This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

Received 5 March 2021; revised 1 April 2021; accepted 5 April 2021.

Digital Object Identifier 10.1109/JMW.2021.3072873

Circuit Type Multiple Beamforming Networks

for Antenna Arrays in 5G and 6G Terrestrial
and Non-Terrestrial Networks
Y. JAY GUO 1 (Fellow, IEEE), MARAL ANSARI 1 (Student Member, IEEE),
AND NELSON J. G. FONSECA 2 (Senior Member, IEEE)
(Invited Paper)
1
Global Big Data Technologies Centre, University of Technology Sydney, Ultimo, NSW 2007, Australia
2
Antenna and Sub-Millimetre Waves Section, European Space Agency, 2200 Noordwijk, The Netherlands
CORRESPONDING AUTHOR: Y. JAY GUO (e-mail: Jay.Guo@uts.edu.au).

ABSTRACT To support the ever-increasing demand on connectivity and datarates, multiple beam antennas
are identified as a critical technology for the fifth generation (5G), the sixth generation (6G) and more gen-
erally beyond 5G (B5G) wireless communication links in both terrestrial networks (TNs) and non-terrestrial
networks (NTNs). To reduce the cost and power consumption, there is a marked industrial interest in adopting
analogue multiple beam antenna array technology. A key sub-system in many of such antenna arrays is the
circuit type multiple beamforming network (BFN). This has led to a significantly renewed interest in and
new technological developments of Butler matrices, Blass matrices, and Nolen matrices as well as hybrid
structures, mostly at millimeter-wave frequencies. To the best of the authors’ knowledge, no comprehensive
analysis and comparison of circuit type multiple BFNs have been properly reported with focus on 5 G and
6 G applications to date. In this paper, the principle of operation, design, and implementation of different
circuit type multiple BFNs are discussed and compared. The suitability of these sub-systems for 5 G and
B5G antenna arrays is reviewed. Major technology and research challenges are highlighted. It is expected
that this review paper will facilitate further innovation and developments in this important field.

INDEX TERMS Fifth generation (5G), sixth generation (6G), beyond 5G (B5G), multiple beam antenna
arrays, circuit type beamforming networks (BFNs), Blass matrix, Butler matrix, Nolen matrix, terrestrial
network (TN), non-terrestrial network (NTN).

I. INTRODUCTION spaceborne networks in a multi-layer communication archi-

The global deployment of the fifth generation (5G) of wireless tecture, with studies recently initiated on the topic by the 3rd
and mobile communication systems is accelerating, and the Generation Partnership Project (3GPP) to elaborate require-
technical race on the sixth generation (6G) has started in ment use scenarios [8], [9].
earnest in many parts of the world [1], [2]. 5G promises signif- At the core of 5G and beyond 5G (B5G) systems are multi-
icantly increased capacity, massive connectivity, low latency, beam antenna arrays which enable unprecedented connec-
and game-changing new applications [3]. 6G is expected to tivity and high spectrum efficiency [10]–[12]. Digital beam-
provide much greater coverage, significantly reduced cost and forming, including Multiple-Input-Multiple-Output (MIMO)
energy consumption, and higher intelligence empowered by signal processing, is the most flexible approach for generat-
machine learning technologies [4]. The ambition to provide ing individually steerable multiple beams. Undoubtedly, fully
global and seamless connectivity calls for the integration of digital beamforming with massive antenna arrays serves as a
non-terrestrial networks (NTNs), capable of extending and powerful technology to meet some of the most challenging
complementing terrestrial networks (TNs) in areas under- desired features of future wireless communication networks,
served or out of reach [5]–[7]. This includes airborne and particularly in rich scattering environments [1]. However, this

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 1, NO. 3, JULY 2021 1
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

single point. This frequently arises in the context of open-air

events such as music concerts and sports competitions. Dif-
ferent sectors of the crowd are covered by separate multiple
narrow beams. Because these events are often one-time or
annual events, there is growing interest in high capacity cell-
on-wheels (COW) systems with a multiple BFN solution pro-
viding horizontally sectorized multiple cell coverage. These
can be moved in to cover the particular event. Ad-hoc net-
works are also primordial in emergency scenarios, where the
TN is disrupted by unforeseen circumstances such as natural
disasters and power outage. The deployment of a temporary
aerial base station (BS) can provide the connectivity required
to facilitate first aid intervention.
For 5G and B5G, there are new challenges for multiple
BFNs. They need to be flexible to support unequal input and
output ports as well as arbitrary beam directions, accommo-
date two dimensional (2D) beamforming, be compact, highly
integrated and multiband or wideband. It is also expected
FIGURE 1. General Layout of a multibeam cellular antenna using 1D that most of the multiple BFNs will be at mm-wave frequen-
circuit type BFNs.
cies [18]. This calls for innovative designs of Butler matrices,
Nolen matrices and various hybrid configurations. To this
approach generally leads to high power consumption and end, this paper provides a comprehensive review of various
hardware cost, which prevent its use in applications such circuit type BFNs. Different solutions and their basic operat-
as large-scale low-cost networking as well as small airborne ing principle, application, and performance are classified and
and space-borne platforms with limited power supply. Quasi- compared. Latest developments are reported, with particular
optical techniques also serve as an alternative analogue beam- focus on printed circuit board (PCB) technology, and current
forming technology. However, they are more suited for higher research challenges are discussed. Methods to enhance the
mm-wave and terahertz (THz) frequencies, as the line of sight performance of antenna arrays based on circuit type BFNs,
(LoS) wave propagation becomes dominant. A cost and en- including the antenna gain, the efficiency, the scanning range,
ergy efficient solution well fitting the microwave domain is the side lobe levels, etc., are presented, and several examples,
analogue multiple beam antenna arrays based on circuit type mainly for applications in cellular communication systems,
beamforming networks (BFN). Such BFNs typically consist are given. To the best of our knowledge, this is the first
of fundamental microwave/millimeter-wave (mm-wave) cir- comprehensive review of circuit type BFNs covering various
cuit components such as power dividers, couplers, crossovers, approaches to support current and future wireless communi-
phase shifters and switches. The best known circuit type BFN cation systems.
is the Butler matrix [13], [14], while the Nolen matrix is
emerging as an attractive alternative [15], [16]. When inte-
grated properly with antenna arrays, circuit type BFNs can
deliver superior performance at low-cost and with moderate II. CIRCUIT TYPE BEAMFORMING NETWORKS
power consumption. A. GENERAL DEFINITIONS
A conceptual layout of a multibeam cellular antenna using A circuit type BFN is a physical layer element of an array
one dimensional (1D) circuit type BFNs is shown in Fig. 1 system that distributes signals with the amplitudes and phases
where the azimuth sector is divided into several cells. Hori- required to produce a desired angular distribution of emit-
zontal linear arrays are fed with horizontal networks which ted radiation, when operating in transmission. A 1D BFN is
produce a set of beams in the azimuth plane. Separate inputs to generally connected to a linear array, defined as a set of N
the horizontal network excite the separate beams. The inputs elementary antennas, arranged along a given axis. The output
to the horizontal networks that apply to a particular azimuth ports of the BFN are often referred to as array ports. The
beam are excited in the normal way by a vertical network com- BFN also has M input ports, or beam ports, as each port
prising a power divider with or without electrical down-tilt provides the excitations to form a given beam. Both M and
which forms the elevation pattern for the particular azimuth N are integers and the BFN is often described as a M × N
beam. The stack of horizontal arrays operates as a planar array device. A multibeam antenna is realized if all of the input ports
with beamforming in both azimuth and elevation. Separate are driven simultaneously. One beam can also be synthesized
vertical networks are required for each azimuth beam and each exciting multiple inputs with the same signal. In reception,
polarization [17]. the BFN combines impinging signals in a reciprocal way, as
Another existing application of multiple BFNs is where an it is generally a passive device. Using well-established PCB
area of extremely high traffic density must be served from a technologies, BFNs can be fully integrated with an array of

2 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

antennas into a single substrate, providing cost-effective solu-

tions.
A circuit type BFN is fully characterized by its reduced
scattering matrix, S, with dimensions N × M and containing
the array coefficients in transmit, such that b = S a, where a
and b are the network input and output complex amplitude
vectors of dimension M and N, respectively. In other words,
the component ai , i = 1. . .M, of the vector a corresponds to
the complex electric field amplitude applied at the ith input
port, while the component b j , j = 1. . .N, of the vector b
corresponds to the complex electric field amplitude delivered
to the j th output port and corresponding array element. The
remaining terms of the complete scattering matrix are gen-
erally neglected under assumption that all ports are matched
and mutually decoupled between input ports and between
output ports. When the matrix S is orthogonal, i.e. S T · S ∗ = I,
where S T and S ∗ are the transpose and complex conjugate of
S, respectively, and I is the unitary matrix, the BFN is said
to be theoretically lossless. This means that the BFN itself
does not add losses besides ohmic losses introduced by the
transmission line technology used to implement the circuit. In
some cases, the BFN may be theoretically lossless in transmit
and lossy in receive, or vice versa [19]. An orthogonal BFN
produces orthogonal array coefficients, which produce in turn
multiple orthogonal beams with constraints on pointing direc-
tions, side lobe levels and overlap between adjacent beams.
These constraints on radiation patterns produced by orthogo-
nal BFNs are well detailed in [20], [21]. In this section, the
main circuit type BFNs are detailed. FIGURE 2. (a) Functional schematic representation of the M-entry Blass
matrix and (b) details of a node [24].

B. THE BLASS MATRIX

Introduced in the early 1960’s, the Blass matrix was the first
described method to provide multiple beams for antenna sys- be approximated by:
tems [22]. It consists of an array of N radiating elements
excited in series by M feeder lines. These lines are intercon- 
m−1
 − jφ 
 n−1
Tmn ≈ j sin θmn e− jφmn e pn cos θ pn cos θmq , (1)
nected by directional couplers at their crossover points and
p=1 q=1
they are terminated with matched loads at each line end, pro-
viding a travelling wave operation that facilitates the design. with the convention that the products are equal to 1 when
A Blass matrix can be sized for use with any number of the lower index is higher than the upper index, i.e. when
array elements and any number of beams. Unfortunately, due m and/or n are equal to 1. The formula is exact for the
to the resistive terminations, the efficiency of the Blass matrix first line, m = 1, but neglects multipath in the network for
is generally low. The less orthogonal the reduced S matrix, m > 1. The approximation is acceptable for nodes with very
the higher the losses. The size of the linear array also dictates low coupling level, typically required in large arrays. This
the level of losses at the terminations. The larger the array, the provides an understanding of the illustration in the seminal
lower the losses. The schematic of an M-input Blass matrix, paper by Blass [22] where the pointing of the beam is driven
as well as the associated notations and the details of a node, by the inclination of the feeder line, in first approximation.
are shown in Fig. 2. A given node Nmn is constituted of a The length of the transmission lines in the network may be
lossless directional coupler, having a coupling parameter θmn adjusted to provide true time delay operation, thus leading to
such that the coupled amplitude feeding the elementary an- frequency-independent beam directions [23].
tenna line n is equal to sin θmn while the direct port amplitude The main advantages of the Blass matrix are the ability to
which propagates to the next node on the feeder line m is generate multiple beams pointing in arbitrary directions, with
cos θmn . This assumes the coupler is perfectly matched and the flexibility on radiating pattern shapes and crossover levels.
two inputs, and outputs respectively, are perfectly decoupled. However, it exhibits higher losses due to the use of termination
The node also contains a phase shifter on the coupled port loads. These properties are mostly the consequence of the
characterised by the angle φmn . The transmission coefficient Blass matrix not being an orthogonal BFN. Ideally, one would
between the input port m and the elementary antenna n may like to design a Blass matrix, defining the parameters of the

VOLUME 1, NO. 3, JULY 2021 3

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

nodes for a given set of desired excitation vectors, such to

minimize losses while accounting for multipath within the
network. The general problem has no exact solution, except
in the case of a two-beam Blass matrix design. The method
was published by William R. Jones, with applications to
monopulse radars [25]. More recently, Mosca et al. introduced
a method in [24] that is applicable to any size of Blass ma-
FIGURE 3. Schematic representation of the conventional symmetric 4 × 4
trix. It uses a set of independent (i.e. orthogonal) beams, or Butler matrix [13].
excitation vectors, that form a base describing all the desired
non-independent beams. The power dissipation is minimized
for the independent beams and not the desired beams, thus
the solution is a good approximation but remains sub-optimal. network is theoretically lossless [14]. There is no beam spac-
The desired beams are generated as a linear combination of ing loss due to the nature of orthogonal beams [20], [36].
input beams, applying a set of complex input values as derived This phase defined array configuration leads to frequency-
in [24]. dependent beam directions. A more general M × N Butler
Optical Blass matrices with potential for multibeam op- matrix is considered in some cases, simply over-sizing the
eration in next-generation cellular wireless systems have at- matrix and terminating unused ports with matched loads. The
tracted much attention. Blass matrices based on optical phase layout may also be adjusted to more specific needs, in what
shifters can ensure a fully flexible selection of the pointing is also referred to as multimode BFNs [37]. This includes
angle of each wireless beam [26], [27]. In such designs, the the use of alternating division and combination functions, as
Mach-Zehnder Interferometers serve as coupler inside the also described in the seminal paper by Butler and Lowe [13],
Blass matrix. to produce cosine and square-cosine amplitude distributions.
Variations of these BFNs, including so-called chess networks,
C. THE BUTLER MATRIX have been reported in the literature [38]–[43]. In most of these
The Butler matrix was first described by Jesse Butler and variants, the orthogonality is compromised, either in transmit
Ralph Lowe in [13], shortly after the introduction of the Blass or in receive, or in both operation modes simultaneously, re-
matrix. The basic concept was also discussed at about the sulting in losses [19].
same time by J. Paul Shelton and Kenneth S. Kelleher in [28]. Typically, 4 × 4 Butler matrices are the most common
The fundamental difference between the Blass matrix and the form, because of their relatively simple structure [44]–[47].
Butler matrix is that the latter is essentially a parallel-fed Fig. 3 shows the schematic configuration of a conventional
BFN, while the former is a series-fed BFN. Allen [29] and symmetric 4 × 4 Butler matrix. For practical communication
Delaney [30] have published Butler matrix designs includ- systems, large arrays are required to increase the gain, reduce
ing precise values and location of all components. However, sidelobe levels (SLL), and have some control of the beam
Moody [14] was the first to report a general and systematic crossover level. While Butler matrices have significantly less
design for Butler matrices in 1964. The development of the components than Blass matrices for a same array size, their
Butler matrix came in parallel with that of the Fast Fourier topology imposes transmission line crossovers, their number
Transfer (FFT) [31], and the similarities between the two were increasing drastically with the array size, thus compromising
quickly acknowledged [32], [33]. Although seldom used in some of the benefits of Butler matrices in planar form.
that configuration, standard Butler matrices were also dis-
cussed in connection with planar arrays [34], and the con-
nection with the multi-dimensional discrete Fourier transform D. THE NOLEN MATRIX
(MD-DFT) was detailed more recently in [35]. The most The Nolen matrix was first described in the Ph.D. dissertation
widely used implementation of Butler matrices with planar of John C. Nolen, published in 1965 [15]. A schematic repre-
arrays is based on a two-stack arrangement of linear BFNs, as sentation of its general layout is illustrated in Fig. 4. While the
described in the seminal paper by Butler and Lowe [13]. original dissertation is not easily accessible, a Technical Note
With the conventional form of Butler matrices, the number by William C. Cummings from the MIT Lincoln Laboratory,
of inputs, M, is equal to the number of outputs, N, and it must published in 1978 and available online, provides some insight
be an integer power of 2 (i.e. N = 2n , where n is an integer). into the original work by Nolen [48]. The Nolen matrix is a
This derives from the use of a 4-port coupler as elementary general form of orthogonal BFN, based on a series-fed topol-
component [28]. The N × N Butler matrix produces N inde- ogy similar to that of the Blass matrix. Cummings described
pendent beams with phase progressions given by [14] an iterative design procedure to define all the couplers and
phase shifters constituting a Nolen matrix. He also demon-
(2k − 1)π
φ = , k = 1. . .N, (2) strated that a Nolen matrix may be simplified and eventually
N reduced to a Butler matrix when the number of ports is an
while amplitude is uniform (i.e. isophoric array configura- integer power of 2. Thus, the Nolen matrix may be seen as
tion). Inputs and outputs are matched and isolated, and the an implementation of the general Discrete Fourier Transform

4 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

TABLE 1. Comparison of the Number of Components in Circuit Type BFNs


The conventional Butler matrix implies a number of beams that is a power of 2,
i.e. M = N = 2n .

distributions, although constrained by the necessary condition

of orthogonality. These works initiated by Fonseca et al. have
triggered a regain of interest for Nolen matrices and designs
are now being reported by other research groups [54]–[56].

III. COMPARISON OF CIRCUIT TYPE BEAMFORMING

NETWORKS
The main circuit type BFNs discussed in Section II are com-
pared here based on metrics of relevance for the targeted 5G
and B5G applications, linked to size, weight, power and cost
(SWaP-C) considerations of importance in modern product
developments.
FIGURE 4. (a) Functional schematic representation of the M-entry Nolen
matrix (b) Details of a node [16].
A. NUMBER OF COMPONENTS
The number of components necessary for designing different
circuit type BFNs or matrices is an important factor in system
designs. Besides the obvious impact on the design effort, af-
(DFT), in the same manner that the Butler matrix is an im- fecting the final cost of the product through non-recurring en-
plementation of the FFT [49]. Being an orthogonal BFN, a gineering, it also constrains the footprint on a PCB and even-
N × N Nolen matrix will produce the same phase progres- tually the size and weight of the final product. In general, there
sions as a Butler matrix, following eq. (2), with the important is an interest in integrating the design as much as possible.
difference that there are no constraints on the possible values With the same notations as in Section II, Table 1 compares the
of N. numbers of components implemented in different circuit type
While the work of Cummings focused on the link between BFNs with M input ports and N output ports. It is assumed
Nolen and Butler matrices, a more recent work revisited the here that all ports of a conventional Butler matrix are used,
design of Nolen matrices taking advantage of the similari- i.e. M = N.
ties with Blass matrices [16]. Indeed, the Nolen matrix may As anticipated in Section II, Butler matrices need less el-
be seen as a lossless variant of the Blass matrix in which ements than Blass or Nolen matrices of equivalent size. In
all the directional couplers below the diagonal have been reality, however, the distinction is not quite so simple as a
removed, and those on the diagonal are replaced by simple Butler matrix requires a large number of crossovers, neces-
bends. Consequently, all the matched loads at the end of the sitating a design with several layers. By contrast, the specific
feeder lines have also been removed. Using the method of structured form of Blass and Nolen matrices leads to a simpler
Mosca et al. for Blass matrices with orthogonal excitations and straightforward planar design, so the Butler matrix turns
and setting the upper limit, sin θ , on coupling values close out to be in some cases as complex as a Blass or Nolen matrix.
to one, i.e. sin θmn < sin θ , for m = 1. . .M and n = 1. . .N To appreciate the complexity of the Butler matrix, a 32 × 32
with sin θ ≈ 1, the resulting design will approximate a Nolen Butler matrix is shown in Fig. 5.
matrix. Note that it is not possible to set the upper limit equal In terms of design effort, there is also a difference between
to one as the algorithm requires to divide by cos θ . However, Butler matrices and Blass or Nolen matrices. Indeed, Butler
considering typical numerical and experimental precision, the matrices rely on hybrid couplers only, with all couplers being
approximation is generally acceptable. Thus, the Nolen matrix identical. This is not the case with Blass and Nolen matri-
may be considered as an asymptotic singular case of a Blass ces, as the series-fed structure requires a larger number of
matrix with orthogonal excitations. unbalanced coupler designs. The larger the array, the larger
Reported designs [16], [50]–[53] indicate that Nolen ma- the range of unbalanced couplers required. A careful selection
trices are comparable in complexity to their equivalent Butler of the coupler technology is essential to enable the necessary
matrices in planar realizations. An interesting feature of Nolen coupling range. Alternatively, a tandem hybrid coupler con-
matrices is their capability to provide non-uniform amplitude figuration with adequate phase adjustment in between the two

VOLUME 1, NO. 3, JULY 2021 5

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

be orthogonal. This condition is difficult to satisfy with mul-

tiple beams and this is the reason why most designs reported
still assume a uniform amplitude distribution.
Orthogonal excitations have an impact on the beams, con-
straining side lobe levels, pointing directions, overlap between
beams, etc. The link between lossless networks and radiation
performance has been addressed in detail by Stein [20] and
White [21]. In the case of linear arrays, a uniform excitation
provides a first side lobe at about 13 dB below peak directivity,
while beam overlap between adjacent beams produced by a
lossless BFN is about 4 dB below peak directivity. For some
applications, there may be an interest in reducing SLL to
lower interference and/or increase the overlap between adja-
cent beams to reduce gain roll-off over the service area.

D. FREQUENCY BEHAVIOUR
FIGURE 5. Schematic representation of 32 × 32 Butler matrix with hybrid A last factor for comparison is the frequency behaviour of
couplers [49]. these matrices. The topology presented for Blass or Nolen
matrices has a naturally narrower frequency band than a But-
ler matrix. Indeed, the phase adjustment for all directional
coupler outputs, before taking the phase shifters into account,
couplers may be used to produce unbalanced couplers [48],
can be applied rigorously only at the center frequency. The
but this comes at the expense of a more complex implementa-
significant difference in electrical path length in series-fed
tion.
BFNs implies that, when deviating from this design frequency,
the matrix performance is degraded. This phenomenon does
B. LOSSES not occur in Butler matrices since the line lengths are all
Another key element of comparison is the intrinsic losses of of the same order, by design. However, constant phase shift
the circuit type BFN, which eventually will affect the power over frequency leads to frequency-dependent beam directions.
efficiency of the final product. As detailed in Section II, Blass While the resulting beam-squint may be acceptable within a
matrices are lossy by design as they dissipate power in the limited fractional frequency bandwidth of typically 10 to 20%,
matched loads that terminate the feeder lines, while Butler and this may become critical for wide band or even multiple-band
Nolen matrices are theoretically lossless. For BFN designs designs. For such applications, Blass matrices may be prefer-
with relatively small number of radiating elements and/or a able as they can be designed with time delay rather than phase
relatively strong constraint on the directional couplers, losses delay [23].
in the matched loads may reach unacceptable levels. For this Another point for consideration when discussing frequency
reason, Blass matrices are generally preferred for larger ar- behaviour is that most unbalanced coupler designs tend to
rays, while Butler and Nolen matrices are more suited for have reduced bandwidth when compared to balanced couplers
smaller array designs, typically up to 16-element arrays in with similar technology. The larger the coupling unbalance,
practice, considering a value that is a power of 2. the narrower the frequency bandwidth. This may translate
While Butler matrices are expected to be the most power into a reduced frequency bandwidth with acceptable return
efficient design, this may be compromised in the case of a loss or a more frequency-dependent coupling value. A careful
multi-layer implementation, as layer-transitions will add to selection of the coupler solution is required to overcome this
the BFN losses and alignment errors may impact the overall limitation.
performance. In this respect, Nolen matrices may be seen as a The advantages and disadvantages of each circuit type BFN
better compromise between losses and complexity. are summarized in Table 2.

C. CONSTRAINTS ON THE EXCITATION IV. METHODS TO ENHANCE THE PERFORMANCE OF

One considerable advantage of Blass and Nolen matrices over CIRCUIT TYPE BEAMFORMING NETWORKS
the Butler matrix is the possibility of realizing flexible ampli- Different multibeam array applications have distinct require-
tude excitations, while Butler matrices in their standard form ments. The efficiency of the design, operating frequencies and
impose a uniform amplitude excitation. In the case of Blass related fractional bandwidth, size and in some cases weight,
matrices, this flexibility comes at the expense of increased SLL, crossover level and beamwidth, number of the beams,
losses. The less orthogonal the excitations, the higher the the ability to realize different polarizations, and wide scanning
losses. In the case of Nolen matrices, the excitations do not range are some of the essential requirements which must be
necessarily need to be uniform in amplitude, but they have to taken into consideration in the design process.

6 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

TABLE 2. Comparison of Different Circuit Type BFNs

For instance, in cellular base station applications, achieving

multibeam sectorization over a ±60◦ azimuth coverage with
−10 dB beam crossover levels are two crucial requirements.
High gain and efficient antennas are always needed in the
antenna systems to compensate for the dielectric and metallic
losses. Such antennas are also desirable for compensation of
high path losses, especially at mm-wave frequencies where
the atmospheric attenuation rate is high due to the resonance
of oxygen molecules. In the following section, some of the
reported methods which can improve traditional designs to
achieve specific requirements are reviewed.

A. LOWERING SIDE LOBE LEVEL FIGURE 6. Schematic representation of modified Butler matrix utilized
power dividers for tapered amplitude law [58].
If the desired signal enters the main beam while interfering
signals enter the sidelobes, then lowering the sidelobes rela-
tive to the main beam improves the signal to interference ratio [59], [64]–[66]. A schematic representation of the resulting
in the receiver. This is desirable in all communication systems. BFN structure is shown in Fig. 6. Li et al. proposed a novel
The sidelobes are also unwanted for transmitters because they N × N Butler matrix using an additional circuit realized with
produce radiation in directions where there are no users. This equal ratio power dividers between the Butler matrix and radi-
could cause interference to other users and lead to unnecessary ating elements to feed a 2N array [64]. Such a circuit is simple
waste of energy. The optimum SLL of an N-element uniform but lossy and difficult to be integrated in a compact system.
array is about −13.6 dB [11]. However, many systems cannot Fonseca and Ferrando demonstrated that the modified Butler
achieve a SLL of less than −10 dB due to effects such as matrix with low SLL may be replaced by a N × 2N Nolen
the mismatch between feeding network and antenna array and matrix having similar properties [58]. The use of a Nolen
mutual coupling between elements [62]. matrix avoids the issues related to the additional crossovers
Several techniques have been reported in the literature to introduced by the power dividers.
reduce the SLL in beamforming antenna arrays. Reduced Tapered aperture illumination utilizing lossy networks has
SLL for Butler matrix fed linear arrays was first introduced been reported in the literature [61], [67]–[69]. Another solu-
by Shelton. He showed the SLL of small linear arrays fed tion is the use of dissipative attenuators to achieve the desired
by a Butler matrix may be improved increasing the number amplitude distribution [70] where a combination of aperture
of elements of the array and using the “covering condition” attenuation and feed design, including the use of overlapping
[63]. Using more elements is the first solution. However, this orthogonal feeds, is considered. External attenuators are asso-
will increase the cost and complexity of the network. It also ciated with significant insertion losses and lower efficiencies.
increases the size of the feed network. Arbitrary amplitude is generated using a 4 × 8 Butler matrix
Butler matrix designs utilizing power dividers to increase with Chebyshev distribution in [71]. The use of a four-port
the number of array elements and to taper the amplitude across network and variable phase shifters provides −20 dB SLL.
the antenna array have been reported in the literature [57], This value could reach −50 dB, only by altering the power

VOLUME 1, NO. 3, JULY 2021 7

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

TABLE 3. Reported Techniques to Reduce the SLL in Circuit Type BFNs

(SIW) [75] becomes a serious alternative, albeit with a larger

circuit footprint. Therefore, some effort has been made to
achieve a level of miniaturization, across the frequency range
of interest for wireless communication devices.

1) REDUCED NUMBER OF ELEMENTS

With the design of Butler matrices, crossovers have always
been a problem, making the structure bulky. Butler matrices
without crossovers were proposed in [76], [77] and without
a 45◦ phase shifter in [78], [79]. A low-cost mm-wave 4 × 4
Butler matrix using microstrip technology is presented in [77],
FIGURE 7. Circuit controlling the SLLs with Butler matrix. The values of the without any crossovers on a single layer. This topology is
amplitude and phase distribution at the output ports are indicated on the different from the traditional left-to-right arrangement of the
right. φc stands for the phase of the cross coupler [73]. basic components. It places the input ports on the outer side
of the layout and the output ports on the inner side. Thus,
ratios of power splitters. In [72], the possibility of SLL re- there are no overlaps of the four signal paths. This topology
duction is investigated using multibeam conformal antenna was used for a SIW 4 × 4 Butler matrix in [80], including the
arrays fed by a modified Butler matrix with compensating antenna array, and it was extended to a 4 × 8 Butler matrix
phase shifters and crossovers. in [81] by introducing four power splitters in order to excite
Tekkouk et al. [73] presented a mm-wave beamforming an eight-element array.
array based on a 4 × 4 Butler matrix using a dedicated circuit Ansari et al. [60] presented a microstrip-to-slot line transi-
composed of four 90◦ couplers with different coupling factors tion to avoid the use of excessive crossovers in Butler matrices
(P1-P4), eight cross couplers, and phase shifters to control the with a larger number of outputs. Their proposed structure is
SLL, as shown in Fig. 7. The beamforming antenna array can operating at 28 GHz with a size of 3.5λ × 1.5λ, which is
cover an angular area of ±43◦ with SLL less than −17.5 dB only 50% of the area that an equivalent conventional matrix
and beam overlap of −10 dB between adjacent beams over the would occupy. The design architecture of a symmetrical 3 × 3
entire scanning range. uniplanar Nolen matrix is introduced in [55]. Compared to
It is to be noted that there is always a compromise be- previously reported 3 × 3 Nolen matrices, the design reduces
tween the SLL and crossover level relating to beamwidth, the number of phase shifters and does not require outside
which should be considered for specific applications. This is phase compensation lines. The design architecture and the
a direct implication of the orthogonal excitations. A possible final prototype are shown in Fig. 8.
workaround is to use two interleaved arrays, thus achieving
much higher overlap without compromising efficiency but at 2) MULTI-LAYER CONFIGURATION
the cost of some hardware duplication [74]. To conclude the Another drawback of having an excessive number of compo-
discussion, some reported circuit type BFNs with reduced nents is the large footprint associated with the BFN layout.
SLL are compared in Table 3. This issue can be resolved by using a multi-layer configuration
which is facilitated by the development of SIW technologies.
B. MINIATURIZING THE CIRCUIT LAYOUT Since SIW is a closed structure, several layers can be directly
Bulky antennas cannot be easily integrated to practical wire- stacked on top of each other without influencing the trans-
less systems. Traditional designs of circuit type BFNs have mission performance. To demonstrate the reduction of the
bulky layouts when they employ more components for im- dimensions of a 4 × 4 Butler matrix, dual-layer configurations
plementation in large arrays, and the technology used is typi- were developed in [82], [83]. There are two main advantages
cally transmission line-based. This may be problematic, espe- of using this dual-layer configuration. First, eliminating the
cially at low frequencies. At mm-wave frequencies, insertion crossovers leads to significantly shorter path lengths, thus,
losses are prohibitive with traditional transmission line types, reducing the insertion losses. Second, it also reduces the in-
such as microstrip lines, and substrate integrated waveguide plane footprint of the Butler matrix by half. Another example

8 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

FIGURE 10. (a) Photograph of the coupler prototype. (b) Prototype of the
4 × 4 Butler matrix realized by lumped-element coupler connected to a
monopole array [86].

FIGURE 8. (a) Architecture of the symmetrical 3 × 3 Nolen matrix (b)

Photograph of the fabricated the 3 × 3 Nolen matrix [55].

FIGURE 11. 3 × 3 Nolen matrix with lumped-element couplers and phase

shifters (a) Simulated layout. (b) Fabricated photograph [54].

3) LUMPED ELEMENTS CIRCUITS

Miniaturised lumped-element devices, such as coupler and
crossover structures, are also popular among researchers as a
solution to reduce the size of the feed network in beamforming
FIGURE 9. Simulated model of the dual-layer 8 × 8 Butler matrix [84]. antenna arrays. For instance, Gandini et al. presented in [86]
a lumped-element unit cell for designing compact BFNs. This
structure is shown in Fig. 10.
of a dual-layer SIW 8 × 8 Butler matrix developed in [84] is Ren et al. employed lumped-element couplers to minimise
shown in Fig. 9. the size of the Nolen matrix, especially for low frequency
As illustrated above, a dual-layer configuration can help applications [54]. The simulated layout and fabricated proto-
remove some of the crossovers to decrease the losses and to type of the proposed 3 × 3 Nolen matrix feeding network are
improve the compactness of the BFN layout. Using coplanar shown in Fig. 11.
waveguide (CPW) multilayer technology is another solution.
In [85], a 4 × 4 Butler matrix with the slot-coupled directional 4) MINIATURIZED TRANSMISSION LINES
coupler using CPW multilayer technology is described to As mentioned above, SIW is a good candidate technology to
avoid any crossing lines in the matrix. Such design reduces reduce losses at mm-wave frequencies. However, this typi-
the size of the proposed circuit. cally leads to larger circuit boards, which may compromise

VOLUME 1, NO. 3, JULY 2021 9

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

the system integration. For this reason, several alternative in a beam squint with frequency that limits its application.
transmission lines, derived from SIW technology, such as For the directional couplers, which are the critical components
half-mode substrate integrated waveguide (HMSIW) [87], in such networks, the conventional branch line coupler has a
ridge substrate integrated waveguide (RSIW) [88], folded C- narrow bandwidth [97]. Afifi et al. reported a wideband Butler
type substrate integrated waveguide (FCSIW) [89], ridged matrix using ridge gap waveguide technology. The structure
half-mode substrate integrated waveguide (RHMSIW) [44], presents 21.25% bandwidth, which shows about three times
have been proposed to reduce the circuit dimensions while more bandwidth than the reported counterparts [45]. A novel
preserving planar transmission characteristics and compatibil- 4 × 4 Butler matrix with a wide fractional bandwidth of
ity with low-cost PCB processes. Among them, FCSIW has 56.4% and compact size is presented in [98]. The design used
been widely used because it has the lowest loss figure while a swap in a vertically installed planar structure to implement
providing size miniaturization. In [90], a mm-wave multibeam the quadrature coupler and crossover required by the Butler
array antenna using FCSIW technology is proposed. A minia- matrix topology.
turized 4 × 4 Butler matrix and a single-branch slot-array are In [99], a planar wideband circular polarized (CP) beam
described, where the designs exhibit 40% and 33.2% reduc- steering antenna array with a Butler matrix network is pro-
tion in occupied surface for the Butler matrix and the whole posed utilizing the rotation technique. This technique is
multibeam array antenna, respectively, in comparison with widely used for achieving CP radiation and enhancing the
their standard SIW counterparts, while demonstrating similar impedance and axial ratio bandwidths [99], [100]. In [101],
RF performance. Zhu et al. presented a wideband multibeam antenna array
Substrate integrated suspended line (SISL) is another based on a three-beam Butler matrix, while the method is ap-
promising alternative transmission line technology that has plicable for designing larger beamforming circuits. The BFN
been proposed and applied for both passive and active circuit comprising wideband quadrature and phase shifters is devel-
designs [91], [92], as it features good properties, including low oped using striplines, and the overall network has 46% frac-
loss, low cost, and high integration. An SISL Butler matrix tional bandwidth covering the frequency bands from 1.7 GHz
using patch element and honeycomb concept was proposed to 2.7 GHz. This structure is shown in Fig. 12.
in [92]. The feed network is composed of four couplers, A broadband Nolen matrix using SIW technology is studied
two crossovers implemented using patch elements for ease in [53]. Being serial feeding networks, Nolen matrices usually
of fabrication and low conductor loss, as well as four phase introduce significant phase dispersion in their standard form,
shifters. The honeycomb concept provides a multicavity struc- thus limiting their effective bandwidth. Generally speaking,
ture where every component of the Butler matrix is isolated the Nolen matrix when compared to its Butler matrix coun-
from each other, so they can be designed and adjusted sepa- terpart exhibits narrowband performances due to significant
rately. This makes the system integration more efficient and phase dispersion introduced by the unequal electrical paths
flexible. Although the developed SISL Butler matrix benefits connecting one input to every output. Therefore, using a
from self-packaging mechanism, it has a complex circuit be- more parallel topology for the Nolen matrix balancing electri-
cause it employs a large number of platted vias, and multilayer cal paths should produce wider frequency bandwidth perfor-
substrates are required to create housing for each component. mances. In [53], broadband frequency operation is achieved
In this regard, perfect magnetic conductor (PMC) packag- with adequate coupler delay compensation, resulting in a more
ing is an effective technique for device shielding. In [93], a parallel matrix topology. H-plane couplers with continuous
PMC packaging concept was utilized at mm-wave frequen- aperture are preferred in the proposed design as opposed to the
cies, which suppresses the higher order cavity modes, im- cross-couplers previously used in the serial configuration. The
proves insertion losses, and helps in developing packaged coupler, phase shifter, and final prototype of this design are
microstrip lines (PMSLs) and double-ridge gap waveguide shown in Fig. 13. The design achieves excellent results over a
(DRGW) transmission lines. 11.7% fractional frequency bandwidth centered at 77 GHz.
Blass matrices can be designed as a true time-delay net-
C. EXTENDING THE FREQUENCY BANDWIDTH work, thereby resolving the bandwidth limitations. In this
The active impedance of an array changes with scanning regard, Lialios et al. introduced a mm-wave wideband multi-
angle. The aforementioned circuit type BFNs use a set of layer Blass matrix for communications between small cells
power dividers, couplers, crossovers, and phase shifters to and base stations or between base stations and the gate-
form the feed network. As long as each component is wide- way [102]. The structure is shown in Fig. 14. The major
band with a stable output phase, the circuit type feed network disadvantage of Blass matrices is their low efficiency because
can achieve a wide operating bandwidth in terms of amplitude of the lossy nature of the matrix.
and phase characteristics. In this regard, several wideband
BFNs have been reported using wideband couplers [83], [94],
crossovers [79], Wilkinson dividers [95] and Schiffman phase D. ENABLING MULTI-BAND SYSTEMS
shifters [96]. As mobile communication standards evolve from genera-
There are numerous challenges with the design of wideband tion to generation, mobile operators have ambitious plans to
structures. For example, the use of fixed phase shifters results combine the multiple frequency bands into multibeam base

10 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

FIGURE 13. (a) Topology of the H-plane short-slot coupler. (b) Topology of
unequal with unequal length phase shifter. (c) Manufactured 4 × 4 Nolen
matrix at 77 GHz [53].
FIGURE 12. (a) 3D model of constructed stripline components used in the
design (b) Layout of the three-beam Butler matrix (c) Prototype of the
overall structure [101].

stations. To satisfy the demand in multi-band wireless com-

munication systems, several dual-band and multi-band BFNs
have been reported.
Dual-band multibeam antenna arrays are commonly built
by employing dual-band radiating elements [103], [104] or
interleaving radiating elements operating at upper and lower
frequency bands [105]. However, one of the major limitations
in the realization of dual-band multibeam antenna arrays is
the distance between radiating elements in order to keep grat-
ing lobes at a reasonable level. Wincza et al. presented an
approach in which the antenna array consists of N radiating
elements operating at the higher frequency range and N/2
dual-polarized antenna elements operating at the lower fre-
FIGURE 14. Layout of a circular mm-wave Blass matrix [102].
quency range integrated in a common aperture [106].
The most popular approach when making a dual-band cir-
cuit type BFN is to convert all the constituted components of
the single band structure into dual-band ones. In this regard, equivalent to the combination of a dual-band ±90◦ BLC and a
a novel dual-band branch line coupler (BLC) for designing dual-band ±45◦ /±135◦ phase shifter. This feature eliminates
a dual-band Butler matrix was introduced in [107], [108]. the requirement of additional phase shifter for the design of
The proposed BLC generates a phase difference of ±45◦ and Butler matrices. The BLC and basic connections for designing
±135◦ between the outputs ports, which enables it to perform dual-band Butler matrix are shown in Fig. 15.

VOLUME 1, NO. 3, JULY 2021 11

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

FIGURE 15. (a) Proposed dual-band branch line coupler. (b) Basic
connections for designing a dual-band butler matrix [107].

FIGURE 16. Photograph of the fabricated proposed dual-band 4 × 4 Butler

matrix [108]. FIGURE 17. (a) 2D antenna feeding network using six 3 × 3 Nolen
matrices with nine input ports 1–9. (b) Photograph of the fabricated 3 × 3
Nolen matrix and 2D beamforming phased array. (c) Simulation results of
Another new dual-band 4 × 4 Butler matrix with dual-band the proposed 2D beamforming phased array, generating nine unique
radiation beams with special values on elevation and azimuth (θ, φ) when
3 dB quadrature BLC and dual-band ±45◦ phase shifter is input ports 1–9 are excited [56].
reported in [108]. The design is capable of performing dual-
band operation with the 2.7 band ratio at 1.0 GHz and 2.5 GHz
center frequencies. A photograph of the fabricated 4 × 4 dual- Researchers have been attempting to combine 1D Butler
band Butler matrix realized in microstrip technology is shown matrices to create BFNs for 2D multibeam antenna arrays.
in Fig. 16. The classic topology for such a BFN consists of two sets of
sub-BFNs orthogonally connected to each other; they realize
E. ENABLING 2D SCANNING beam forming properties in both the horizontal and vertical
Most reported works on circuit type BFNs have been focused planes [109], [110]. A novel 3 × 3 Nolen matrix for 2D beam-
on 1D designs. In practice, many applications would need forming applications is presented in [56]. The design has been
2D antenna arrays with a wide scanning range. For example, realized in microstrip technology operating at 5.8 GHz. The
in cellular multibeam antenna arrays, in order to ease the proposed Nolen matrix employs couplers with arbitrary phase
deployment and to cope with the mast sways of lamp posts differences and generates relatively flexible phase differences
and other assumed 5G small cell deployment sites during at its output ports. The 2D feeding network structure and the
operation, scanning in 2D is required. However, such scanning corresponding simulated results of the planar phased array are
capability increases significantly the complexity of the BFN, shown in Fig. 17.
especially for the cases where a large number of beams are The key challenge in designing 2D multibeam antenna ar-
required. rays with 1D circuit type BFN is the large size. In this regard,

12 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

FIGURE 18. (a) Configuration of the 2D scanning multibeam array

antenna. (b) Details of the basic components. (c) Photographs of the
fabricated multibeam array antenna [111].

Yang et. al [113] proposed a symmetrical 2D BFN at 94 GHz.

The network is compactly implemented using two multi-
folded Butler matrices and four couplers in LTCC technology,
so as to realize eight symmetric beams in 2D with stable gains.
Compact 2D multibeam array antenna fed by planar cascaded FIGURE 19. (a) Topology of conventional 16 × 16 BFN. (b) Proposed
Butler matrices for mm-wave communication was reported topology of uniplanar 16 × 16 BFN. (c) Proposed topology of 16 × 16 BFN
with eight-port crossover [112].
in [114]. Lian et al. reported several 2D multibeam antenna
arrays in planar and uniplanar structures [111], [115]–[117].
In one of the recent works, a compact mm-wave 2D scanning
multibeam antenna array based on a SIW 16-way BFN is intersections simultaneously and can reduce the total number
proposed. The design covers a reduced area of 3λ × 12λ and of path intersections from 16 to only 4 as shown in Fig. 19.
is an attractive candidate for 5G communication systems. This The structure is uniplanar and does not suffer from multilayer
structure is shown in Fig. 18. design with typical drawbacks such as fabrication complexity,
Eight-port hybrid couplers and crossovers are proposed to high cost, high losses caused by the transmission between
realize uniplanar 2D multibeam BFNs. As an example, they separated layers.
are implemented in a Butler matrix to enable uniplanar 2D A 2D scanning magneto-electric dipole antenna array fed
BFN in [112]. The proposed crossover can address four path by a printed ridge gap waveguide (PRGW) Butler matrix is

VOLUME 1, NO. 3, JULY 2021 13

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

proposed in [118]. Four scanning beams are created with

beam angle direction θ0 and φ0 calculated as (35◦ , −135◦ ),
(35◦ , −45◦ ), (35◦ , 45◦ ), and (35◦ , 135◦ ) when fed from port
1 to 4, respectively. Wang et. al [114] utilized the concept of a
3D-to-planar cascaded Butler matrix topology transition to re-
alized a 2D multibeam antenna array at mm-wave frequencies.
The beam steering system is compact and creates 16 spatially
distributed orthogonal beams scanning a conical space with a
maximal cone angle of 77.4◦ in the elevation plane and 136.8◦
in the azimuth plane by feeding at different inputs.

F. TRANSMISSION LINES TECHNOLOGIES

Various realization techniques and associated transmission
line technologies have been reported for the fabrication of
circuit type multiple BFNs. These include microstrips [16],
[56], [60], [119], striplines [101], low-temperature co-
fired ceramic [120], ridged-waveguides [73], hollow wave-
guide [121], SIW [53], [115], [122], half-mode SIW [123],
ridged half-mode SIW (RHMSIW) [44], coplanar waveguie
(CPW) [85], to list a few examples.
Different technologies have their pros and cons. The de-
signs realized with microstrip lines suffer from high di-
electric losses, and there are radiation losses especially at
high frequencies. However, microstrip technology is cheap
and easier to be integrated with other components in RF
systems. Besides, with such technology, there is no need
for SIW/waveguide-to-microstrip transitions. SIW technology
FIGURE 20. (a) Exploded view of the proposed Butler matrix. (b) Test
has been developed as a compromise between the low-loss and setup of the proposed Butler matrix. The two halves of the matrix are
excellent high-frequency performance of rectangular waveg- connected via four phase matched SMA cables [44].
uides, and the integrability of planar transmission lines. How-
ever, SIW structures occupy space and have a large size, which
opens another door for the development of new techniques to recently by Fonseca et al. in [124], [128]. Fig. 21 shows the
miniaturize their structures. connecting network enabling that interesting property and a
In [44], Der et al. presented an integration of ridge picture of the device under test.
half-mode SIW (RHMSIW) 4 × 4 Butler matrix with tun-
able phase shifters fabricated using the same technology to V. FUTURE WORK AND CHALLENGES
achieve semi-continuous beam steering for 5G applications. Based on the application requirements, one or more of the
This work reported over 70% miniaturization, which shows methods listed in this paper may be employed to produce
RHMSIW technology is a promising solution for implement- circuit type BFNs suitable for 5G and B5G wireless com-
ing miniaturized beamforming devices. Fig. 20 shows this munication systems. Increasing the bandwidth, improving the
structure. scanning range, and reducing the fabrication cost are among
For applications in harsh environment, such as atmospheric some of the challenges to be addressed in future activities.
satellites and low Earth orbit (LEO) satellites, waveguide is To provide higher system capacity or diversity, current and
generally the preferred technology, using CNC milling manu- future wireless communication systems must support dual
facturing. At mm-wave frequencies, the size of the device re- polarization. Some existing designs can be found in [117],
mains acceptable and the reduced insertion losses are appeal- [129]–[134], and they typically require duplication of hard-
ing. To further reduce the dimensions of the BFNs in wave- ware to achieve dual polarization or a double-sized beamform-
guide technology, 2D couplers have been proposed in [125] to ing network [135]. Simplifying the design and reducing the
combine both stages of beamforming (vertical and horizontal cost of circuit type BFNs for dual-polarized systems is key
beamforming) in a same structure suitable for planar arrays. to the wireless industry. To this end, a new way of feeding
Various designs were reported in K-band, including 16 × 16 dual-polarized antenna arrays using a single layer network is
and 64 × 64 Butler matrices [125]–[127]. Most 2D scanning presented in [136]. However, it works only for a single beam.
planar arrays using BFNs produce a square lattice of beams. Significant research is required to advance the state-of-the-art
The beam overlap may be improved using a triangular lattice in this field.
of beams instead. The first demonstration of a 2D Butler Other challenges associated with circuit type BFN based
matrix generating a triangular lattice of beams was reported arrays are the limited isolation between beams, which are

14 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

of cells constituting the service area, providing dynamic re-

source allocation [139]–[142]. Future systems are expected to
benefit from a hybrid beamforming approach to reduce the
number of control points required in a phased array, while
providing some level of flexibility, such as a more continuous
beam steering or higher pointing direction granularity. This
is of interest when the number of beams is much lower than
the number of radiating elements and is a promising field of
research for cost-effective alternatives to conventional phased
array antennas.
Currently, THz spectrum is being considered an important
part of the 6G spectrum, especially for achieving extreme
datarates over short distances. For these systems, we envis-
age two antenna technologies to become dominant. One is to
employ lenses to achieve high gain without suffering from
the losses in transmission lines. The quasi-optical nature of
sub-THz and THz waves renders lenses effective in creating
high gain antennas. When very high gain is needed, one can
combine lenses with digital beamformers to construct a hybrid
array [1]. The second promising technology is photonic beam-
forming networks. In this approach, radio frequency phase
shifts are obtained either using a single mode fibre or variable
optical delay lines [143], [144].
Generally speaking, new systematic approaches would be
needed to create energy and cost optimized solutions for 5G
and B5G wireless communications networks.

VI. CONCLUSIONS
FIGURE 21. (a) CAD view of the connecting network generating a 5G and B5G wireless communications networks demand cost
triangular lattice of beams from a 2D Butler matrix. (b) Modified 2D Butler and energy efficient analogue multibeam antenna arrays. To
matrix under test [124].
date, the most promising solution to realize such arrays in
the lower part of the frequency spectrum considered for 5G,
labelled FR1 and corresponding to below 6 GHz frequency
attributed to imperfections in the matrix and variations in the bands, is to employ circuit type BFNs. This paper presented
impedances presented by the array radiators. Further chal- a comprehensive overview of circuit type BFN for mutibeam
lenges are the difference in beam shapes, scan losses and antenna arrays. Different circuit type BFNs were analyzed,
limited scanning angle. These are all important directions for particularly focusing on Butler matrices, Blass matrices and
future research. Nolen matrices. Various methods to improve the performance
Another interesting direction of research is the development of circuit type BFNs were discussed. New research challenges
of hybrid antenna arrays employing a combination of circuit to advance the state of the art in the field were also reviewed.
type BFNs and digital array processing techniques, otherwise It is anticipated that novel beamforming techniques and
known as hybrid arrays [137], [138]. Conventional phased technologies will emerge addressing the specific needs in the
arrays typically produce a very small number of simultaneous mm-wave frequency range, with further miniaturization and
steered beams using analogue phase shifters, often limited to integration. It is hoped that this paper will facilitate the dis-
a single beam capability. MIMO digital signal processing can cussion and research in this important field.
support dynamic and steerable multiple beams but it requires
high power consumption and hardware cost, while analogue
multiple beamforming networks produce multiple but fixed REFERENCES
beams at low cost. In a cellular network architecture, in- [1] Y. J. Guo and R. W. Ziolkowsk, Advanced Antenna Array Engineering
for 6G and Beyond Wireless Communications. Hoboken, NJ, USA:
cluding also some satellite and more generally non-terrestrial Wiley, 2021.
networks, continuous beam steering may not be necessary and [2] W. Hong et al., “Multibeam antenna technologies for 5G wireless
the multiple BFNs presented serve as a cost effective technol- communications,” IEEE Trans. Antennas Propag., vol. 65, no. 12,
pp. 6231–6249, Dec. 2017.
ogy for sub-dividing the field of view, or service area, into [3] I. Ahmed et al., “A survey on hybrid beamforming techniques in 5G:
fixed cells. Beam hopping is considered in satellite systems Architecture and system model perspectives,” IEEE Commun. Surveys
when the number of active beams is lower than the number Tuts., vol. 20, no. 4, pp. 3060–3097, Oct.-Dec. 2018.

VOLUME 1, NO. 3, JULY 2021 15

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

[4] X. You et al., “Towards 6G wireless communication networks: Vision, [31] J. W. Cooley, P. A. W. Lewis, and P. D. Welch, “Historical notes on the
enabling technologies, and new paradigm shifts,” Sci. China Inf. Sci., fast Fourier transform,” Proc. IEEE, vol. 55, no. 10, pp. 1675–1677,
vol. 64, no. 1, pp. 1–74, 2021. Oct. 1967.
[5] S. Cioni, R. De Gaudenzi, O. Del Rio Herrero, and N. Girault, “On the [32] J. P. Shelton, “Fast Fourier transforms and Butler matrices,” Proc.
satellite role in the era of 5G massive machine type communications,” IEEE, vol. 56, no. 3, pp. 350–350, Mar. 1968.
IEEE Netw., vol. 32, no. 5, pp. 54–61, Sep./Oct. 2018. [33] W. Nester, “The fast Fourier transform and the Butler matrix,” IEEE
[6] A. Guidotti et al., “Architectures and key technical challenges for 5G Trans. Antennas Propag., vol. 16, no. 3, pp. 360–360, May 1968.
systems incorporating satellites,” IEEE Trans. Veh. Technol., vol. 68, [34] J. P. Shelton, “Multibeam planar arrays,” Proc. IEEE, vol. 56, no. 11,
no. 3, pp. 2624–2639, Mar. 2019. pp. 1818–1821, Nov. 1968.
[7] F. Rinaldi et al., “Non-terrestrial networks in 5G beyond: A survey,” [35] P. Angeletti, “Multiple beams from planar arrays,” IEEE Trans. Anten-
IEEE Access, vol. 8, pp. 165 178–165 200, 2020. nas Propag., vol. 62, no. 4, pp. 1750–1761, Apr. 2014.
[8] X. Huang, J. A. Zhang, R. P. Liu, Y. J. Guo, and L. Hanzo, “Airplane- [36] W. Kahn and H. Kurss, “The uniqueness of the lossless feed network
aided integrated networking for 6G wireless: Will it work?”, IEEE Veh. for a multibeam array,” IRE Trans. Antennas Propag., vol. 10, no. 1,
Technol. Mag., vol. 14, no. 3, pp. 84–91, Sep. 2019. pp. 100–101, 1962.
[9] A. Anttonen, P. Ruuska, and M. Kiviranta, “3GPP nonterrestrial net- [37] P. Angeletti and M. Lisi, “Multimode beamforming networks for space
works: A concise review and look ahead,” VTT Res. Rep. No VTT-R- applications,” IEEE Antennas Propag. Mag., vol. 56, no. 1, pp. 62–78,
00079-19, 2019. Feb. 2014.
[10] S. Sun and T. S. Rappaport, “Multi-beam antenna combining for [38] S. P. Skobelev, “Methods of constructing optimum phased-array anten-
28 GHz cellular link improvement in urban environments,” in Proc. nas for limited field of view,” IEEE Antennas Propag. Mag., vol. 40,
IEEE Glob. Commun. Conf., 2013, pp. 3754–3759. no. 2, pp. 39–50, Apr. 1998.
[11] R. J. Mailloux, Phased Array Antenna Handbook. Norwood, MA, [39] D. Betancourt and C. del RioBocio, “A novel methodology to feed
USA: Artech House, 2017. phased array antennas,” IEEE Trans. Antennas Propag., vol. 55, no. 9,
[12] H. J. Visser, Array and Phased Array Antenna Basics, vol. 2005. New pp. 2489–2494, Sep. 2007.
York, NY, USA: Wiley, 2005. [40] N. Ferrando and N. J. G. Fonseca, “Investigations on the efficiency
[13] J. Butler and R. Lowe, “Beam-forming matrix simplifies design of of array fed coherently radiating periodic structure beam forming
electronically scanned antenna,” Electron. Des., vol. 9, pp. 170–173, networks,” IEEE Trans. Antennas Propag., vol. 59, no. 2, pp. 493–502,
1961. Feb. 2011.
[14] H. Moody, “The systematic design of the Butler matrix,” IEEE Trans. [41] N. J. G. Fonseca, “Discussion on tapered amplitude multiple beam
Antennas Propag., vol. 12, no. 6, pp. 786–788, 1964. forming networks for array antenna design,” in Proc. Int. Symp. An-
[15] J. Nolen, “Synthesis of multiple beam networks for arbitrary illumina- tennas Prop., Jeju, Korea, Oct. 2011, pp. 1–4.
tions,” Ph.D. dissertation, Radio Div., Bendix Corp., Baltimore, MD, [42] N. J. G. Fonseca, “Design and implementation of a closed cylindrical
USA, 1965. BFN-fed circular array antenna for multiple-beam coverage in az-
[16] N. J. G. Fonseca, “Printed S-band 4x4 Nolen matrix for multiple beam imuth,” IEEE Trans. Antennas Propag., vol. 60, no. 2, pp. 863–869,
antenna applications,” IEEE Trans. Antennas Propag., vol. 57, no. 6, Feb. 2012.
pp. 1673–1678, Jun. 2009. [43] H. E. A. Laue and W. P. du Plessis, “A checkered network for im-
[17] Y. J. Guo and B. Jones, “Base station antennas,” in Antenna Engi- plementing arbitrary overlapped feed networks,” IEEE Trans. Microw.
neering Handbook, J. Volakis ed. New York, NY, USA: McGraw-Hill Theory Techn., vol. 67, no. 11, pp. 4632–4640, Nov. 2019.
2018, ch. 10. [44] E. T. Der, T. R. Jones, and M. Daneshmand, “Miniaturized 4 × 4 But-
[18] W. Roh et al., “Millimeter-wave beamforming as an enabling tech- ler matrix and tunable phase shifter using ridged half-mode substrate
nology for 5G cellular communications: Theoretical feasibility and integrated waveguide,” IEEE Trans. Microw. Theory Techn., vol. 68,
prototype results,” IEEE Commun. Mag., vol. 52, no. 2, pp. 106–113, no. 8, pp. 3379–3388, Aug. 2020.
Feb. 2014. [45] I. Afifi and A.-R. Sebak, “Wideband 4 × 4 Butler matrix in the
[19] N. J. G. Fonseca, “Discussion on reciprocity, unitary matrix and loss- printed ridge gap waveguide technology for millimeter wave applica-
less multiple beam forming networks,” Int. J. Antennas Propag., vol. tions,” IEEE Trans. Antennas Propag., vol. 68, no. 11, pp. 7670–7675,
2015, no. 946289, pp. 1–9, 2015. Nov. 2020.
[20] S. Stein, “On cross coupling in multiple-beam antennas,” IRE Trans. [46] H. Ren et al., “A novel design of 4 × 4 Butler matrix with relatively
Antennas Propag., vol. 10, no. 5, pp. 548–557, 1962. flexible phase differences,” IEEE Antennas Wireless Propag. Lett.,
[21] W. White, “Pattern limitations in multiple-beam antennas,” IRE Trans. vol. 15, pp. 1277–1280, Dec. 2015.
Antennas Propag., vol. 10, no. 4, pp. 430–436, 1962. [47] Q. Wu, J. Hirokawa, J. Yin, C. Yu, H. Wang, and W. Hong,
[22] J. Blass, “Multidirectional antenna - A new approach to stacked “Millimeter-wave multibeam endfire dual-circularly polarized antenna
beams,” Proc. IRE, vol. 48, no. 3, pp. 402–402, 1960. array for 5G wireless applications,” IEEE Trans. Antennas Propag.,
[23] N. J. G. Fonseca, A. Ali, and H. Aubert, “Cancellation of beam vol. 66, no. 9, pp. 4930–4935, Sep. 2018.
squint with frequency in serial beamforming network-fed linear array [48] W. C. Cummings, “Multiple beam forming networks,” MIT Lincoln
antennas,” IEEE Antennas Propag. Mag., vol. 54, no. 1, pp. 32–39, Lab., Lexington, MA, USA, Tech. Note 1978-9, Apr. 1978.
Feb. 2012. [49] N. J. G. Fonseca, “An investigation of Blass and Nolen matrices,”
[24] S. Mosca, F. Bilotti, A. Toscano, and L. Vegni, “A novel design CNES Toulouse Space Center, Toulouse, France, Tech. Note 152,
method for Blass matrix beam-forming networks,” IEEE Trans. An- Nov. 2007.
tennas Propag., vol. 50, no. 2, pp. 225–232, Feb. 2002. [50] N. J. G. Fonseca, “Study and design of a S-band 4 × 4 Nolen matrix for
[25] W. R. Jones and E. C. DuFort, “On the design of optimum dual-series satellite digital multimedia broadcasting applications,” in Proc. 12th
feed networks,” IEEE Trans. Microw. Theory Techn., vol. 19, no. 5, Int. Symp. Antenna Technol. Appl. Electromagn. (ANTEM URSI/CNC),
pp. 451–458, May 1971. Montréal, QC, Canada, Jul. 2006, pp. 481–484.
[26] C. Tsokos et al., “Analysis of a multibeam optical beamforming net- [51] A. Ali et al., “Novel two-layer 4 × 4 SIW Nolen matrix for multi-beam
work based on Blass matrix architecture,” J. Lightw. Technol., vol. 36, antenna application in Ku-band,” in Proc. 3rd Eur. Conf. Antennas
no. 16, pp. 3354–3372, 2018. Propag., Berlin, Germany, Mar. 2009, pp. 241–243.
[27] C. Tsokos et al., “Optical beamforming networks supporting multi- [52] T. Djerafi, N. J. G. Fonseca, and K. Wu, “Planar Ku-band 4 x 4 Nolen
beam and multicast operation,” in Proc. 22nd Int. Conf. Transparent matrix in SIW technology,” IEEE Trans. Microw. Theory Techn.,
Opt. Netw., 2020, pp. 1–5. vol. 58, no. 2, pp. 259–266, Jan. 2010.
[28] J. Shelton and K. Kelleher, “Multiple beams from linear arrays,” IRE [53] T. Djerafi, N. J. G. Fonseca, and K. Wu, “Broadband substrate in-
Trans. Antennas Propag., vol. 9, no. 2, pp. 154–161, Mar. 1961. tegrated waveguide 4 × 4 Nolen matrix based on coupler delay
[29] J. Allen, “A theoretical limitation on the formation of lossless multiple compensation,” IEEE Trans. Microw. Theory Techn., vol. 59, no. 7,
beams in linear arrays,” IRE Trans. Antennas Propag., vol. 9, no. 4, pp. 1740–1745, Jul. 2011.
pp. 350–352, 1961. [54] H. Ren, H. Zhang, and B. Arigong, “Ultra-compact 3 x 3 Nolen matrix
[30] W. P. Delaney, “An RF multiple beam-forming technique,” IRE Trans. beamforming network,” IET Microw. Antennas Propag., vol. 14, no. 3,
Mil. Electron., no. 2, pp. 179–186, 1962. pp. 143–148, 2020.

16 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

[55] P. Li, H. Ren, and B. Arigong, “A symmetric beam-phased array fed by [78] G. Tian, J. Yang, and W. Wu, “A novel compact Butler matrix without
a Nolen matrix using 180◦ couplers,” IEEE Microw. Wireless Compon. phase shifter,” IEEE Microw. Wireless Compon. Lett., vol. 24, no. 5,
Lett., vol. 30, no. 4, pp. 387–390, Apr. 2020. pp. 306–308, May 2014.
[56] H. Ren, H. Zhang, Y. Jin, Y. Gu, and B. Arigong, “A novel 2D 3 x 3 [79] H. Nachouane et al., “Broadband 4 × 4 Butler matrix using wideband
Nolen matrix for 2D beamforming applications,” IEEE Trans. Microw. 90◦ hybrid couplers and crossovers for beamforming networks,” in
Theory Techn., vol. 67, no. 11, pp. 4622–4631, Nov. 2019. Proc. Int. Conf. Multimedia Comput. Syst., 2014, pp. 1444–1448.
[57] K. Wincza, S. Gruszczynski, and K. Sachse, “Reduced sidelobe four- [80] T. Djerafi and K. Wu, “A low-cost wideband 77-GHz planar Butler
beam antenna array fed by modified Butler matrix,” Electron. Lett., matrix in SIW technology,” IEEE Trans. Antennas Propag., vol. 60,
vol. 42, no. 9, pp. 508–509, 2006. no. 10, pp. 4949–4954, Oct. 2012.
[58] N. J. G. Fonseca and N. Ferrando, “Nolen matrix with tapered [81] Y. Cao, K. Chin, W. Che, W. Yang, and E. S. Li, “A compact
amplitude law for linear arrays with reduced side lobe level,” in 38 GHz multibeam antenna array with multifolded Butler matrix for
Proc. 4th Eur. Conf. Antennas Propag., Barcelona, Spain, Apr. 2010, 5G applications,” IEEE Antennas Wireless Propag. Lett., vol. 16,
pp. 1–5. pp. 2996–2999, Sep. 2017.
[59] S. Gruszczynski, K. Wincza, and K. Sachse, “Reduced sidelobe four- [82] A. A. M. Ali, N. J. G. Fonseca, F. Coccetti, and H. Aubert, “Design
beam N-element antenna arrays fed by 4 × N Butler matrices,” IEEE and implementation of two-layer compact wideband Butler matrices
Antennas Wireless Propag. Lett., vol. 5, pp. 430–434, May 2006. in SIW technology for Ku-band applications,” IEEE Trans. Antennas
[60] M. Ansari, H. Zhu, N. Shariati, and Y. J. Guo, “Compact planar Propag., vol. 59, no. 2, pp. 503–512, Feb. 2010.
beamforming array with endfire radiating elements for 5G applica- [83] S. Karamzadeh et al., “Compact and broadband 4 × 4 SIW Butler ma-
tions,” IEEE Trans. Antennas Propag., vol. 67, no. 11, pp. 6859–6869, trix with phase and magnitude error reduction,” IEEE Microw. Wireless
Nov. 2019. Compon. Lett., vol. 25, no. 12, pp. 772–774, Nov. 2015.
[61] S. Trinh-Van, J. M. Lee, Y. Yang, K. Lee, and K. C. Hwang, “A [84] L.-H. Zhong, Y. Ban, J. Lian, Q. Yang, J. Guo, and Z. Yu, “Miniatur-
sidelobe-reduced, four-beam array antenna fed by a modified 4 × 4 ized SIW multibeam antenna array fed by dual-layer 8 × 8 Butler ma-
Butler matrix for 5G applications,” IEEE Trans. Antennas Propag., trix,” IEEE Antennas Wireless Propag. Lett., vol. 16, pp. 3018–3021,
vol. 67, no. 7, pp. 4528–4536, Jul. 2019. Oct. 2017.
[62] D. M. Pozar and B. Kaufman, “Design considerations for low sidelobe [85] M. Nedil, T. A. Denidni, and L. Talbi, “Novel Butler matrix using
microstrip arrays,” IEEE Trans. Antennas Propag., vol. 38, no. 8, CPW multilayer technology,” IEEE Trans. Microw. Theory Techn.,
pp. 1176–1185, Aug. 1990. vol. 54, no. 1, pp. 499–507, Jan. 2006.
[63] J. Shelton, “Reduced sidelobes for Butler-matrix-fed linear arrays,” [86] E. Gandini, M. Ettorre, R. Sauleau, and A. Grbic, “A lumped-element
IEEE Trans. Antennas Propag., vol. 17, no. 5, pp. 645–647, Sep. 1969. unit cell for beam-forming networks and its application to a miniatur-
[64] W.-R. Li et al., “Switched-beam antenna based on modified But- ized Butler matrix,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 4,
ler matrix with low sidelobe level,” Electron. Lett., vol. 40, no. 5, pp. 1477–1487, Apr. 2013.
pp. 290–292, 2004. [87] Q. Lai, C. Fumeaux, W. Hong, and R. Vahldieck, “Characterization
[65] J. Hirokawa, K. Morimoto, and M. Ando, “Beam-switching single- of the propagation properties of the half-mode substrate integrated
layer post-wall waveguide slot array controlling crossover & side- waveguide,” IEEE Trans. Microw. Theory Techn., vol. 57, no. 8,
lobes,” in Proc. 2nd Eur. Conf. Antennas Propag., Edinburgh, U.K., pp. 1996–2004, Aug. 2009.
Nov. 2007. [88] Y. Ding and K. Wu, “A 4 × 4 ridge substrate integrated waveguide
[66] K. Wincza et al., “Reduced sidelobe multibeam antenna array with (rsiw) slot array antenna,” IEEE Antennas Wireless Propag. Lett.,
broadside beam fed by 4 × 8 Butler matrix,” in Proc. Int. Symp. vol. 8, pp. 561–564, Apr. 2009.
Antennas Propag., Hobart, Australia, Nov. 2015, pp. 1–3. [89] N. Nguyen-Trong, T. Kaufmann, L. Hall, and C. Fumeaux, “Vari-
[67] B. Strait, “Antenna arrays with partially tapered amplitudes,” IEEE ational analysis of folded substrate-integrated waveguides,” IEEE
Trans. Antennas Propag., vol. 15, no. 5, pp. 611–617, Sep. 1967. Microw. Wireless Compon. Lett., vol. 25, no. 6, pp. 352–354,
[68] F. E. Fakoukakis, G. A. Kyriacou, and J. N. Sahalos, “On the design of Jun. 2015.
Butler-like type matrices for low SLL multibeam antennas,” in Proc. [90] Q. Sun, Y. Ban, J. Lian, Y. Liu, and Z. Nie, “Millimeter-wave multi-
6th Eur. Conf. Antennas Propag., Prague, Czech Republic, Mar. 2012, beam antenna based on folded c-type SIW,” IEEE Trans. Antennas
pp. 2604–2608. Propag., vol. 68, no. 5, pp. 3465–3476, May 2020.
[69] A. Fragola, M. Orefice, and M. Pirola, “A modified Butler matrix for [91] Y. Wang, K. Ma, N. Yan, and L. Li, “A slow-wave rat-race cou-
tapered excitation of scanned arrays,” in IEEE Antennas Propag. Soc. pler using substrate integrated suspended line technology,” IEEE
Int. Symp. Dig., vol. 4, 2001, pp. 784–787. Trans. Compon. Packag. Manuf. Technol., vol. 7, no. 4, pp. 630–636,
[70] E. DuFort, “Optimum low sidelobe high crossover multiple beam Apr. 2017.
antennas,” IEEE Trans. Antennas Propag., vol. 33, no. 9, pp. 946–954, [92] Y. Wang, K. Ma, and Z. Jian, “A low-loss Butler matrix us-
Sep. 1985. ing patch element and honeycomb concept on SISL platform,”
[71] H. Nasseri, M. Bemani, and A. Ghaffarlou, “A new method for ar- IEEE Trans. Microw. Theory Techn., vol. 66, no. 8, pp. 3622–3631,
bitrary amplitude distribution generation 4 × 8 in Butler matrix,” Aug. 2018.
IEEE Microw. Wireless Compon. Lett., vol. 30, no. 3, pp. 249–252, [93] N. Ashraf, A.-R. Sebak, and A. A. Kishk, “PMC packaged single-
Mar. 2020. substrate 4 × 4 Butler matrix and double-ridge gap waveguide horn
[72] K. Wincza, S. Gruszczynski, and K. Sachse, “Conformal four-beam antenna array for multibeam applications,” IEEE Trans. Microw. The-
antenna arrays with reduced sidelobes,” Electron. Lett., vol. 44, no. 3, ory Techn., vol. 69, no. 1, pp. 248–261, Jan. 2021.
pp. 174–175, 2008. [94] K. Wincza, S. Gruszczynski, and K. Sachse, “Broadband planar fully
[73] K. Tekkouk, J. Hirokawa, R. Sauleau, M. Ettorre, M. Sano, and M. integrated 8 x 8 Butler matrix using coupled-line directional couplers,”
Ando, “Dual-layer ridged waveguide slot array fed by a Butler matrix IEEE Trans. Microw. Theory Techn., vol. 59, no. 10, pp. 2441–2446,
with sidelobe control in the 60-GHz band,” IEEE Trans. Antennas Sep. 2011.
Propag., vol. 63, no. 9, pp. 3857–3867, Sep. 2015. [95] H. Nachouane et al., “Wideband 3 × 4 Butler matrix using Wilkinson
[74] F. Doucet et al., “Shaped continuous parallel plate delay lens with divider for MIMO applications,” in Proc. Int. Conf. Next Gener. Netw.
enhanced scanning performance,” IEEE Trans. Antennas Propag., Serv., 2014, pp. 101–105.
vol. 67, no. 11, pp. 6695–6704, Nov. 2019. [96] K. Wincza and S. Gruszczynski, “Broadband integrated 8 x 8 Butler
[75] K. Wu, M. Bozzi, and N. J. G. Fonseca, “Substrate integrated trans- matrix utilizing quadrature couplers and Schiffman phase shifters for
mission lines: Review and applications,” IEEE J. Microwaves, vol. 1, multibeam antennas with broadside beam,” IEEE Trans. Microw. The-
no. 1, pp. 345–363, Jan. 2021. ory Techn., vol. 64, no. 8, pp. 2596–2604, Aug. 2016.
[76] H. Hayashi, D. A. Hitko, and C. G. Sodini, “Four-element planar But- [97] D. M. Pozar, Microwave Engineering. Hoboken, NJ, USA: Wiley,
ler matrix using half-wavelength open stubs,” IEEE Microw. Wireless 2011.
Compon. Lett., vol. 12, no. 3, pp. 73–75, Mar. 2002. [98] Q. P. Chen, Z. Qamar, S. Y. Zheng, Y. Long, and D. Ho, “Design of
[77] C. Dall’Omo et al., “Design and realization of a 4 × 4 microstrip a compact wideband Butler matrix using vertically installed planar
Butler matrix without any crossing in millimeter waves,” Microw. Opt. structure,” IEEE Trans. Compon. Packag. Manuf. Technol., vol. 8,
Technol. Lett., vol. 38, no. 6, pp. 462–465, 2003. no. 8, pp. 1420–1430, Aug. 2018.

VOLUME 1, NO. 3, JULY 2021 17

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GUO ET AL.: CIRCUIT TYPE MULTIPLE BEAMFORMING NETWORKS FOR ANTENNA ARRAYS

[99] C. Liu, S. Xiao, Y. Guo, M. Tang, Y. Bai, and B. Wang, “Circularly [120] Y. J. Cheng, X. Y. Bao, and Y. X. Guo, “60-GHz LTCC miniaturized
polarized beam-steering antenna array with Butler matrix network,” substrate integrated multibeam array antenna with multiple polariza-
IEEE Antennas Wireless Propag. Lett., vol. 10, pp. 1278–1281, Nov. tions,” IEEE Trans. Antennas Propag., vol. 61, no. 12, pp. 5958–5967,
2011. Dec. 2013.
[100] J. Huang, “A technique for an array to generate circular polarization [121] S. Yamamoto, J. Hirokawa, and M. Ando, “A single-layer hollow-
with linearly polarized elements,” IEEE Trans. Antennas Propag., waveguide 8-way Butler matrix,” in Proc. IEEE Antennas Propag. Soc.
vol. 34, no. 9, pp. 1113–1124, Sep. 1986. Int. Symp., vol. 1A, 2005, pp. 647–650.
[101] H. Zhu, H. Sun, B. Jones, C. Ding, and Y. J. Guo, “Wideband dual- [122] Q. Yang, Y. Ban, K. Kang, C. Sim, and G. Wu, “SIW multibeam
polarized multiple beam-forming antenna arrays,” IEEE Trans. Anten- array for 5G mobile devices,” IEEE Access, vol. 4, pp. 2788–2796,
nas Propag., vol. 67, no. 3, pp. 1590–1604, Mar. 2019. 2016.
[102] D. I. Lialios et al., “Design of true time delay millimeter wave beam- [123] X. Jun-Feng, “A novel Butler matrix based on half-mode substrate
formers for 5G multibeam phased arrays,” Electronics, vol. 9, no. 8, integrated waveguide technology,” J. Microw., vol. 24, no. 3, pp. 5–9,
2020, Art no. 1331. Nov. 2008.
[103] H. Ren et al., “Compact phased array antenna system based on [124] N. J. G. Fonseca et al., “Connecting networks for 2D Butler matrices
dual-band operations,” Microw. Opt. Technol. Lett., vol. 56, no. 6, generating a triangular lattice of beams,” IEEE J. Microwaves, vol. 1,
pp. 1391–1396, 2014. no. 2, pp. 646–658, Apr. 2021.
[104] J. Lee and K. Chang, “Dual-band switched beam array fed by dual- [125] D.-H. Kim, J. Hirokawa, and M. Ando, “Design of waveguide shortslot
band Butler matrix,” Electron. Lett., vol. 47, no. 21, pp. 1164–1165, 2-plane couplers for one-body 2D beam-switching Butler matrix appli-
2011. cation,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 3, pp. 776–
[105] N. Amiri and K. Forooraghi, “Dual-band and dual-polarized mi- 784, Mar. 2016.
crostrip array antenna for GSM900/DCS1800 MHz base stations,” in [126] J. Hirokawa, D. Kim, M. Wakasa, Y. Sunaguchi, T. Tomura,
Proc. IEEE Antennas Propag. Soc. Int. Symp., 2006, pp. 4439–4442. and K. Nishimori, “Measurements of a 64x64-way one-body two-
[106] K. Wincza, K. Staszek, I. Slomian, and S. Gruszczynski, “Scalable dimensional beam-switching hollow-waveguide Butler matrix,” in
multibeam antenna arrays fed by dual-band modified Butler matri- Proc. Eur. Microw. Week, Madrid, Spain, Sep. 2018, pp. 125–128.
ces,” IEEE Trans. Antennas Propag., vol. 64, no. 4, pp. 1287–1297, [127] T. Tomura et al., “A 20-GHz-band 64 × 64 hollow waveguide two-
Apr. 2016. dimensional Butler matrix,” IEEE Access, vol. 7, pp. 164 0 80–164
[107] A. M. Zaidi, B. K. Kanaujia, M. T. Beg, J. Kishor, and K. Rambabu, 088, Nov. 2019.
“A novel dual-band branch line coupler for dual-band Butler matrix,” [128] N. J. G. Fonseca, S.-A. Gomanne, and J. Hirokawa, “Connecting
IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 66, no. 12, pp. 1987– networks for 2D Butler matrices with improved aggregate gain,” in
1991, Dec. 2019. Proc. Int. Symp. Antennas Propag., no. 2E4.2, Osaka, Japan, Jan. 2021,
[108] A. M. Zaidi, “A new dual band 4 × 4 Butler matrix with dual band pp. 285–286.
3 dB quadrature branch line coupler and dual band 45◦ phase shifter,” [129] H. Xin et al., “A two-dimensional millimeter wave phase scanned lens
AEU-Int. J. Electron. Commun., vol. 99, pp. 215–225, 2019. utilizing analog electromagnetic crystal (EMXT) waveguide phase
[109] D.-H. Kim, J. Hirokawa, and M. Ando, “Design of waveguide short- shifters,” IEEE Trans. Antennas Propag., vol. 53, no. 1, pp. 151–159,
slot two-plane couplers for one-body 2D beam-switching Butler ma- Jan. 2005.
trix application,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 3, [130] O. Quevedo-Teruel, M. Ebrahimpouri, and F. Ghasemifard, “Lens
pp. 776–784, Mar. 2016. antennas for 5G communications systems,” IEEE Commun. Mag.,
[110] K. Ding and A. A. Kishk, “2D Butler matrix and phase-shifter group,” vol. 56, no. 7, pp. 36–41, Jul. 2018.
IEEE Trans. Microw. Theory Techn., vol. 66, no. 12, pp. 5554–5562, [131] Q. Liao, N. J. G. Fonseca, and O. Quevedo-Teruel, “Compact multi-
Dec. 2018. beam fully metallic geodesic Luneburg lens antenna based on non-
[111] J.-W. Lian, Y. Ban, Q. Yang, B. Fu, Z. Yu, and L. Sun, “Pla- euclidean transformation optics,” IEEE Trans. Antennas Propag.,
nar millimeter-wave 2D beam-scanning multibeam array antenna fed vol. 66, no. 12, pp. 7383–7388, Dec. 2018.
by compact SIW beam-forming network,” IEEE Trans. Antennas [132] Y. Geng, J. Wang, Y. Li, Z. Li, M. Chen, and Z. Zhang, “Radiation
Propag., vol. 66, no. 3, pp. 1299–1310, Mar. 2018. pattern-reconfigurable leaky-wave antenna for fixed-frequency beam
[112] J.-W. Lian, Y. Ban, Q. Yang, B. Fu, Z. Yu, and L. Sun, “Unipla- steering based on substrate-integrated waveguide,” IEEE Antennas
nar beam-forming network employing eight-port hybrid couplers and Wireless Propag. Lett., vol. 18, no. 2, pp. 387–391, Feb. 2019.
crossovers for 2D multibeam array antennas,” IEEE Trans. Microw. [133] N. J. G. Fonseca, Q. Liao, and O. Quevedo-Teruel, “Equivalent planar
Theory Techn., vol. 68, no. 11, pp. 4706–4718, Nov. 2020. lens ray-tracing model to design modulated geodesic lenses using
[113] W. Yang et al., “94-GHz compact 2D multibeam LTCC antenna based non-euclidean transformation optics,” IEEE Trans. Antennas Propag.,
on multifolded SIW beam-forming network,” IEEE Trans. Antennas vol. 68, no. 5, p. 3410–3422, May 2020.
Propag., vol. 65, no. 8, pp. 4328–4333, Aug. 2017. [134] N. J. G. Fonseca, Q. Liao, and O. Quevedo-Teruel, “Compact
[114] X. Wang, Y. Yang, W. Che, C. Fan, and Q. Xue, “Compact 2D parallel-plate waveguide half-Luneburg geodesic lens in the Ka-
multibeam array antenna fed by planar cascaded Butler matrix for band,” IET Microw. Antennas Propag., vol. 15, no. 2, pp. 123–130,
millimeter-wave communication,” IEEE Antennas Wireless Propag. Feb. 2021.
Lett., vol. 18, no. 10, pp. 2056–2060, 2019. [135] Y. Park, J. Bang, Y. Kim, and J. Choi, “Beam-steerable dual-circular
[115] J.-W. Lian, Y. Ban, J. Zhu, K. Kang, and Z. Nie, “Compact 2D polarized 60 GHz patch antenna array with 8 × 8 Butler matrix,” in
scanning multibeam array utilizing the SIW three-way couplers at Proc. Int. Symp. Antennas Propag., 2019, pp. 1–3.
28 GHz,” IEEE Antennas Wireless. Propag. Lett., vol. 17, no. 10, [136] A. Li and K.-M. Luk, “Single-layer wideband end-fire dual-polarized
pp. 1915–1919, Oct. 2018. antenna array for device-to-device communication in 5G wireless
[116] J.-W. Lian, Y. Ban, J. Zhu, K. Kang, and Z. Nie, “Planar 2D scan- systems,” IEEE Trans. Veh. Technol., vol. 69, no. 5, pp. 5142–5150,
ning SIW multibeam array with low sidelobe level for millimeter- May 2020.
wave applications,” IEEE Trans. Antennas Propag., vol. 67, no. 7, [137] S. Han et al., “Large-scale antenna systems with hybrid analog and
pp. 4570–4578, Jul. 2019. digital beamforming for millimeter wave 5G,” IEEE Commun. Mag.,
[117] J.-W. Lian, Y. Ban, J. Zhu, K. Kang, and Z. Nie, “Uniplanar high-gain vol. 53, no. 1, pp. 186–194, Jan. 2015.
2D scanning leaky-wave multibeam array antenna at fixed frequency,” [138] X. Huang, Y. J. Guo, and J. D. Bunton, “A hybrid adaptive antenna
IEEE Trans. Antennas Propag., vol. 68, no. 7, pp. 5257–5268, array,” IEEE Trans. Wireless Commun., vol. 9, no. 5, pp. 1770–1779,
Jul. 2020. May 2010.
[118] M. M. M. Ali and A.-R. Sebak, “2D scanning magnetoelectric dipole [139] N. J. G. Fonseca and J. Sombrin, “Multi-beam reflector antenna
antenna array fed by RGW Butler matrix,” IEEE Trans. Antennas system combining beam hopping and size reduction of effectively
Propag., vol. 66, no. 11, pp. 6313–6321, Nov. 2018. used spots,” IEEE Antennas Propag. Mag., vol. 54, no. 2, pp. 88–99,
[119] W. F. Moulder, W. Khalil, and J. L. Volakis, “60-GHz two- Apr. 2012.
dimensionally scanning array employing wideband planar switched [140] P. Angeletti and R. De Gaudenzi, “A pragmatic approach to massive
beam network,” IEEE Antennas Wireless Propag. Lett., vol. 9, MIMO for broadband communication satellites,” IEEE Access, vol. 8,
pp. 818–821, Aug. 2010. pp. 132 212–132 236, 2020.

18 VOLUME 1, NO. 3, JULY 2021

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

[141] Y. Kawamoto, T. Kamei, M. Takahashi, N. Kato, A. Miura, and MARAL ANSARI (Student Member, IEEE) re-
M. Toyoshima, “Flexible resource allocation with inter-beam inter- ceived the M.S. degree (Hons.) in electrical and
ference in satellite communication systems with a digital channel- communication engineering from Tabriz Univer-
izer,” IEEE Trans. Wireless Commun., vol. 19, no. 5, pp. 2934–2945, sity, Tabriz, Iran, in 2015. She is currently working
May 2020. toward the Ph.D. degree in electrical engineering
[142] A. Wang et al., “Noma-enabled multi-beam satellite systems: Joint with the Global Big Data Technologies Center,
optimization to overcome offered-requested data mismatches,” IEEE School of Electrical and Data Engineering, Faculty
Trans. Veh. Technol., vol. 70, no. 1, pp. 900–913, Jan. 2021. of Engineering and IT, University of Technology
[143] M. Y. Frankel and R. D. Esman, “True time-delay fiber-optic con- Sydney (UTS), Ultimo, NSW, Australia. Her cur-
trol of an ultrawideband array transmitter/receiver with multibeam rent research interests include multi-beam antenna
capability,” IEEE Trans. Microw. Theory Techn., vol. 43, no. 9, arrays, lens antennas, mm-wave beam scanning,
pp. 2387–2394, Sep. 1995. and beamforming networks.
[144] T. Nagayama et al., “Photonics-based millimeter-wave band remote
beamforming of array-antenna integrated with photodiode using vari-
able optical delay line and attenuator,” J. Lightw. Technol., vol. 36,
no. 19, pp. 4416–4422, 2018.

Y. JAY GUO (Fellow, IEEE) received the bach- NELSON J. G. FONSECA (Senior Member,
elor’s and master’s degrees from Xidian Univer- IEEE) received the M.Eng. degree from Ecole Na-
sity, China, in 1982 and 1984, respectively, and tionale Supérieure d’Electrotechnique, Electron-
the Ph.D. degree from Xian Jiaotong University, ique, Informatique, Hydraulique et Télécommuni-
China, in 1987. cations (ENSEEIHT), Toulouse, France, in 2003,
He is a Distinguished Professor and the Director the M.Sc. degree from the Ecole Polytechnique de
of Global Big Data Technologies Centre (GBDTC) Montreal, Quebec, Canada, also in 2003, and the
at the University of Technology Sydney (UTS), Ph.D. degree from Institut National Polytechnique
Australia. Prior to this appointment in 2014, he de Toulouse – Université de Toulouse, France, in
served as a Director in CSIRO for over nine years. 2010, all in electrical engineering.
Before joining CSIRO, he held various senior tech- He currently works as an Antenna Engineer for
nology leadership positions in Fujitsu, Siemens and NEC in the U.K. His the Antenna and Sub-Millimetre Waves Section, European Space Agency
research interest includes antennas, mm-wave and THz communications and (ESA), Noordwijk, The Netherlands. Since November 2020, he has held an
sensing systems as well as big data technologies. He has published over 550 Honorary Appointment as Professional Fellow at the University of Technol-
research papers including 280 journal papers, most of which are in IEEE ogy Sydney (UTS), Australia. His research interests include multiple beam
Transactions, and he holds 26 patents. antennas for space missions, beam-former theory and design, ground terminal
Prof. Guo is a Fellow of the Australian Academy of Engineering and antennas, transfer of technology from and to terrestrial systems, including
Technology, a Fellow of IEEE and a Fellow of IET, and was a member of 5G networks, and novel manufacturing techniques. He has authored or co-
the College of Experts of Australian Research Council (ARC, 2016–2018). authored more than 220 papers in peer-reviewed journals and conferences
He has won a number of most prestigious Australian Engineering Excellence and has over 50 patents issued or pending.
Awards (2007, 2012) and CSIRO Chairman’s Medal (2007, 2012). He was Dr. Fonseca served as the Chair of the 38th ESA Antenna Workshop in
named one of the most influential engineers in Australia in 2014 and 2015, 2017, and as the Co-Chair of the 2018 IET Loughborough Antennas &
respectively, and one of the top researchers in Australia in 2020. Propagation Conference (LAPC 2018). He is currently serving as an Asso-
He has chaired numerous international conferences and served as guest ciate Editor for IET Microwaves, Antennas and Propagation and for IEEE
editors for a number of IEEE publications. He is the Chair of the Inter- TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, and as a Topic
national Steering Committee, International Symposium on Antennas and Editor for IEEE JOURNAL OF MICROWAVES. He has been also serving as the
Propagation (ISAP). He was the International Advisory Committee Chair Co-Vice Chair of the newly founded IEEE MTT-S Technical Committee 29
of IEEE VTC2017, General Chair of ISAP2022, ISAP2015, iWAT2014 and (MTT-29) on Microwave Aerospace Systems. He is a board member of the
WPMC’2014, and TPC Chair of 2010 IEEE WCNC and 2012 and 2007 IEEE European School of Antennas and Propagation (ESoA) since January 2019
ISCIT. He served as Guest Editor of special issues on “Antennas for Satellite and is also serving as coordinator of the ESA/ESoA course on Antennas for
Communications” and “Antennas and Propagation Aspects of 60-90 GHz Space Applications, for which he was voted best lecturer by the participants
Wireless Communications,” both in IEEE TRANSACTIONS ON ANTENNAS AND of the 2020 edition. He is the elected EurAAP Regional Delegate representing
PROPAGATION, Special Issue on “Communications Challenges and Dynamics Benelux for the term 2021–2023. He received several prizes and awards,
for Unmanned Autonomous Vehicles,” IEEE JOURNAL ON SELECTED AREAS including the Best Young Engineer Paper Award at the 29th ESA Antenna
IN COMMUNICATIONS, and Special Issue on “5G for Mission Critical Machine Workshop in 2007, an ESA Teamwork Excellence Award in 2020 and multi-
Communications, IEEE NETWORK MAGAZINE. ple ESA technical improvement awards.

VOLUME 1, NO. 3, JULY 2021 19