Normally you find the frequency content by using a discrete fourier transform. This takes a number of data points and transforms them from time domain to freque

Normally you find the frequency content by using a discrete fourier transform. This takes a number of data points and transforms them from time domain to frequency domain. It's basically a mapping from one vector space to another. And the frequency domain vector space has fixed base vectors. The basis vectors correspond to frequencies, and being basis vectors they are actually independent of the data. What is data dependent is how much energy those frequencies have. A given window size will have a minimum frequency it can detect. So the low frequency cutoff is actually implicit. If you want to get lower frequencies you have to use longer windows, and that loses temporal precision, so you have to do a tradeoff. In a fascinating twist that is actually closely related to the heisenberg uncertainty principle.