Vector Fitting of Noisy Frequency Responses via Smoothing Regularization

We present a simple and effective strategy to compute reduced-order rational macromodels from noisy frequency responses. The reference macromodeling engine is the basic Vector Fitting (VF) scheme, which is well known to be sensitive to noise in the training data. This problem is here avoided by augmenting the VF cost function with a penalization term related to the second derivative of the model, which effectively acts as a regularizer. The results obtained on a set of noisy measurements of a Surface Acoustic Wave (SAW) filter demonstrate the effectiveness of proposed approach in rejecting noise and producing smooth models.