Analysis of Adaptive Automatic Gain Control for Nonlinearly Amplified 16-QAM

Automatic gain control (AGC) is required for all multilevel modulations to align the received signal with decision boundaries defined by the nominal constellation mapping. AGC implementation in most legacy receivers is near optimal for linear, additive white Gaussian noise channels, but is suboptimal if the received signal is distorted by nonlinear amplification. In this paper we consider judicious scaling of the receiver´s AGC to reduce the symbol error probability (SEP) of nonlinearly amplified 16-QAM. For a given degree of distortion and signal-to-noise ratio (SNR), we determine the SEP of 16-QAM as a function of a real scalar multiple of the matched filter output. The theoretical SEP is minimized with respect to this scaling parameter to provide the optimum receiver gain. An AGC algorithm that converges to this value in steady-state is optimal for both linear and nonlinearly distorted signals. The optimal gain is independent of the AGC implementation and is employed to determine the scaling parameter for specific AGC algorithms. We consider the performance of two decision-directed AGC algorithms commonly implemented in hardware and determine their mean steady-state performance. The SEP performance of these algorithms is investigated over a range of nonlinear distortion levels and a broad SNR range. We demonstrate a reduction of up to 10 times in the SEP by employing the optimally scaled AGC algorithm versus the legacy AGC algorithms. The analytical results were corroborated by a real-time emulation test bed with a space-qualified traveling wave tube amplifier.