Channel Estimation and Hybrid Precoding for Millimeter Wave Cellular Systems


Ahmed Alkhateeb, Omar El Ayach, Geert Leus, and Robert W. Heath, Jr.


Submitted to the IEEE Journal of Selected Topics in Signal Processing, September 2013.


Millimeter wave (mmWave) cellular systems will enable gigabit-per-second data rates thanks to the large bandwidth available at mmWave frequencies. To realize sufficient link margin, mmWave systems will employ directional beamforming with large antenna arrays at both the transmitter and receiver. Due to the high cost and power consumption of gigasample mixed-signal devices, mmWave precoding will likely be divided among the analog and digital domains. The large number of antennas and the presence of analog beamforming requires the development of mmWave-specific channel estimation and precoding algorithms. This paper develops an adaptive algorithm to estimate the mmWave channel parameters that exploits the poor scattering nature of the channel. For single-path channels, tools from compressed sensing are leveraged to derive a lower bound on the estimation success probability using the proposed algorithm, and find the optimal training power allocation among the adaptive stages of the algorithm to maximize this success probability. These tools also provide a mathematical basis by which the proposed algorithm is extended to tackle the multi-path mmWave channel estimation problem. Using the estimated channel, this paper proposes a new hybrid analog/digital precoding algorithm that overcomes the hardware constraints on the analog-only beamforming, and approaches the performance of digital solutions. Simulation results show that the proposed low-complexity channel estimation algorithm achieves comparable precoding gains compared to exhaustive channel training algorithms. The results also illustrate that the proposed channel estimation and precoding algorithms can approach the coverage probability achieved by perfect channel knowledge even in the presence of interference.

Available on IEEE

The Matlab code of the paper is available here