ADC Bit Optimization for Spectrum- and Energy-Efficient Millimeter Wave Communications Jinseok Choi, Brian L. Evans, and Alan Gatherer Project Supported by Huawei MOTIVATION I III POWER CONSTRAINED BIT ALLOCATION Algorithm: Power Constrained ADC Bit Allocation Challenge 1: Optimal # of ADC Bits MmWave Massive MIMO Large bandwidth to achieve multi-gigabit data rates Small antenna sizes due to high carrier frequency Large antenna arrays to compensate large pathloss Set power constraint p Sort channel gains to be Compute Mmax Set Binary Search at stage s with Convex optimization problem # of activated RF chains Modeled Psw KKT condition Real number relaxation Millimeter Wave Spectrum [Pi11] Closed-form optimal solution : Aggregated channel gain at jth RF chain For Goal Reduce uplink power consumption at base station Need to reduce power consumption at ADCs Challenge 2: Non-Linearity in Total Power Consumption non-linearity compute bs compute Optimal solution Enforce positivity & append zeros unknown optimal Optimal Nact ▷ binary search O(log(Nr)) Approach Exploit beamspace sparsity in mmWave MIMO channels : Apply analog processing (beamspace projection) ADC bit allocation subject to a total power constraint : Some ADCs/RF chains will be turned off to save power : Other ADCs will have a variable number of bits compare total quantization error for If Ms has minimum total quantization error # of activated RF chains at search stage s Non-linear ADC resolution switching power ▷ training switching power model map of Ms to nearest integer, then return else go to smaller half MODELS & ASSUMPTIONS II IV VALIDATION & CONCLUSIONS * ULA: Uniform linear array Multiuser Massive MIMO Uplink Ergodic Spectral Efficiency w. MRC Energy Efficiency Nu users, each with single antenna Narrowband channel Nr ULA* antennas at base station (Nr >> Nu) Known channel state information at receivers Received signals after analog combining y yq Tx power User symbols AWGN Analog combiner Hybrid receiver with adaptive-resolution ADCs Millimeter Wave Channel Quantization Model L major propagation paths (limited scattering) Array response vector under ULA Additive quantization noise model Variable number of quantization bits bi for each ADC Angle of Arrival Complex path gain Array response vector Conclusions Quantization noise Quantization gain matrix Beamspace channel Pathloss Achieves highest spectral/energy efficiency for low-resolution ADCs Eliminates most of quantization distortion with min. power consumption Enables existing state-of-the-art digital combiners to be employed Allows more power consumption for downlink communication System parameters Cell radius Min dist. Noise fig. fc BW 200 m 30 m 5 dB 28 GHz 1 GHz # rx ant. # RF chains # users # paths Tx Power 256 128 10 13 20 dBm where where Personal homepage https://sites.google.com/site/jinseokchoi89/ November, 2017