Hi all, 
I'm convolving a vector x with another vector h, both of length N. I need a matrix corresponding to convolution with h, not just conv(x,h), so I'm assembling the discrete convolution matrix A as follows

M = 2*N-1;
A = sparse(M,N);
for i=1:N
        A(i:min(M,i+N-1),i) = h(1:min(N,M-i+1));
end
%convolution would happen as y = A*x, but I just need A

However creating this matrix is very slow. Anyone have any tricks to creating A quickly? I can't think of any way to vectorize this any more

Please see post #8 
http://www.mathworks.com/matlabcentral/newsreader/view_thread/292941

Bruno