[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
not equivalent?
From: |
Albert F. Niessner |
Subject: |
not equivalent? |
Date: |
07 Nov 2002 15:46:15 -0500 |
I have two vectors (series and spacing) of 500x1 elements each. If I
then do the following (N = length(series);):
k=0:N-1;
W = e.^(-j*2*pi*(spacing*k)/N);
spectrum = W' * series;
I get some answer. Because of the size of W, it may be worth while to
loop over very large vectors so I can also do it this way:
w = -j*2*pi*spacing/N;
for k=1:N
W = e.^(w*(k-1));
spectrum(k) = sum (series .* W);
endfor;
I then used the same vectors in both cases and get different results --
max (abs((abs(r1) - abs(r2))) < 10^-10 but max(abs(r1 - r2)) > 450. I
have not checked the phase angle because I am not sure that information
would reduce my confusion.
So, why are the results different? Is there some implied precision
difference between the matrix-vector multiplication and multiplying
followed by and accumulation?
Example data has been attached.
Al Niessner
example.oct.gz
Description: GNU Zip compressed data
- not equivalent?,
Albert F. Niessner <=