Research On Key Techniques In Wireless Receiver With Low-Precision Quantization

As the date rates and bandwidths of communication systems scale up, the cost and power consumption of high-precision (e.g.,8-12 bits) ana-log-to-digital converters (ADCs) become prohibitive. One possible approach to relieve this bottleneck is to redesign communication systems with the starting assumption that the receiver employs ADCs with drastically reduced precision (e.g.,1-4 bits). Recent information-theoretic analysis for the AWGN channel shows that, the use of 2-3 bit ADCs leads to only a small but accepta-ble degradation in channel capacity. Furthermore, a sensible choice of uni-form pulse amplitude modulated (PAM) input, with quantization thresholds set to perform maximum likelihood hard decisions, achieves performance close to what an optimal input and quantizer pair attain, which makes the low-precision ADC easier to be realized. This has motivated more detailed investigations of signal processing for key receiver functionalities when ADC precision is reduced, including problems in carrier synchronization, channel estimation, equalization and so on.Currently, the research on low-precision signal processing techniques usually assumes an ideal automatic gain control (AGC). However, as the de-cline of the ADCs'precision and the rise of the accuracy requirement in AGC, how to perform an accurate gain control with a low-precision ADC is a cru-cial problem. The research on AGC with low-precision ADC is still in its ear-ly stage. This paper investigates the AGC problem with low-precision quan-tizer with a goal that the input signal can exactly match the quantizer with an effective estimate, which can maximize the use of the finite precision quan-tizer and make the low-precision ADCs practicable. In addition, we also study the channel estimation algorithm for fading channels with low-precision quantizer. The main research and contributions of this paper are as follows.First of all, a maximum likelihood (ML) estimator for the input signal amplitude of PAM signal is proposed and the gain is controlled based on the estimation. In addition, by minimizing the normalized mean squared error (NMSE), we derive the optimal dither, with which the performance in high SNR range is improved. With the assistance of AGC, bit error rate is effec-tively reduced and we can almost achieve the capacity of the ideal gain con-trol.Considering the relationship between the quantizer and the input signal amplitude, we derive the optimal quantizer for amplitude estimation of PAM signal. Based on this optimal quantizer, the paper proposes an iterative esti-mator under the maximum a posteriori (MAP) criterion. The input signal amplitude is dynamically adjusted based on the MAP estimation using in-formation of the received sequence to match the optimal quantizer, which improves the estimation performance.In order to achieve a good estimation performance, the thresholds are designed to maximize the mutual information between the observations and the input signal amplitude and a greedy estimation algorithm is proposed. Different from the MAP estimator, this method considers the distribution of the input amplitude. Due to the dynamic change of the output distribution, the thresholds need to be optimized after each observation and results in a high complexity. However, the performance of this greedy algorithm achieves sig-nificant improvement and with the training sequence length N=40, the per-formance is closed to that of the ideal gain control.Giving consideration of both performance and complexity, a particle fil-ter based estimation algorithm is proposed. First, a set of random samples is generated to represent the distribution of the input signal amplitude and the thresholds for the following signals are designed based on these random sam-ples. Then one can update the distribution according to quantization outputs, adjust the particle positions to represent the new distribution and set up a new set of thresholds. This particle filter algorithm has a large performance gain compared with MAP estimator and ML estimator, while the complexity is ef-fectively reduced compared to the greedy algorithm.Furthermore, a channel estimation method with optimal dither for low-precision quantizer is proposed and its performance is analyzed. The dither is designed based on the MAP criterion. An effective estimation can be obtained with the help of dither, which is close to the performance of the full precision minimum mean square error (MMSE) estimator with a short train-ing sequence.In conclusion, we proposed four signal amplitude estimate algorithms for automatic gain control with low-precision quantizer:ML estimator, iterative MAP algorithm, particle based estimation algorithm and greedy method with optimal quantization threshold. These algorithms can obtain an accurate esti-mation of the input signal amplitude with short training sequence, based on which an effective gain control can be made to guarantee the following signal processing progresses. In addition, the paper verifies channel estimation method with a low-precision quantizer and the results show that an acceptable performance can be achieved with a short training sequence. The research on the AGC with low-precision quantization in this paper lays the foundation for the application of low-precision ADC in a communication system receiver.