Design of limiting amplifier models for nonlinear amplify-and-forward distributed estimation

This paper describes the design of limiting amplifier models to be used in the implementation of non-linear amplify-and-forward distributed estimation systems. The use of nonlinear amplifiers in amplify-and-forward distributed estimation can yield significant gains in sensor power efficiency compared to the more conventional linear amplifiers. To ensure similar compression characteristics across all individual sensors and thus allow the receiver to perform estimation, predistortion is utilized to fit all amplifiers to a common shaping function similar to their inherent compression characteristics. When designing this shaping function, knowledge of the receiver noise and the accuracy of the amplifier predistortion is critical to estimating system performance. Analytical expressions for the asymptotic variance are derived for two non-linear amplifier models. The two models analyzed are the scaled hyperbolic tangent and the Cann model. It is found the Cann model with low sharpness can improve power added efficiency in a distributed estimation system. A set of Java-DSP functions is introduced to aid in teaching predistortion design tradeoffs.