[arnaud]

I could not find the original reference yet but here's an example for a population of chinese individuals, check figure 8 for typical scatter: http://cdn.intechopen.com/pdfs-wm/45612.pdf

 

[MRC001]

I could not find the original reference yet but here's an example for a population of chinese individuals, check figure 8 for typical scatter: http://cdn.intechopen.com/pdfs-wm/45612.pdf

I've been wondering whether the HRTF for deeply inserted IEMs (like Etymotics) is different from over-the-ear headphones like HD-600s. Intuitively, it seems it should be different. Etys like the ER-4S are deeply inserted and play the music right near the eardrum, while over-the-ears are entirely outside the ear. One would seem to be affected by the shape of the outer ear, the ear canal, etc. which the former completely bypasses.
 
What made me think of this is my custom audiologist molded earplugs. As a flute player I sit in front of the brass, which is very loud. And I use them for practicing piccolo, which is also very loud (110 dB SPL measured at the head while playing). They are Etymotic ER-15s which have close to flat frequency response. The audiologist cut them too long/deep intentionally, so we could fit them until they bumped my eardrums (ow!), then remove and slice a fraction of a millimeter off to make them shorter, then re-fit them, repeat until they don't touch my eardrums. This gets them as deep and close to the eardrum as possible to eliminate occlusion which they do nicely. The end result is an earplug that attenuates outside sounds yet without muffling or changing the tonal balance.

 

[arnaud]

Wow, I know audiologists typically try to go pass the second bend for custom ear plugs, but all the way to the ear drum is a first! Next step=>mechanical driver on the ear drum itself . More seriously, yes, iems don't see the outer ear effects, which probably explains why standard stereo records can't ever really sound out of your head when played through these.
 
They can excel at depth and placement through for the good BAs I tried / have owned though. I guess either because the stereo effect works a bit (time/phase differences between the 2 channels) and/or the brain uses a lot of the ambient information / instrument decay to recreate a sense of distance and space (we can certainly perceive depth / room size from a monoral recording).
 
In regards to the deep insertion into the ear canal, it seems rather dangerous to me (what if someone bumps into you?). Also, you still have to carry sound waves from the BA all the way to the mouth of the shell, this duct will have similar kind of resonances along the length as your actual ear canal. Maybe the benefit is in having the iem shell canal about same length as your ear canal so as to get the resonances at about the same frequencies?
 
For the Etys, they work surprisingly good for the money, especially in light to the price escalade in the last 5 years in a world where 3-4 driver BA units rule .

 

[MRC001]

... you still have to carry sound waves from the BA all the way to the mouth of the shell, this duct will have similar kind of resonances along the length as your actual ear canal. Maybe the benefit is in having the iem shell canal about same length as your ear canal so as to get the resonances at about the same frequencies...

What is "BA"?
 
Given this difference, it would seem that the frequency response of over-the-ear headphones should be designed for a different HRTF than IEMs. And reviewers should use a different HRTF when measuring them.

 

[arnaud]

BA stands for Balanced Armature, a type of transducers the ety's and many iems use.

You bring a good point about target curve for iem vs. over the ear. I am not familiar with the details there unfortunately but my intuition is that while the target does not need to be different, the transducer response must be more significantly manipulated in case of iem due to the need to reproduce the effect of the outer ear while over the ear phone doesn't.

 

[MRC001]

BA stands for Balanced Armature, a type of transducers the ety's and many iems use.

You bring a good point about target curve for iem vs. over the ear. I am not familiar with the details there unfortunately but my intuition is that while the target does not need to be different, the transducer response must be more significantly manipulated in case of iem due to the need to reproduce the effect of the outer ear while over the ear phone doesn't.

Etymotic also makes moving coils. I'm listening to an MC-5 right now. It has ruler flat response, very low distortion and lifelike reproduction of natural acoustic sounds, just like the Etymotic BAs. The only drawback to the MC is it doesn't reproduce the top octave. I wonder if that is the benefit of BA over MC - high frequency response.
 
Subjectively, the lack of top octave doesn't impact voicing or timbre. Most of that and music energy in general is < 10k Hertz which the MC-5 handles with aplomb. The impact is a lack of "space" or "air" and fine HF details in small percussive sounds like castanets. Most recordings aren't good enough to capture these anyway, but the lack of these freqs is evident on very fine recordings.

 

[higbvuyb]

  What is "BA"?
 
Given this difference, it would seem that the frequency response of over-the-ear headphones should be designed for a different HRTF than IEMs. And reviewers should use a different HRTF when measuring them.

Both headphones and earphones are measured at the eardrum level and then compared to an eardrum-level reference, so you don't need a different hrtf model.
 
The main difference is that a headphone goes:
headphone response > dummy head outer ear > ear canal simulator > eardrum level
Referenced to:
speaker response > dummy head outer ear > ear canal simulator > eardrum level reference
 
So, in theory, the transfer function of the dummy head's outer ear cancels out, and so you actually listen to your 'own' outer ear.
 
IEMs skip the outer ear stage so it's almost like you're listening through the dummy head's outer ear rather than your own (and there is some individual variation which is lost)
 
Though on the other hand IEMs don't suffer from all the positioning variation that headphones have.

 

[MRC001]

  Both headphones and earphones are measured at the eardrum level and then compared to an eardrum-level reference, so you don't need a different hrtf model.
 

Makes sense. In that case, headphones and IEMs must be designed with a different HRTF - that is, they must produce different frequency response curves to "sound" the same. But they can be measured using the same HRTF.

 

[jf4828]

Sorry to dig up an older topic.  I've been researching this topic a bit recently and thought I'd share some neat resources.  Found on the "HRTF" wikipedia page, there are a list of HRTF databases at the bottom of the article.  If you have access to Matlab, you can easily graph HRTF's (derived from head related impulse response measurements) for a relatively large sample of people at various elevations and azimuths using the material from UC Davis.  It includes ITD etc.  It's a really nice set of data and all you should need to gain a good understanding of what's going on.  I'm actually trying to do the opposite.  I've been designing loudspeakers for quite some time and I'd like to see if I can use DSP to make speakers sound more like some of my favorite IEMs.  I've heard some good speakers, but I've never heard mids/highs like I have on some good headphones....  I know there may be other reasons for this but I'd like to give it a try to see what happens
 
Anyways, the HRTF's for people vary quite a bit.  The difference between stand mounted and head mounted measurements from the OP do not match the UC davis data at all.  As far as I understand, the UC Davis data is blocked canal and accounts for pinna, head, torso etc.  I've attached 3 graphs showing the HRTF of 3 random individuals from that data.  Note that I chopped the frequency but the first dip happens right after kHz and the peak on the first and 3rd is at ~4khz.  If you go through the full database, you may gain an understanding of why some people find sound more annoying then others; these were random, not the outliers!  I can imagine a +/-10dB variation at 3khz might make you go from loving to hating music  
 
This is such an interesting topic.  I feel like there is so much untapped potential in a deeper understanding in this arena....

 

 

 

[RRod]

With things like Oculus Rift coming out, I think we'll be seeing more about 3D headphone spatialization. There recently were new standards released for 3D audio as well (see here, but perhaps you've seen it already). As you noted, the effects from our ears/body that our brains expects are quite individualized, so hopefully we see systems come out that allow for better customization but that are easy enough for end-users to handle.

 

[MRC001]

I'm not surprised that individual HRTFs are so different. No wonder even well trained listeners disagree on what kind of microphones, recording, headphones, speakers etc. sound most like the real thing. No recording or playback is perfect, though some are closer to perfection than others. Different people are more or less sensitive to different audible artifacts, however subtle they may be.
 
On top of that there is the completely separate question of preferences. Many people prefer one kind of sound to another, even if it is less accurate (meaning it sounds less like the real thing). And even "the real thing" is a moving target. Even live acoustic music sounds quite different depending on room in which it's performed, where the musicians are and where the listener is.

 

[arnyk]

  I'm not surprised that individual HRTFs are so different. No wonder even well trained listeners disagree on what kind of microphones, recording, headphones, speakers etc. sound most like the real thing. No recording or playback is perfect, though some are closer to perfection than others. Different people are more or less sensitive to different audible artifacts, however subtle they may be.
 
On top of that there is the completely separate question of preferences. Many people prefer one kind of sound to another, even if it is less accurate (meaning it sounds less like the real thing). And even "the real thing" is a moving target. Even live acoustic music sounds quite different depending on room in which it's performed, where the musicians are and where the listener is.

 
 
Individual HRTFs are largely dependent on the size and shape of heads that vary greatly in real life. 
 
The ear's response to headphones and earphones similarly are subject to individual differences in physiology. Some people who fit hearing aids have references about where to file off some plastic if there are certain kinds of situations with the customer's ears.
 
Preferences seem to converge on flat response of a kind, but the ear is marvelously adaptive and again tolerates nominal imperfections well. Thus, we can still hear that tiger sneaking up on use, no matter what kind of field of plain we are in.