On the Impact of Spillover Losses in 28 GHz

Rotman Lens Arrays for 5G Applications

M. Ali Babar Abbasi, Harsh Tataria, Vincent F. Fusco, and Michail Matthaiou
Institute of Electronics, Communications and Information Technology (ECIT), Queen’s University Belfast, Belfast, U.K.
e–mail:{m.abbasi, h.tataria, v.fusco, m.matthaiou}@qub.ac.uk

Abstract—This work demonstrates the sensitivity of lens an-

tenna arrays operating at millimeter-wave (mmWave) frequen-
cies. Considering a Rotman lens array in receive mode, our
investigation focuses on its most imperative defect: aberration
of electromagnetic (EM) energy. Aberration leads to spillover of
electric fields to neighboring ports, reducing the lens’ ability to
focus the EM energy to a desired port. With full EM simulations,
we design a 28 GHz, 13 beam and 13 array port Rotman lens
array to characterize its performance with the aforementioned
impairment. Our findings show that the impact of aberration is
more pronounced when the beam angles are close to the array
end-fire. More critically, the corresponding impact of aberration
on the desired signal and interference powers is also investigated
for an uplink multiuser cellular system operating at 28 GHz. The
presented results can be used as a reference to re-calibrate our
expectations for Rotman lens arrays at mmWave frequencies. Fig. 1: Rotman lens schematic diagram.
I. I NTRODUCTION the strongest scatterer(s) in the far-field propagation channel.
Radio-frequency (RF) lens-enabled antenna arrays have Most commonly, this is done via a network of analog phase
considerable potential to reduce the hardware complexity at shifters interfaced with the array elements (see e.g., [6]).
millimeter-wave (mmWave) frequencies [1]. Nevertheless, However, it is well known that mmWave phase shifters tend
lens arrays suffer from the inherent quantization of the to be significantly lossy, and often incapable of providing
beamspace. In addition, errors due to imperfections in the lens precise phase shifts over the required system bandwidth. More
construction itself have a critical impact on its performance. accurate phase shifters based on tunable materials often need
Almost all of the existing studies on lens-based architectures to be driven by an external control signal, leading to higher
as a radio transmitter or receiver consider the lens as an ideal circuit complexity and material cost [7]. Rotman lenses pro-
(perfect) RF device (see e.g., [1–4] and references therein). A vide an attractive alternative to phase shifter networks, which
common topology in the literature is known as the Rotman are increasingly considered [1–4]. A Rotman lens provides
lens, which is designed with three focal points where lens switch-less multi-beam operation of an antenna array with a
operation is perfect. At all other focal points, lens aberration wide beam-steering range. Figure 1 presents the generalized
occurs [5], causing spillover losses - as electromagnetic (EM) schematic diagram of a Rotman lens with 𝑁bp beam-ports and
energy which may be desired for a particular beam port 𝑁ap array ports, all connected to the corresponding tapering
also leaks into neighboring beam ports. This counteracts the and transmission lines. The area annotated as wave medium
functionality of the lens, causing high levels of desired signal propagation is where wave superposition takes place. The
loss. The impact of this physical artifact on the overall system tapering and transmissions lines are normally used to artifi-
performance remains uncharacterized in the literature, and in cially introduce time delays such that the waves propagating
this paper, we close this gap. More specifically, with a 13 × 13 through the transmission lines are phase aligned along the port
Rotman lens antenna array at 28 GHz, we investigate the terminations. A separate set of ports generally referred to as
impact of aberration on an uplink multiuser cellular system the dummy ports, are introduced to increase the adjacent beam
operating with analog and baseband processing. In doing so, port isolation. The dummy ports are normally terminated in
via EM simulations, we characterize the inherent limitations a matched load, and thus result in minimizing the reflections
of Rotman lens arrays by studying the spillover levels as a from the side walls of the lens. The beam side and array side
function of the azimuth direction-of-arrivals (DoAs). curvature is defined using design parameters: on-axis focal
II. D ESIGN P RINCIPLES OF ROTMAN L ENS A RRAYS length 𝑓1 , off-axis focal length 𝑓2 , their ratio 𝛽 = 𝑓2 /𝑓1 ,
Successful operation of mmWave cellular systems will the focal angle 𝛼, the sweep angle 𝜑max , and the lens ex-
requires pencil-like beam-steering capability in order to ac- pansion factor defined as 𝛾 = sin (𝜑max ) / sin (𝛼). A detailed
curately steer/receive the transmit/receive gain towards/from description of the above parameters can be found in [8]. While
realizing a specific Rotman lens geometry, the aforementioned
This work was supported by the EPSRC, UK, under grant EP/P000673/1
and EP/EN02039/1. parameters define a convex polygon, where the beam and the

978-1-5386-1197-5/18/$31.00 ©2018 IEEE

 (a) (b)

Fig. 2: EM simulation of a 28 GHz Rotman lens using FDTD

method. The labels 1–13 denote the beam ports, 14–26 denote
the array port while 27 and 28 denote the dummy ports.

array port focal points can be identified [8]. The part of the lens
around these points is usually referred to as the port segment,
and is connected to tapered lines which guide the propagating
wave towards corresponding transmission line.
III. 28 GH Z E LECTROMAGNETIC D ESIGN (c)
OF A ROTMAN L ENS A RRAY Fig. 3: Surface E-field distribution 200𝜇m inside the substrate
layer at multiple angle of arrivals at (a) 𝜃 = 0∘ , (b) 𝜃 = 12.5∘
Here we present the design of a 28 GHz Rotman lens
and (c) 𝜃 = 26.5∘ (color map: normalized for all figures).
constructed on a 0.64 mm thick Taconic-RF 60 substrate
(𝜖r = 6.15, tan (𝛿) = 0.0038) using microstrip technology.
In our design, 𝑁ap = 𝑁bp = 13, and two dummy ports are in addition to spillover towards the opposite ports (9 - 12) in
considered. The lens’s parallel plate region was synthesized case of DoA 3; a trend which is not significant at broadside-
by the design parameters of a tri-focal Rotman lens model like angles. Note that the DoA 2 and DoA 3 cases do not
coincide with any of the designed focal points of the Rotman
[3, 8]. The predefined design parameters were 𝑓1 = 5𝜆,
𝛽 = 0.9, 𝛼 = 30∘ , 𝜑max = 30∘ , and the array steering angle lens array, and hence are purely demonstrating the spillover
𝜃 was set to 50∘ . The tapering lines for all the ports are effects on the Rotman lens focusing. This result highlights one
of the major limitations of the Rotman lens, and demonstrates
3𝜆 in length. Moreover, the Finite-Difference-Time-Domain
(FDTD) method was used to characterize the lens in full-wave the fact that the EM focusing is more accurate towards the
electromagnetic (EM) simulation. The physical lens geometry broadside excitation angles. To quantify the field leakage into
is presented in Fig. 2 with port numbering defined in the the neighboring ports of the lens, we show an observation
caption. Although this specific example was constructed using curve 200𝜇m inside the Taconic substrate layer (see Fig. 4 (a)).
a standard synthesis method, it is worth noting that with a The one-dimensional plot of electric field as a function of the
careful selection of the design parameter of the parallel plate Rotman curve length for all DoA cases is presented in Fig. 4
region, the lens performance can be improved further [3]. (b). The field maxima along the curve length are indicative
of 13 ports. It is interesting to note that the field distribution
IV. R ESULTS AND D ISCUSSION for DoA1 is fairly symmetric. This is unlike DoA2, which
A set of EM simulations were carried out in which array reveals a contour with an uneven distribution of fields, one
ports of the lens were excited by phase ramped power signals where the spillover profile is vastly different for the central
representing multiple DoAs along the azimuth plane. For the and the edge ports. The final case in Fig. 4 (b) the contour for
sake of brevity, only three distinct scenarios are presented DoA3, where the wave converging point falls at the boundary
in Fig. 3. In the first case (denoted by DoA1), the beam is of two ports. Catastrophically, the power is distributed between
propagating from a potential transmitting source which could two concurrent ports almost equally. Here the effects of RF
be located at 𝜃 = 0∘ relative to the antenna array broadside reflections can also be observed in addition to the spillover,
(see Fig. 1). Based on the electric field distributions shown in where peaks in the contour are observed between 20-40 mm.
Fig. 3 (a), it is evident that the maximum power is converged In order for one to understand the ultimate impact of aber-
at the central beam port, i.e., port 7, while a small portion ration, we simulate an uplink multiuser MIMO system, where
of the power is spilled over to the neighboring ports due to two mobile terminals simultaneously transmit uplink data to
aberration. One can also notice from Figs. 3 (b) and (c), that a Rotman lens-enabled array. Upon receiving the signals, the
the amount of spillover increases as 𝜃 is varied from 12.5∘ Rotman array is fed with a network of RF switches, followed
(DoA 2) to 26.5∘ (DoA 3). One can also observe the reflection by two complete down-conversion chains to recover the trans-
 1

Cumulative Distribution Function (CDF)

0.5

Without Spillover
With Spillover
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Multiuser Interference Power [Absolute]

(a)
104 0.5
6
Port Indices 7 DoA1 With Spillover
3 Without Spillover
6 8 0
1 2 3 4 5 9 10 11 12 13 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Absolute Electric Field [v/m]

0 Normalized Desired Signal Power [Absolute]

-40 -30 -20 -10 0 10 20 30 40
104
6
DoA2
Fig. 5: Desired signal and interference powers at a user
3 terminal with 13×13 Rotman lens array receiving signals from
two simultaneous users at an average operating SNR of 0 dB.
0
-40 -30 -20 -10 0 10 20 30 40
104 interference power is not concentrated on a particular port,
6
DoA3
but rather distributed to a set of ports. It is worth noting that
3
overall, it is the ratio of the desired signal to the interference
0
-40 -30 -20 -10 0 10 20 30 40
power which is seen at a user terminal. The dominance of one
Relative Position on Focal Curve [mm] over the other is a function of all involved system parameters,
(b) as well as aberration levels. Hence, in order not to obfuscate
the findings, we avoid discussing the ratio explicitly.
Fig. 4: (a) Indication of the observation curve in the Rotman
lens substrate. (b) Electric field distribution along the observa- V. C ONCLUSIONS
tion curve depicting the field spillover at multiple DoAs when We present an investigation into the spillover losses of a 28
𝜃 = 0∘ , 𝜃 = 12.5∘ , and 𝜃 = 26.5∘ . GHz Rotman lens array. Via the aid of full EM simulations,
for a 13×13 Rotman array, we conclude that the impact of
mitted data at baseband. We employ the use of maximum-ratio spillover is significantly more pronounced at beam angles
(MR) baseband combining to separate the multiple arriving closer to the array end-fire. More accuracy in the spatial
streams. The 28 GHz far-field propagation channel is sim- focusing is found when the arriving beam at the Rotman array
ulated following the classical double-directional description is around the broadside direction, corresponding to exciting the
[9]. Precisely, we consider the total number of scattering central ports. The impact of aberration on the desired signal
clusters in the propagation channel to be 4, with each cluster and interference characteristics were investigated. To the best
contributing the total of 5 sub-paths. The instantaneous path of the authors’ knowledge, such a type of investigation is
gains are assumed to be Gaussian distributed with zero-mean missing from the lens literature.
and unit variance. The far-field array steering vector of a linear R EFERENCES
array is utilized with a uniform beam scanning range in the [1] A. M. Sayeed and J. Brady, “Beamspace MIMO channel modeling
forward half space of the array (𝜃 ∈ [−90∘ , 90∘ ]). Large-scale and measurement: Methodology and results at 28GHz,” in Proc. IEEE
GLOBECOM (Workshops), Dec. 2016.
fading (geometric attenuation and shadow fading/blockage) is [2] H. Tataria, et al., “Uplink interference analysis with RF switching for
modeled on both links using the classical power loss model lens-based millimeter-wave systems," in Proc. IEEE ICC, May 2018.
described in [2]. Due to space constraints, we avoid presenting [3] Y. Gao, et al., “Rotman lens based hybrid analog-digital beamforming in
massive MIMO systems: Array architectures, beam selection algorithms
further information regarding the simulation setup, as well and experiments”, IEEE Trans. Veh. Technol., vol. 66, no. 10, pp. 9134-
as the exact mathematical descriptions of the desired signal 9148, Oct. 2017.
and interference powers with MR processing. These can be [4] Y. Zeng and R. Zhang, “Millimeter wave MIMO with lens antenna array:
A new path division multiplexing paradigm," IEEE Trans. Commun., vol.
found in [2]. Figure 5 depicts the desired signal and multiuser 64, no. 4, pp. 1557-1571, Apr. 2016.
interference power cumulative distribution functions (CDFs) [5] M. Jiang, et al., “Metamaterial-based thin planar lens antenna for spatial
with and without aberration, at an operating SNR of 0 dB. One beamforming and multibeam massive MIMO,” IEEE Trans. Antennas
Propag., vol. 65, no. 2, pp. 464-472, Feb. 2017.
can observe that with aberration, there is naturally a loss in the [6] W. Roh, et al., “Millimeter-wave beamforming as an enabling technology
desired signal power, more pronounced at probability values for 5G cellular communications: Theoretical feasibility and prototype
around 0.5. This is due to the leakage of the desired EM energy results," IEEE Commun. Mag., vol. 52, no. 2, pp. 106-113, Feb. 2014.
[7] R. Mendez-Rial, et al., “Hybrid MIMO architectures for millimeter wave
to the neighboring ports, allowing the power to spread out. In communications: Phase shifters or switches?," IEEE Access, vol. 4, pp.
stark contrast, a reduction in the interference power is observed 247-267, Jan. 2016.
with aberration. This is due to the fact that the spreading of [8] R. C. Hansen, “Design trades for Rotman lenses," IEEE Trans. Antennas
Propag., vol. 39, no. 4, pp. 464-472, 1991.
the EM energy across multiple ports effectively helps to de- [9] M. Steinbauer, et al., "The double-directional radio channel," in IEEE
correlated the uplink signals in the RF domain, such that the Antennas and Propag. Mag., vol. 43, no. 4, pp. 51-63, Aug 2001.