Quote:
I liked the sound enough to have played it four times so far. I do believe the KU 100 is very seriously 'broken' for 3d sound localization though.
I really must get a copy of How to Win Friends and Influence People...
I'm going to go ahead and disagree with you on that one (about the KU 100 being 'broken').
I have used at least four different mannequin heads in my day job(s) over the past 20 years - some costing as much as $25k, and there are (to be fair) differences in localization that can occur among head variants. Most of this comes down to the type of equalization that each mannequin head employs (and to some extent, the presence / absence of a torso). Some mannequin heads have switchable equalization to free field, but this is really only best suited for lab use and to make the mannequin head's pressure response mimic that of a typical free-field lab microphone (i.e. B&K type 4189). Free-field equalization (assumes that the source is directly in front of the mannequin, approx 2m away, anechoic conditions) can be used in certain head variants, and this really affects the timbre when recording / acquiring data in non-anechoic conditions. Clearly, very, very few recordings (for entertainment) are made in such boundary conditions, and this is logical as the absence of natural reverb would make for some pretty 'dry' music.
Conversely, diffuse field equalization is also technically incorrect (which is what the KU 100 is equalized to) for any acquisitions that take place in anything other than a wholly diffuse field (i.e. a reverb chamber suitable for use according to the guidelines outlines in ASTM C-423 or its ISO equivalent). However, for recording, if I had to choose between free and diffuse field equalization (and I had no options to alter the EQ), I would choose diffuse simply because it's closer to the situation / boundary conditions in which one would record music, and further away from the free field equalization curve (which would make recordings overly bright due to the HF lift in the correction).
Again, in my opinion, neitehr the free or diffuse field is technically 'correct', save for using the mannequin explicitly in those boundary conditions (those sound-fields). What makes the greatest sense to me is that the equalization be venue-specific, and it's been my experience that one can get quite close to this when done as a post process via signal processing (convolution). In essence, this allows you to alter the default equalization curve (a correction if you will) to flat in-situ response. What I have found, subjectively, is that when one compares the diffuse field equalized recording to the corrected one, the latter tends to sound more natural - but this is a function of how close the conditions in which one records are to a truly diffuse field; the closer the environment is to diffuse, the less import the in-situ correction seems to have if starting with a diffuse-field equalized system.
The B&K 4100 can be switched to free or diffuse field EQ, but nothing in-between. The variants from Head Acoustics feature free, diffuse, ID (independent of direction) and a 'user' EQ (assuming you have the software to 'talk' to the DSP box). ID tends to be a pretty good 'go-to' equalization, because it only accounts for the ear canal resonance and so seems to me to be a good approach as it negates the effect of the pinna, shoulders, and torso (taking into account only the effects of the ear canal and the cavum (and possibly the cymba - I don't recall right now) .
In all other (commercial) mannequin heads (Cortex, KEMAR (now GRAS), B&K 4100, and Neumann) there is (at least not that I know of) no user-defined EQ, but again, the EQ can become whatever you like as a post process (iwthin the limits of DSP).
However, what I have seen, over and over (for the past 20 years now working in product sound quality) is the importance of context and the visual cortex in terms of imaging - especially when running juried listening tests. I think I mentioned it elsewhere in the thread how when people evaluate binaural playback seated where the recordings were made, the localization becomes much more believable and 'natural'...but that same recording taken out of context sounds different. I think this is a primary factor in imaging, and EQ is an almost-secondary factor (again, there's no rule here, because one has to consider boundary conditions relative the EQ with which one starts).
I base these comments on my own experience using binaural for juried playback in juried studies, but also on similar observations made by those who use the methodology for evaluating performace of in-car OEM audio systems - they usually run their juried tests in a vehicle interior (listener sits in the vehicle using headphoens) as this often yields much more reliable data about the subjective elements of the sound system. That is, when they play the same sounds to jurors outside of the automotive interior, many users report the exact same recording as soundling less 'real' than when they hear the exact same recording (played back at the same (and 1:1) loudness) played back in the vehicle interior. Given that there's no change to the recording, or to the playback boundary conditions (because the jurors wear headphones), it seems pretty well established that context and visual references play a very big role in localization. Mind you, this happens pretty much independent of mannequin type (using similar equalizations).