A technique for interpolating a series of samples includes constructing a mathematical model of the series that describes its large signal behavior. The model is subtracted from the original series to yield a residue. A discrete Fourier transform (DFT) is taken of the residue, and the DFT is zero-padded. An inverse DFT of the padded result yields an interpolated residue, which is then added back to the mathematical model to construct an interpolated version of the series of samples. Using this technique, accurate interpolation can generally be attained even when the series of samples is not coherently sampled.