The series of consecutive time signals after sampling and truncation is Xn (n), and its spectrum function Xn (ejw) does not change with the completion of the end of the series, the signal frequency resolution is fs/n. the completion of zero at the end of the sequence can only increase the resolution of the signal spectrum display. In other words, if the information in the signal spectrum is distorted due to the leakage or mixing of the spectrum during discretization or window Truncation in the time domain, then no matter how to complete the DFT, can no longer recover lost information.
To increase the frequency resolution of a signal, only increase the sampling frequency of the signal or increase the truncation length N (longer the signal duration) of the sequence ).
1) Zero-padding after data ------- the signal frequency resolution cannot be improved
Although the signal spectrum does not change after the completion of zero at the end of the sequence, after the completion of zero at the L-point DFT, the calculated spectrum is actually the original signal spectrum at [* PI) intervals are sampled at intervals of equal intervals, which increases the number of points for real spectrum sampling and changes the position of the sampling points. This will show more details about the original signal spectrum. Therefore, adding zeros after data can overcome the fence effect.
2) zero data interval ------- the signal frequency resolution cannot be improved
3) Data Interpolation
It is equivalent to improving the sampling rate of the signal and improving the frequency resolution of the signal.