Quote Originally Posted by skeptic
IMO, the difference between the sound field at a live performance and a binaural performance is the fact that the live performance creates a vector field where each element of sound whether arriving at you directly from the instruments or reflected as an echo is a vector having both magnitude and direction. It is therefore external to you and as you move your head, your brain can draw conclusions about its source. Binaural sound consists of two scalar fields having only magnitude. This may sound like a fine distinction but it isn't. It is the crucial difference. It is very difficult to recreate a vector field from a series of scalar fields especially if the source for the scalars, the headphone drivers, moves with your head.
But this is only a dynamic, in relation to your movement. Let's say your head is held in one place by a vise, and listeing to a speaker playing an high frequency band limited warble tone, for example. The vector of the reflections/origin in the room will remain fixed. Let's add some variable: have you and your viced head on a turntable rotating at a fixed rate. the sound source itself is stationary, your sensory system is the variable. I completely agree that the movemen of the head(which is of course natural) while listening is important for a portion of the localization cues, as i made note of in an eariler response. However, the way the ear produces a directional cues is by the specific combing/filtering effects it produces in high frequencies at persepctive angles relative to the source(HRTF=the complex ineraction of the canal, pinnae, head shape, ear shape/size). It seems to me, that the head sensory system is dynamic variable in this instance, not the source it self. If a sound is produced over these headphones to match the HRTF of the listener using a dynamic DSP processer, and linking a rate change sensor to the headphone to control/direct the DSP, the localizatin cues can be just about as good as the 'real' thing. It is conceivable(though I can't point to a completelely succesful example yet) that this could be applied to a linear recording of an acoustic event(using still proper space/timing of micrpohones to represent head, this is critical--though deletion of a specific ear/canal model is probably feasible, i believe), and the comb/filter effects nescarry applied by a DSP according to the rate/change sensor that measures your head position, applied to a headset to emulate correct perspective.

-Chris