[sdiy] ot: rotating speaker simulation or stupid approach

[Richard Wentk]

At 10:05 04/07/2003 +0200, you wrote:
> > The difference is that you're not getting the wavefront compression that's
> > created if you move a single speaker.
> >
> > Look at it this way. If you move a boat through water [...]
> > Technically you're creating what's called a phased array, which has very
> > different properties to a moving point source. A moving sound point source
> > physically, and literally, compresses the spacing of the wavefronts ahead
> > of it. A phased array can be used to steer a beam of wavefronts by using
> > phase cancellation, but it doesn't produce any Doppler effects.
>
>Richard, you would almost convince me. But I have just recently
>built something like that, not even using speakers, just
>a tapped delay line and a scanner to interpolate between
>the different delay times (like the Hammond scanner vibrato does),
>and the frequency modulation is clearly audible.

Okaaaay - but this is not the same as the original idea. Scanning a delay 
line *will* create frequency shifting, because you're effectively moving a 
pick-up wrt to the input wavefronts, which is another way to simulate 
doppler effects. (You can move the sound or you can move a microphone. 
Either way will work.) The fact that you're doing it electronically doesn't 
matter. You're still simulating the effect of compressing wavefronts ahead 
of your pick-up point and stretching them out behind you - in just the same 
way that if you're flying over a sea full of plane waves, waves ahead of 
you will appear to bunch up and waves behind you will appear to spread out, 
because of the relative velocities involved.

But straightforward panning of a sound along an array of speakers won't do 
this. You'll get the effect of varying distance, but that will be it.

Anyway, the original question was whether you could interpolate impulse 
responses. I don't believe you can, because if you try to interpolate 
impulse responses with different lengths you're going to run into all kinds 
of interesting problems, not least of which is that your impulse response 
set is going to vary hugely depending on listening position and relative 
angle from the source.

It would make more sense to close-mike the impulse responses to eliminate 
as much delay as possible, and add a delay with a variable tap to simulate 
the frequency shifts. But in fact that still won't work because what you'd 
really have to do is take the impulse pattern of a *moving* horn, not a 
still one, to simulate the differential effects caused by horn velocity. 
This will be especially true if you put the horn inside a box.

Richard