zero padding question

Started by January 31, 2008
I have a question that has been bugging me fo rsome time and I am hoping someone here can help me get to the bottom of it. It is a zero padding & FFT issue.

Here goes;
If I take 2 random signals for deconvolution (say 1000 samples and 1500 samples long with same sampling time). If I zero pad both out to 4096, take FFTs, devide one by another, then take iFFT, I get a graph of one signal deconvolved by the other.

However, I was lead to believe that how much zero padding you do will not affect the fundamental information in the signals. BUT, if I zero pad both signals out to 8192 before FFTing, deviding & iFFT, I get a totally differnt graph with the spikes, etc in a different position.

Does anyone know why zero padding the signals by a particular length gives a different answer to zero padding both the signals by a different amount?