Adaptive Automatic Gain Control for Nonlinearly, Distorted Constellations

Automatic gain control (AGC) is a key part of any demodulator that supports multilevel linear constellations. The AGC scales the received signal such that the constellation points line up with the nominal mapping. The scaling is adjusted to minimize the average distance between the received constellation point and its hard-decision equivalent. This approach has been shown to be optimal for undistorted constellations. However, it is not optimal for constellations distorted by a nonlinear amplifier. This paper investigates the relationship between AGC scaling and nonlinear distortion for coded and uncoded communications links. We present both simulation and experimental results that demonstrate significant improvement in the bit error rate (BER) by changing the AGC adaptation algorithm. The decision directed AGC algorithm is augmented with a least-mean-square (LMS) algorithm to compensate for the nonlinear distortion. This simple change can be implemented on legacy receivers with negligible increase in computational complexity.