To: atm{at}shore.net
From: mdholm{at}telerama.com
Reply-To: mdholm{at}telerama.com


Recently I translated a 2D-FFT routine from C to VB5.  I haven't tested it
or made any use of it yet, but, assuming it works, it will be the core of
what you need to do the star test simulation (I think.)  I was thinking of
doing something similar.  Don't know if I have time.

Jim Burrows sent me some basic info on how to use the 2D-FFT to calculate
diffraction patterns.  I have it somewhere.  One thing I don't have much
idea about is how to include the effects of spider and central obstruction.
 As Jim related it, a typical FFT uses a rather course grid of sample
points.  With that coarse sampling, it would be impossible to directly
model sipder vanes, and the central obstuction would be pretty crudely
modeled.

One little point is that the FFT routine wants arrays with sizes in powers
of 2.  So at some point, the array will have to become 256x256 or 512x512. 
This can be done by scaling, or by padding.  Also, the FFT wants complex
numbers. You get these by calculating the path lengths from the mirror to
the image point.  It is complicated to explain, but not to hard to write
once you know the math.

There are, according to Jim, some rather non-obvious relationships between
resolution (of the model) and array size.  The two are inversely related
across the FFT.  (I hope I haven't got that screwed up.)

I think doing the simulation for a single star and no spider or obstruction
is within my capability, though it will be far from a fast process.

Obviously, others have solved this problem with all the details in place.

Mark Holm
mdholm{at}telerama.com

--- BBBS/NT v4.00 MP