Particle Filtering Based Automatic Gain Control for ADC-Limited Communication

As the date rates and bandwidths of communication systems scale up, low-precision (e.g., 1-4 bits) analog-to-digital converter (ADC) is one approach to reduce the cost and power consumption for sampling rates on the order of gigahertz. Because of the significant quantization error of the low-precision ADCs, the conventional digital automatic gain control (AGC) is not applicable any more. In this work, we investigate the problem of AGC for pulse amplitude modulation (PAM) signaling over the AWGN channel. First, the optimal thresholds are derivated by maximizing the Fisher information. Then a particle filter based estimator is proposed. We divide the training sequence into several slots. In each slot, particles are generated to approximate the distribution of the input signal amplitude, which is updated basing on the quantizer outputs in previous slots. We adjust the thresholds according to the amplitude particles of the input training signal and finally obtain a maximum a posteriori probability (MAP) estimate. We obtain notable performance in terms of uncode bit error rate (BER), compared with one-shot maximum likelihood (ML) estimator with the same training length. The performance is closed to that of ideal AGC over different input amplitude values.