Problem with complex number multiplication and division

Hello, thanks for looking at this, this is potentially a very stupid question.

I'm having a problem with signs for complex conjugates when I multiply complex numbers two ways. In the first way, I have a simple calculation:

    Rws(n,m)+j*w(m)*Iws(n)-Cws(n,m)*j/w(m)

to give me a value (in this case, it was 1.366e+09 + 5.937e+09i). My matrices are [9 800] in size, and I wanted to not do this in loops, but instead vectorize the calculation. So, instead, I change it to this:

    Rws+j*Iws*w-bsxfun(@rdivide, Cws*j, w);

and I get the same answer with a small twist: the sign on the complex number was switched (1.366e+09 - 5.937e+09i). This change in sign is prevalent in all other members in the matrix, but the calculation was identical. In this, Rws is a 9x800 matrix, Iws is a 9x1 vector, w is a 1x800 vector, and Cws is a 9x800 matrix.

I can bypass this problem by switching the addition with the subtraction in the second equation, but I wanted to know if there was a reason I was getting this mismatched sign. Any ideas?

Thanks for your help, I think I'm just missing something simple here.

EDIT: Included some matlab files for clarification, use run_it to run the program, check for the variables _forloop and _vectorized for reporting.

[David Young]

I can't replicate the problem. This code gives identical results for Result1 and Result2.

Rws = rand(9, 800);
Cws = rand(9, 800);
Iws = rand(9, 1);
w = rand(1, 800);

Result1 = zeros(9, 800);
for n = 1:9
    for m = 1:800
        Result1(n,m) = Rws(n,m)+j*w(m)*Iws(n)-Cws(n,m)*j/w(m);
    end
end

Result2 = Rws+j*Iws*w-bsxfun(@rdivide, Cws*j, w);

[John D'Errico]

My guess is you transposed something along the way. Since you have not actually shown us an explicit example, so we cannot know.

[Brian]

First off, thanks to both for answering. I'll go into a little more of my problem. I can attach the m files I was working on modifying, if that will help you get an idea.

I took a look through it again, and the only thing I can think of is its a multiplication problem when I string along my vectorization.

I included the original MATLAB files: you can run the program with run_it, and the variables to look for are compliance_forloop, compliance_vectorized, inertiance_forloop, inertiance_vectorized. Zws is just a sum of resistance with those two components.

I can bypass this problem by switching addition with subtraction, but I'd rather try to understand the problem.

Thanks!