Fast and accurate interpolation of frequency domain response from non-uniform sampling

Field computation of wideband frequency response is very costly, especially when high resolution data is required to capture the sharp peak or null structures in frequency domain. In this paper a fast interpolation algorithm is proposed to accurately reproduce the high resolution frequency data from sparse and non-uniform samples. Based on Cauchy method, the system frequency response is represented by a ratio of two polynomials, whose coefficients can be estimated by using statistical approaches such as total least square (TLS). Nonuniform sampling is implemented in Cauchy method such that new samples at arbitrary frequencies can be added to the existing sample pool to continuously improve the interpolation result. In this proposed algorithm, the density of sample points is automatically adjusted according to the curvature variation of the frequency response, so that accurate frequency response is generated while the least sample points are taken to minimize the computational load.