A Two-frequency Subtraction Method to Improve Spectral Peak Identifications

We discuss a two-frequency subtraction technique to reduce the energy leakage in a Fourier spectrum. In our method, the two strongest frequency components are determined by finding the periodogram over an interval such that the two frequencies will not interfere with each other. Such a method allows the subtraction of the two main frequency components more accurately from the original signal. The energy leakage from the main components is minimized to allow identification and more accurate determination of weaker components. Statistical error from the subtraction technique can be several times smaller than the FFT method. We show that the subtraction method is relatively robust for signals with varying amplitude or frequency.