Adaptive Sampling for Terahertz Time-Domain Spectroscopy and Imaging

We propose an adaptive sampling algorithm to improve the acquisition efficiency for terahertz time-domain spectroscopy (THz-TDS). Most THz-TDS measurements scan the delay line with constant speed and the data acquired have constant time steps. Our algorithm exploits the fact that the useful information within THz signals tends to cluster at certain positions: efficient sampling can be done by adaptively increasing the sample rate in regions containing more interesting features. The algorithm was implemented by programming a linear optical delay line. Depending on the experiment parameters, the sampling time of a pulse can be reduced by a factor of 2-3 with only slight degradation in accuracy, possible sources of error are discussed. We show how adaptive sampling algorithms can improve the acquisition time in applications where the main pulse is the primary concern.