not look like a measurement but more like a theoretical, idealized expectation. I will assume here that the "predicted FR" is indeed close to the smoothed, unequalized FR in room of the speaker in the study. This FR can be approximated by a -0.5dB/oct tilted flat curve.

I added the smoothed (1/3rd octave) PIR of Genelec 8341 from Amir's test for comparison. The speaker in the study looks like having a rather wide radiation pattern (even wider in HF than the Genelec).

Trained listeners seem to accept (actually prefer) a considerably altered on-axis curve in return for an adjustment of the room frequency response, because the "trained preference" differs from the "predicted FR" that might be seen as the "natural FR" of the speaker.
It seems like the trained listeners prefer more LF and a reduced level in HF (a more tilted FR) than this speaker produces with flat direct sound.
EQ is plotted in yellow.

This rises a couple of questions.
Is a flat direct sound - contrary to general wisdom - not the most preferred option with wide speakers?
Would a speaker with narrower radiation that combines flat direct sound with a more tilted in room FR be preferred over the speaker in the study?

Black: Trained Preference	Green: -1 dB/oct (dark) ±1dB (light)
Red: Predicted FR (of speaker in room)	Red: Shifted Copy of Predicted FR
Yellow: Approximate EQ for Trained Preference
Blue: PIR for Genelec 8341 (smoothed) for comparison