Estimation of RF PA nonlinearities after cross-correlating power supply current and output voltage

Abstract

ee08064, dee05001, jms]@fe.up.pt II. ESTIMATION OF NONLINEARITYIt has been shown that a weakly nonlinear, memoryless, time-invariant, PA's behaviour can be expressed by a thirdorder polynomial [11]. This fact has been explored to develop in-circuit test circuits based in the observation of the input and output RF voltage signals [4,12,13]. Likewise, one can estimate the PA's power transfer characteristic after finding the polynomial Pout = /31 Pin + /32 Pin 2 + /33 Pin 3 that best fits a set of observed input-output power pairs, and whose coefficients allow us to compute Pin1dB and Pin IIP3 points. Following procedures similar to those presented in [11][12][13] one can obtain, respectively, values. From this set of coordinates one can obtain the polynomial that best fits the PA's input/output characteristic and whose coefficients allow one to calculate 1 dB compression ( Pin1dB ) and third order intercept ( PinIIP3 ) points values. The methodology being presented here relies also on the polynomial fitting approach, however, ordinates are provided by the cross-correlation between output voltage and power supply current, instead. The remaining of the paper presents in section II the procedure being proposed to estimate Pin1dB and PinIIP3 from the power transfer characteristic curve, together with validating simulation results. Section III shows how power measures can be obtained from the cross-correlation between voltage and current signals, as well as, the mixer based circuit being used as a correlator. Section IV highlights the main conclusions.Abstract -The present paper describes developments carried-out on estimating 1 dB compression and thirdorder intercept points after the cross-correlation between power supply dynamic current and output voltage ofradiofrequency power amplifier. The underlining theory and a circuit that allows implementing this measurement on-chip are presented. Simulation results, including the analysis of optimum stimuli amplitudes and the Monte Carlo analysis to circuits' process variations are presented. These show that good accuracy can be obtained with relatively simple measurement conditions.