# FFT analysis for series of different length

i have some data series in time domain which are in different length. i'm using FFT to transform them to frequency domain and now i want to calculate their average. Since their lengths are not equal i'm doing: (1) normalized the series before applying FFT (2) appending zeros in the end of sequences to complete the max series length

however, once i'm applying inverse FFT after averaging the series, i'm getting discontinuous results. the result up to the point of equal length is exactly the average, but from the point of the the difference it's completely out of bounds.

• What does it mean to average two sequences of different lengths? If you just want to zero pad to same length and then do $(x[n]+y[n])/2$ then why not just do that in the time domain? – Batman Jun 18 '15 at 11:28