On Achievable Rate For Massive MIMO Under Low-resolution Output Quantization

with the number of mobile Internet users and the number of mobile terminals grow-ing rapidly in recent years,the demands from mobile Internet users to high-definition video and other high-quality services are increasing gradually,which has brought tremen-dous pressure to the existing wireless communication systems and technology.In order to alleviate the pressure,the frequency bandwidth and the number of antennas will be a substantial increase in the next generation of wireless communication systems.Mas-sive MIMO,millimeter wave technology and other technologies greatly improve the energy efficiency and spectrum efficiency.However,such technologies introduce new challenges for the hardware design.A main challenge is the power assumption with analog-to-digital converters(ADCs).The use of few-and especially one-bit ADCs rad-ically changes both the theory and practice of communication.A general analytical framework based on generalized mutual information is applied to the analysis of massive multiple-input-multiple-output systems with low-resolution output quantization.For a given input distribution and decoding rule,the generalized mutual information is the rate below which the average probability of error-aver-aged over the ensemble of codebooks——decays to zero as the block length tends to infinity,and above which this average tends to one.For Gaussian codebook ensem-ble and nearest-neighbor decoding rule,an equivalence relationship is established for general nonlinear transceiver distortion,that the effective signal-to-noise ratio based on the generalized mutual information is consistent with the heuristically derived signal-to-quantization-noise ratio based on Bussgang theorem.Physically,the quantization process is implemented with ADCs,but the nonlinear-ity characteristics of ADCs pose difficulties to the analysis.The additive quantization noise model(AQNM)is a widely used,strong analytical model to approximate the prop-erties of quantizer.Therefore,we use the additive quantization noise model to analyse the approximate rate of the system and compare it with the achievable rate of quantized system.Simulation results show that the approximation rate based on additive quantiza-tion noise model and the achievable rate of quantized system are different in single user scenario,but this inconsistency can be remedied by taking into account the correlation within the quantization noise vector.The parameters of the generalized mutual information involve the combiner,the covariance matrix of the quantized output vector,the correlation vector of the quantized output vector and the transmitted signal.This thesis proposes a linear combiner with lower complexity which matches to AQNM rather than to reduce the system overhead.The simulation results show that this suboptimal linear combiner brings the performance gain close to the optimal linear combiner both in BER and rate.