Bias Compensation When Identifying Static Nonlinear Functions Using Averaged Measurements

When estimating the input-output characteristic of a static nonlinear function, input-referred noise may induce estimation bias, if model identification based on simple averages of input and output measurement data is performed. This paper considers estimation of static nonlinearities based on polynomial functions and input-output averaged data. It first illustrates how the estimation bias originates and then it describes a procedure to compensate its effects. Both simulation and experimental results are shown. Experiments are carried out to estimate the voltage-to-voltage characteristic of a diode-based electrical circuit. Practical considerations are made regarding the minimum number of samples needed to perform compensation effectively.