[TronII]

I've been looking into frequency response curves and a few things strike me as odd: Why does the ISO 226:2005 equal-loudness contour differ from the free field response by up to 30 dB in the lower frequencies? And, while we're at it: doesn't the concept behind the Harman target seem dubious? Wouldn't it make more sense to have audio engineers EQ a headphone to match a speaker tuned to have a neutral response, average their responses, and generate a curve based on that? Doing it based on consumer opinion and how speakers get a "warm tilt" outside of an anechoic chamber smells like MQA-esque marketing, especially coming from a speaker company.

 

[castleofargh]

I've been looking into frequency response curves and a few things strike me as odd: Why does the ISO 226:2005 equal-loudness contour differ from the free field response by up to 30 dB in the lower frequencies?

I don't think there is a secured answer for that, or that there will ever be one.

And, while we're at it: doesn't the concept behind the Harman target seem dubious? Wouldn't it make more sense to have audio engineers EQ a headphone to match a speaker tuned to have a neutral response, average their responses, and generate a curve based on that? Doing it based on consumer opinion and how speakers get a "warm tilt" outside of an anechoic chamber smells like MQA-esque marketing, especially coming from a speaker company.

the resulting curves from Harman are an accumulation of experiments. decades of testing on speakers gave them confidence that what we preferred was also what felt neutral to us on speakers, and that happened to be not flat at our ears, but flat speakers put in some reference room(so not actually a flat response at the listener's position). as their work on the subject has pretty much redefined the speaker industry, you can assume that it's not that controversial.
so when starting to work on a target for headphones, their first assumption was that they should aim for that same response for the listener. that was the objective approach you talk about, they just had already reached a specific conclusion about perceived neutral/preferred FR. so they went to apply a correction to test that idea on headphones(on a pair of HD800 I think), but it turned out that the subjects seemed to have a different preferred target, so they moved on to try and determine what it was(while also trying to test various ideas/questions) because after all they sell stuff so knowing what people prefer is pretty relevant to them ^_^.
the final curves are averages, so some people won't feel like the sound is neutral or nice to them, that's expected. but the all project is mostly good scientific method at work IMO.

 

[TronII]

The Harman response mostly seemed dubious to me because of the idea behind it: headphones should sound like good speakers in a "good room" instead of an anechoic chamber. I had no idea there was so much research behind it beyond consumer testing

 

[castleofargh]

The Harman response mostly seemed dubious to me because of the idea behind it: headphones should sound like good speakers in a "good room" instead of an anechoic chamber. I had no idea there was so much research behind it beyond consumer testing

where did you see Harman state that the sound should reproduce the FR of an anechoic chamber? I believe that their assumption from the start has been that we'd like the sound of flat speakers in a reference room. that they still decided to propose free field and diffuse field along with other target responses for the tests so that listeners can decided how they feel about said curves, that's simply thorough testing not relying only on assumptions. in this case the experiment agreed with them that the preferred response was neither free field nor diffuse field. but further testing also showed that listeners did not prefer the FR of speakers in a reference room(expected TBH as FR alone means we're dismissing the impact of reverb, non linearity, tactile bass, and crap). realizing that their assumption wasn't correct, they simply proceeded by following what the data from listeners told them. anytime the listeners preferred something specific, they simply went that way. in other tests, people had a basic bass and treble EQ to set what they wanted. they also tested for difference in taste with ages and culture, and went to test people on different continents, etc. the all thing has been published into maybe a dozen papers over the last years(and I probably missed some), with the latest one even taking care of change between headphone and IEMs. it's been a rather long process.
most papers might be hard to find now, so you'll have to pay AES to get them . but Sean Olive has been interviewed a bunch of times, so you can probably get a good deal of information by finding some youtube vids or some podcasts with him. just by listening to him talk for 5mn, anybody can tell that the guy is a scientist and not a carpet seller.

 

[71 dB]

I've been looking into frequency response curves and a few things strike me as odd: Why does the ISO 226:2005 equal-loudness contour differ from the free field response by up to 30 dB in the lower frequencies? And, while we're at it: doesn't the concept behind the Harman target seem dubious? Wouldn't it make more sense to have audio engineers EQ a headphone to match a speaker tuned to have a neutral response, average their responses, and generate a curve based on that? Doing it based on consumer opinion and how speakers get a "warm tilt" outside of an anechoic chamber smells like MQA-esque marketing, especially coming from a speaker company.

I believe ISO 226:2003 is the current standard, but equal-loudness contours don't matter, because they are the same for all listening scenarios. Mixing music on fixed levels takes care of that.

This very compex topic beginnig with how balanced speakers sound and how music is mixed for them etc. Harman target tries to take into account how a typical room increases bass level due to acoustics (longer reverberation time + room modes at low frequencies), the torso+head effect of the listener around 3000 Hz and the attenuation of highest frequencies. Hardly dubious.

 

[bigshot]

Headphone users sometimes don't understand that the room is as much a part of the sound as the speakers. In the real world, we never hear sound isolated perfectly like we do in headphones It always has a complex envelope of reflections wrapped around it that not only affect timing, but also affect response as well. That envelope just isn't there with headphones. Adjusting the curve a bit to simulate that natural quality of sound takes a little bit of the curse off the sterile sound of headphones and makes it sound more natural. Different rooms affect the sound differently, so there isn't a hard and fast formula for this. But the Harman curve is a great place to start.

 

[TronII]

I think I've made a goof with the OP, I'm sorry.