I'm trying to distinguish eletrical engines by the magnetic field they create. I have no idea whether this is a smart idea, it's possible at all, whatsoever. However it brought me to a basic question about the Fourier transformation to which I couldn't find an answer that I understand :)
I am sampling with a magnetometer at 160Hz. I'm taking the 160 samples I get each second and calculate the DFT. When I plot the DFT over time I see pretty much a flat surface along all frequencies. The amplitudes vary depending of the motor or distance of the magnetometer but I am not seeing what I was hoping to see - some characteristic frequencies of the motors to reliably distinguish them.
This got me wondering whether it is necessary to get rid of frequencies that cannot be detected before applying the fourier transformation. As I understand it, sampling with 160Hz allows me to detect frequencies up to 80Hz. Would it be necessary to use a lowpass filter and get rid of all frequencies below 80Hz? A frequency of e.g. 160Hz is a multiple of 80Hz, 40Hz,..., so wouldn't it "leak" into the 80Hz, 40Hz,... buckets?