Modelling and Compensation of Nonlinear Distortion in Ultra Wideband Step Frequency Radars

Abstract

Non-ideal behavior of ultra wideband step frequency radars cause artifacts in the etimated impulse responses. These artifacts are discribed by a mathematical model where real-time nonlinearities can be exactly represented by a linear, time-warying model in an implicit time domain. The age and crosstalk between the transmitter and receiver. Simulations show that sompensation methods based on the model with parameter estimation from calibration measurements compensate for the modeled radar artifacts. These compensation methods represent a generalization of a well-known approuch for equalization of linear distortion in step frequwncy radars.

The model and the compensation methods have to be extended to work with radars with IQ imbalance and radars that only sample the real part of the frequency response. Abrupt changes in the frquency response of the radar equipment can be a problem for real domain radars using a Hilbert transformer for retreivel of the quadrature component, but the proposed calibration methods take this into account. In addition, when only a single calibration unit is available, a smoothness constraint is added to the parameters to be able to estimate a complex model for linear distortion.

The suggested methods have been tested and confirmed both by simulations and laboratory tests. In simulations the proposed compensation methods were outperformed by simpler legacy methods when using few calibration units, but when many calibration units were used, the proposed methods were better. The same trend was observed in a laboratory test. As the number of spurious components can be quite large, it is not feasible to compensate for all of them. The legacy methods do not take asny spurious components into account.

The suggested methods compensated satisfactorily for crosstalk and linear distortion in the laboratory test, but only a slight improvement has been observed when compensating for a single spurious component. It is assumed that more spurious components have to be included in the compensation to achieve significant improvements, but then the model or the parameter estimation methods have to be improved to not require so many calibration units that the method becomes infeasable.