Characterization and modeling of drift noise in Fourier transform spectroscopy: implications for signal processing and detection limits

Abstract

A theoretical analysis of long-term drift noise in Fourier transform spectroscopy is presented. Theoretical predictions are confirmed by experiment. Fractional Brownian motion is employed as a stochastic process model for drift noise. A formulation of minimum detectable signal is given that properly accounts for drift noise. The spectral exponent of the low-frequency drift noise is calculated from experimental data. A frequency-dependent optimal spectrum averaging time is found to exist beyond which the minimum detectable signal increases indefinitely. It is also shown that the minimum detectable signal in an absorbance or transmission measurement degrades indefinitely with the time elapsed since background spectrum acquisition.