Wideband Channel Estimation for Hybrid Beamforming Millimeter Wave Communication Systems with Low-Resolution ADCs
Millimeter wave (mmWave) communications is a promising technology to increase data rates to users. In mmWave communications, large antenna arrays are needed to form directional beams to overcome the high pathloss. Power consumption becomes a significant challenge. By reducing the number of RF chains and adapting the analog-to-digital converter (ADC) resolution in each RF chain, power consumption can be dramatically lowered.
One receiver architecture to achieve both power reduction goals uses analog combining and digital beamforming. The output of the analog combiner is sparse. That allows the number of RF chains and the number of ADC bits to be tailored to teach beam. Some beams have such low received signal-to-noise ratio that no RF chain would need to applied to it.
In this hybrid beamforming architecture with low-resolution ADCs, conventional channel estimation algorithms such as least squares and minimum mean square error estimation not perform well because they do not exploit the sparsity in the channel impulse responses.
For wideband channel estimation in mmWave systems, WNCG Professor Brian L. Evans and students Mr. Junmo Sung and Mr. Jinseok Choi have evaluated compressed sensing algorithms such as orthogonal matching pursuit (OMP) and generalized approximate message passing (GAMP) as well as the least squares (LS) estimator for comparison purposes. Through simulation, they have shown that one-bit ADCs using OMP outperform infinite-bit ADCs using LS in an SNR range of interest. Using OMP, low-resolution such as 3 and 4 bits can achieve channel estimation performance close to the infinite bit resolution case across the SNR range of -10 dB to 15 dB.
This research was supported by Huawei Technologies.