Frequency Dependent Beamforming - A Time Efficient Solution to the Beam Alignment Problem in Millimeter Wave Communications
Buch in deiner Nähe kaufen
Beschreibung
Millimeter-wave frequencies offer a wide spectrum and high data rates, but their use is associated with high propagation losses. These losses can be compensated for by using directional links between the transmitter and the receiver. Beamforming is hereby key to generate array manifolds with maximum energy transfer at each receiver location to improve the quality of the link. However, in conventional systems, e.g., phased arrays, beamforming is associated with consecutive probing of a large number of potential beams from a given analog codebook leading to large signaling overhead. Especially in dense network situations or fast-changing environments, the classical method of brute force iterative testing of all potential beams limits the overall performance of the system. In this thesis, a beamforming network based on true-time delays that intentionally violates the so-called narrowband assumption so that the generated beam pattern becomes frequency dependent, is proposed as a promising alternative to classical iterative beamformer training. Compared to the brute-force approach, a hardware implementation based on true-time delays of multiples of the reciprocal of the band- width can be used to generate frequency dependent beamformers that simultaneously address all angular directions at different baseband frequencies. We exploit this prop- erty and present low-complexity digital signal processing algorithms to both accelerate analog beamformer selection and provide the most accurate angle estimation possible. Our results will help in highly congested millimeter-wave networks with a large number of users to overcome the limitations of the classical method of iterative beam-forming testing as defined in the 801.11ay standard and we will propose a framework which is able to reduce the signaling overhead of orchestrating all user requirements. Link quality and overall network performance can be improved using frequency de- pendent beamforming, where receivers need only perform a simple spectrum analysis to find their own best analog beamformer or angle of departure estimation. We show how to select the analog beamformer faster than in the classical brute-force approach and how to estimate the beam angle with a mean absolute error less than 1◦ with low hardware complexity, i.e., array sizes of M = [8, 16] antenna elements. A base station equipped with such a frequency dependent beamforming device assists all receiver selecting their best analog beamformers and, ultimately, helps to maximize the number of active users and data sum-rates. It is interesting to note that increasing the number of antenna elements M does not immediately increase the accuracy of the angle estimation. Only the resolution, i.e., distinguishing two individual paths with almost equal angle of departure can be improved. Hereby, the Rayleigh criterion defines a well known minimum angular resolution for given antenna array physical dimension. Increasing the number of frequency sampling points, however, minimizes the quantization error when sampling the power spectrum and, therefore, helps to minimize estimation error.
Autor*in
Christoph Jans