jgd@acl.lanl.gov (Jerry Glen Delapp) (03/16/90)
In article <3215@pur-phy> murphy@pur-phy (William J. Murphy) writes: Path: lanl!cmcl2!yale!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!cica!iuvax!noose.ecn.purdue.edu!mentor.cc.purdue.edu!pur-ee!pur-phy!murphy From: murphy@pur-phy (William J. Murphy) >Well perhaps. But since there are *many* similar boards available for >a PC, I don't see why they would switch? > >JAV Let me give just one simple for instance why I would switch in a second (if I had the funds to buy a new system). Try doing an FFT on the PC that is 32K array of doubles. Let's see that would be 256K for the signal and 256K twice for the returned arrays. (complex don't ya know) That makes 768K of data. Now consider what kind of crap you will have to use to program the EMS or EMM memory on the PC versus the simple act of mallocing on the Amiga and just doing the calculation? That is enough reason for me. I know, I have tried it on both, the Amiga worked without a hitch. Now I am waiting to buy the 68030 expansion card to make such a calculation feasible timewise. If you don't mind a slightly strange format for storage, you can compute the FFT in 256K for input and 256K for output, by performing a complex FFT on the real data, followed by an unravel function. The trick is that, in a real-to-complex FFT, the first and middle values of the output both have zero imaginary component, and the second half is the complex conjugate of the first half (in reverse order). The method described above doesn't store the second half since it's easily generated from the first half, and uses the imaginary component of the first value to store the real component of the middle value. I don't remember the exact unravel calculation off the top of my head, but it was no more complicated than a radix-2 FFT pass. I think that this would allow you to wedge this calculation into memory without resorting to EMS or EMM (until you wanted to unpack it). You can still get the power spectrum from this format very easily.