Analysis and Mitigation of SideLobe Degradation due to Quantized control in mmWave 5G Phased Arrays
Abstract
Antenna arrays are one of the most important parts of the RF communication systems. With
the advancement in the eld of 5G mobile communication, electronically steerable arrays, particularly
phased arrays have undergone necessary evolutionary developments in the past decade.
In addition to the advantage of electronic scans using phase shifters or true time delay devices,
phased arrays provide control on sidelobe synthesis. This is possible through the use of a variety
of amplitude weighting schemes that have been developed from polynomials, optimization
algorithms, or sampling of continuous sources. Apart from this, adaptive phased arrays can
spatially lter jamming sources (or strong interference sources) by synthesizing appropriate
array weights which in turn improves the signal to interference ratio (SINR). Phased arrays
may also o er the advantage of collecting incoming signals from di erent directions by forming
multiple beams.
In 5G communication systems, moderately sized phased arrays have emerged as promising
candidates. Though the exact array size may depend on the link budget calculations, recent
studies suggest the use of moderately sized arrays. Three types of arrays have been explored for
5G mobile communications, namely the fully analog phased arrays, fully digital arrays, and the
hybrid analog-digital arrays. The hybrid systems are believed to be the most suitable candidate
for the 5G communication systems due to the
exibility in beamforming using both the digital
and RF beamformers. The use of both the beamformers makes the formation and scanning of
the multiple beams feasible in practice.
RF analog beamformers consist of phase shifters and attenuators, both of which have nite
precision. Besides, the dynamic range of these attenuators is also nite. Phase shifters and
attenuators with 5 and 6 bits are available today. This was not the case in the past, where
phase shifters and attenuators with 1-3 bits were mostly considered due to the high cost,
size, and power requirements of their higher precision counterparts. The quantized control
o ered by the RF beamformers causes impairments in the antenna array patterns such as
degradation in side lobe level (SLL), beam pointing errors, gain reduction, and occurrence of
grating lobes. With the improvement in the precision of phase shifters (5 to 6 bits), the beam
pointing errors are negligible. But, the degradation of the sidelobe level is considerable. Most
of the proposed techniques for the improvement of the sidelobe level are based on the random
searching of quantized phases and amplitudes. Randomization of quantized phase shifts leads
to the derangement of periodic and correlated quantization errors, which leads to the reduction
in the parasitic sidelobe peaks. These random search based algorithms can be computationally
ine cient both in terms of the number of iterations and the convergence characteristics. The
required number of iterations can be very large, in addition to being uncertain. None of these
techniques exploit the high precision of present-day beamformers.
Some of the aforementioned issues are analyzed on this research by considering uniform
linear arrays (ULAs) of various sizes. We also consider the e ects of the nite dynamic range
of the attenuators which have not received signi cant attention in previous works. Although
the majority of this work assumes isotropic radiators as antenna elements, the e ects of the
directional pattern are also included where required.
In the rst part of this thesis, a detailed analysis of the quantized beamforming network
(BFN) is presented. The e ects of amplitude quantization are rst discussed for Chebyshev
and DPSS arrays. An analysis of achievable highest sidelobe level (HSLL) for moderate to high
target HSLL values (25 to 80 dB) is introduced. The study is focused on attenuators with
more than three bits and a xed dynamic range. Various combinations of attenuator bits and
dynamic range are also analyzed. True (unquantized) values for phase shifts were assumed
as the reference for analyzing amplitude quantization. Next, phase quantization e ects are
investigated for phase shifters of 3 5 bits in the whole visible scanning range. The pattern of
HSLL degradation in the scanning range is veri ed by plotting array factors for various scan
angles. This analysis assumes unquantized DPSS amplitudes in all the investigated cases.
An algorithm to minimize HSLL degradation in antenna arrays is proposed. The proposed
algorithm is based on ordered binary decision making on the perturbation of phases by LSB
of the phase shifter. Quantized values of DPSS amplitudes are used for all the results. The
convergence characteristics of the proposed algorithm and its computational complexity are
also discussed in detail. It is also observed that although the proposed algorithm converges in
only two iterations, it works only for larger arrays (N > 32). Thus, a new technique is required
for small and moderately sized arrays.
A new closed-form expression for quantizing the phases based on progressive phases to scan
without HSLL degradation is proposed. This approach generates a set of modi ed phases
suitable for small and moderately sized arrays, and can be applied to a dual and triple beam
scenarios. It is shown that the proposed approach can be easily extended to planar arrays. The
results of electromagnetic simulations of an array of coaxial-fed microstrip patch antennas using
CST microwave studio is used to verify the proposed approach for linear arrays. The HSLL
improvements with quantized DPSS amplitudes are also studied. Since it is observed that in the
case of null steering, the phase-only compensation is insu cient, simultaneous compensation
of amplitude and phase is proposed to improve the null depth. The trade-o between null
depth and width with the proposed algorithm is also discussed. It is expected that these new
approaches will improve beamforming and steering strategies in 5G mm-wave and other modern
applications of phased arrays.