Standing wave pattern suppression in the terahertz frequency domain spectroscopy by means of windowed Fourier filtering

Terahertz frequency domain spectroscopy provides a superior resolution in the broad range from 50 GHz to 5 THz. However, temporal frequency drift and the standing wave pattern in the spectrometer, which often act simultaneously, produce intensive coherent noise in the transmittance spectra of the measured samples. The paper presents a spectrum processing method allowing to reduce the coherent noise and thus significantly enhance the accuracy of spectral data analysis. The idea of the method is to remove the small-period (0.25 GHz) oscillations in the standing wave pattern of the measured signal by applying the windowed Fourier filtering. The large-period (4 GHz) oscillations in the measured spectra are broadened and then used to compensate the frequency drift of the two consecutive measurements of the signal with the sample and without it. The proposed approach is tested on the measured spectra of metamaterial and silicon wafer. Its advantage over the classical method based on the averaging of adjacent data points is confirmed. Our work benefits to optimization of the frequency-domain terahertz systems and paves the way for fast and accurate analysis of spectral data.