Research On Massive MIMO Receivers Under Low-resolution Quantization

By deploying tens of to hundreds of antennas at the base station(BS)and simulta-neously serving tens of single-antenna users in the same time-frequency resource block,massive multiple-input-multiple-output(MIMO)is able to achieve unprecedented gain in both spectral efficiency and energy efficiency.As a result,it is envisioned as a key technology to meet the stringent requirements of the next generation cellular network and has attracted highlighted attention in the recent several years.Regarding to hard-ware cost and circuit power consumption,however,the large antenna array deployed imposes great challenge on the system design and will possibly soon become the sys-tem bottleneck.Therefore energy-efficient solutions are desired to realize the promised profits of massive MIMO and meanwhile strike a reasonable balance between system performance and hardware cost as well as energy consumption.Noticing that analog-to-digital converter(ADC)accounts for a large portion of the circuit power consumption of the receiver,and that the power consumption of ADC scales roughly exponentially with the bit-width,low-resolution quantization is therefore well accepted as a promising route to the energy-efficient design of massive MIMO receivers.For quantized multi-antenna systems,there is no closed-form expression for the channel capacity.Thus in this work,we leverage the information-theoretic tool of gen-eralized mutual information(GMI)to derive and further maximize the achievable rates of quantized massive MIMO systems under either frequency-flat fading or frequency-selective fading.Moreover,in addition to linear receivers,we also explore the spectral efficiency gain of nonlinear receiver.The main contributions of our work are summa-rized as follows.1)For uplink massive MIMO systems under frequency-flat fading,we propose a mixed-ADC architecture to strike a reasonable tradeoff between spectral efficiency and energy efficiency.We start by focusing on the deterministic single-input-multiple-output(SIMO)channel scenario.Conditioned on a given linear receiver and a given ADC switch vector,we derive a closed-form expression of the achievable rate,and based on which further optimize the linear receiver as well as the ADC switch scheme.Then we extend the analytical results to ergodic time-varying channel scenario and ex-plore the impact of round-robin channel training and the residual estimation error on the spectral efficiency.We further extend the analytical results to the multi-user access sce?nario,and investigate the tradeoff between spectral efficiency and energy consumption.Finally,numerical results are presented to validate the analytical results.2)We extend the mixed-ADC architecture to the uplink massive MIMO systems under frequency-selective fading.Particularly,we employ orthogonal frequency divi-sion multiplexing(OFDM)to mitigate inter-symbol interference,and propose a linear frequency-domain equalizer to alleviate inter-carrier interference.We also start by fo-cusing on the deterministic SIMO channel scenario.Given a linear equalizer and an ADC switch vector,we derive a closed-form expression of the achievable rate,and in turn optimize the linear equalizer as well as the ADC switch vector to maximize the achievable rate.Then we extend the analytical results to ergodic time-varying channel scenario and investigate the impact of round-robin channel training and the residual es-timation error on the spectral efficiency.We further extend the analytical results to the multi-user scenario,and propose a low-complexity algorithm to reduce the computa-tional burden of the receiver.Numerical results validate the superiority of the mixed-ADC architecture and thus we envision it as a promising solution for the energy-efficient design of massive MIMO receivers.3)We explore the spectral efficiency gain of minimum mean squared error(MMSE)receiver as compared with the linear MMSE receiver.For analytical simplicity,we fo-cus our attention on one-bit quantized frequency-flat SIMO channels.Again we lever-age GMI to derive the achievable rate of the system.Analytical result reveals that the MMSE receiver achieves exactly the same spectral efficiency as the linear one if there are only two receive antennas.Moreover,numerical results show that the performance gain of the MMSE receiver remarkable increases as the number of receive antennas grows large.