Non-uniform Data

The FFT assumes that the input data is uniformly spaced. But if you are working with a collection of observational data, it may be sampled at irregular time intervals. In this case, the Lomb-Scargle periodogram is often used.

Tip

A nice introduction to the Lomb-Scargle periodogram is provided in the paper Understanding the Lomb-Scargle Periodogram by J. VanderPlas.

SciPy provides the scipy.signal.lombscargle function which implements this algorithm.