A Crucial Question: Does flat response for speakers and harman curve for headphones tend to be individual’s favourite response?

[Sokel]

An easy way to see what consonance and dissonance is,is that little test (can't do it in dual tone so I went to triple by adding a freq that will not mess up acoustically)

Set the freqs exactly like this,
While playing,reduce the second freq (f2) by one step at a time with the little arrows.You will immediately notice what is all about.


DOUBLE CHECK THAT THE VOLUME IS LOW (as always)

 

[Chromatischism]

That's interesting, but it's essentially multi-tone, right?

 

[Sokel]

That's interesting, but it's essentially multi-tone, right?

Normally you do it with two tones but REW wouldn't let me set the f1 and f2 I wanted,automatically scale them down or up.

I know it from music,I can't imagine how you can use it for speakers where the Freqs can't change (and harmonics don't do that as you heard it).

I only posted it because someone has said something about it.If is useful so be it.

 

[someguyontheinternet]

It seems that there is some confusion about what the Harman target represents.

It is not necessarily the best curve for any given individual. However the chance of having a personal preference close to the Harman target is very high. There will be naturally occurring deviations based on the individual HRTF, but they are likely not very significant. Of course this does not deny that such cases can exist. It is simply a matter of probability which is normal when using statistic evaluation.

What Harman research also shows is that experience, nationality, sex and age have no significant bearing on the overall preference ranking - important: not preference rating! By ranking I mean the order of headphones by score. By rating I mean the numerical value score for any given headphone and it's variance among testers.

The reason why the Harman target should be used as an evaluation baseline for headphones is that the likelihood of any given individual having a preference close to the target is statistically higher than having a preference far away from the target.
If a headphone is close to the Harman target and has low distortion, it is capable of conforming to most individual preferences.

 

[MayaTlab]

What Harman research also shows is that experience, nationality, sex and age have no significant bearing on the overall preference ranking - important: not preference rating! By ranking I mean the order of headphones by score. By rating I mean the numerical value score for any given headphone and it's variance among testers.

A caveat to that point : in terms of ranking, the validation article for the over-ears virtual headphones methodology shows that even with a small sample size of six headphones the ranking order was different between the real and virtual headphones. The article takes satisfecit in the idea that the difference was limited to one rank order maximum, but the likelihood that this simply is a byproduct of the small sample size is quite high. For in-ears the validation article isn't really a test of "real" vs "virtual" headphones as the real headphones were never inserted in the listeners' ears, but rather as a test of the influence of non-linear distortion on listeners' preferences. It is quite crucial to know whether or not the virtual headphones methodology (ie measure headphones A on ear simulator, measure replicator headphones, EQ the replicator headphones to match how headphones A measure on the ear simulator) accurately reproduced how specific headphones' models would behave on someone's head, and I don't think that the validation articles are a particularly strong case that they systematically do (actually we already have evidence that they don't in some cases, the best documented being the issue of leakage).

 

[ADU]

Not only amplitude but the bandwidth/q is significant. Psychoacoustically, boosted sub/lower bass will mask other bass and mid frequencies less than an equivalent boost let's say beyond 120-150hz would mask elsewhere. Boosts past this point are almost universally considered bad or deleterious to accurate timbre, usually result in "muddy" sound is the term that's always thrown around, and in contrast you can get away with much more in lower frequencies lower than that before things like intelligibility become a major problem. This will obviously present itself more obviously in some genres, for example in a tune with vocals it will be more obvious to almost everyone.

The fact that harman iem target requires more gain in bass than their oe target leads me to suppose that the lack of low frequencies interacting with our body/pinna/room (which still doesn't explain the quantity of gain- bone conduction?) or perhaps the psychoacoustic effect of partial/full occlusion has an effect on how these frequencies are processed. I would like to assume that completely flat has to be "objectively" wrong as far as timbre accuracy in lieu of reflections, and maybe even as far as general preference is concerned for sounds with incidence close to the ear/different angle than is expected normally. As far as I know variance in low frequency transfer functions between individuals is quite low relative to other bands, so individual physiology explains that even less like it does with higher frequencies. There could be a lot of mechanisms, maybe since internal/infrasound may be more perceptible with occluded ears, we need the higher gain to mask or interpret these frequencies better?

The different bass levels in Harman's in-ear and over-ear targets could also just be an artifact of the somewhat outdated 711 measurement rig and ear coupler they were using for their initial research, as described here...

 

[ADU]

I respect Toole and Olive’s research and I understand that these are preferred response in blind test, though I do think Olive’s sample size is too small (usually 200-300 people in his various papers and only 50-60 in each strata like region, age etc) to safely establish a solid standard/target.
I personally like flat speakers more than coloured speakers (though not the case for headphones, I always feel harman curve having too much bass, I don’t need to eq my HD800s’s bass at all.)
But Are these preferred frequency response necessarily close to individual’s actual favourite response? I mean is it possible that an individual prefers flat and harman but at the same time prefer some different response more?
If that’s the case, wasn’t visiting shops and having as many demo as possible a more effective way than spending time reading measurements and analysing deviation from target(flat/harman) to find your favourite speakers and headphones, especially for non-professionals?
Then it circles back to just listen to the device...

Thanks.
But can you prefer Harman and actually like some significantly different response more?

This has probably already been addressed. But I think what the OP was really trying to ask here is whether an individual's preferences can be different that the subjectively preferred response in Harman's tests. And I think the answer to that is yes.

What the Harman research suggests though is that people generally prefer headphones which approximate the in-ear response of neutral loudspeakers in a room. And the anecdotal evidence seems to largely bear this out imo.

The question from my perspective is whether or not the current Harman target (or targets) really does a good job of accurately capturing or approximating this type of response.

 

[ADU]

A recent livestream with Dr. Olive discussing some of the above...

 

[markanini]

What the Harman research suggests though is that people generally prefer headphones which approximate the in-ear response of neutral loudspeakers in a room. And the anecdotal evidence seems to largely bear this out imo.

The question from my perspective is whether or not the current Harman target (or targets) really does a good job of accurately capturing or approximating this type of response.

It's a general and open ended question.

Looking at the popularity of headphones that match the target indicates that it works better than other targets at the very least.

AutoEq/results/RANKING.md at master · jaakkopasanen/AutoEq

Automatic headphone equalization from frequency responses - jaakkopasanen/AutoEq

github.com

If there's a more specific question, with any interest among people that attend meetups, and such, tests could be carried out to provide statistical proofs.