Novel Dual-Band Beam-Scanning/Switching Network Based on a Hybrid Coupler with Synchronously Tuned Phase Differences

This paper proposes a novel feeding network that enables the antenna array to perform dualband continuous beam scanning and beam switching between forward and backward directions. Specifically, the switchable beam is controlled by the network input and the beam in each direction is continuously steerable at either frequency. To support a wide range of beam-scanning angles, a novel dual-band coupler with synchronously tuned differential phases is applied to the feeding network, together with a pair of phase shifters. The operating principles and design equations of the dual-band beam-scanning/switching network, coupler, and phase shifters will be given in detail. As a proof-of-concept, a 0.9-GHz/1.8-GHz beamscanning/switching network was prototyped alongside a 1W4 linear array. Experimental results show that the array can scan and switch its main beam from -24° to -42° in the backward direction and from 24° to 34° in the forward direction at 0.9 GHz, while the beam scanning is from -12° to -24° and from 12° to 22° at 1.8 GHz.


I. INTRODUCTION
The beam-scanning or beam-switching antenna array in nowadays wireless communications has gained everincreasing interest due to its adaptive capability to increase the signal to noise ratio and thus improve channel capacity in spite of the existing multipath fading and various sources of electromagnetic interference. To produce a scanned beam at a frequency, the composite right/left-handed leaky-wave antenna provides a solution to the fixed-frequency beam scanning by electronically tuning the antenna circuit parameters [1], [2]. On the other hand, the beam-scanning array is generally driven by a feeding network, where the 360 o controllable phase progression between antenna ports is theoretically necessary to scan the entire spatial range. While a greater phase shifting range will extend beam coverage of the antenna system, designing the feeding network becomes more challenging as a result. Thus, a practical solution is to take account of size, complexity, cost, and the achievable performance of a beam-scanning network. In [3], an additional beam-steering network was used to form more beams by appropriately weighting the excitation ratio between two beam ports of the Butler matrix. Tunable phase shifters were incorporated into a conventional 8-way Butler matrix to contribute to a small steering range around the original (switched) beam [4]. For an acceptable quantization loss, the 1-bit or 2-bit PIN-controlled reflectarrays [5]- [9] were realized to alleviate design complexity and element loss necessary for implementing continuous phase shift. Moreover, the array driven by the beam-switching network is an alternative solution to providing multiple characteristic beams toward different angles. The Butler matrix, for example, is a popular choice and its extended capabilities have been extensively studied in the literature [3], [4], [10]- [17]. By replacing quadrature couplers with nonstandardphase-difference couplers [10], [16], the switched beam can be engineered to other angles. In [14] and [15], the filtering Butler matrix is inherited from the filtering 180 o hybrid coupler. To carry out dual-band operation, the Butler matrix was built with dual-band components [12], [13]. Besides, the single-ended-to-balanced power divider can feed the 1×4 differential array to produce two switched beams [17].
It is noted that, in the literature, the beamscanning/switching networks are mainly designed with single-band scheme and the array feeding network is usually on a basis of beam-switching operational mechanism with reduced spatial resolution (by discrete beams). To lend itself to multiple operating standards and to cover the spatial range in a continuous manner, this work proposes a novel, singlelayer, and microstrip feeding network that allows beam scanning and switching at dual frequencies of interest. Specifically, a dual-band hybrid coupler with two equal and tunable differential phases is the building block of the dualband feeding network, as shown in Fig. 1. It will be shown that the proposed coupler can achieve a wide range of tunable phase differences at both frequencies, which will contribute to an extended beam-scanning range. To enable the proposed beam-scanning/switching network to be utilized at two arbitrary frequencies, a detailed coupler study of flexible dual-band realization will also be conducted. The tunable phase progression between adjacent antenna ports in terms of the coupler's differential phase will be given for both inputs and depending on the input port, continuous beam steering can be fulfilled in the forward or backward direction. The operating principles of the proposed feeding network and the design techniques of the dual-band phase tunable coupler and phase shifters will be provided with a thorough discussion. As a proof-of-concept, a microstrip feeding network along with a 1×4 linear array was prototyped at 0.9 GHz and 1.8 GHz and was experimentally tested. The 1×4 array can demonstrate a scanning beam at both frequencies and the main beam scans in the forward or backward direction by switching between two input ports. Fig. 1 shows the schematic diagram of proposed dual-band beam-scanning/switching network, consisting mainly of three identical dual-band phase tunable couplers in two-stage configuration. Two kinds of tunable phase shifters (PS) 1 and 2 are used in-between stages to connect couplers and the connection is controlled by two single-pole double-throw (SPDT) switches. Input ports 1 and 2 are used to switch beam direction; ports 3-6 are the antenna ports of the network and each will be connected to 1×4 antenna elements. Note that to reduce integration complexity, each individual component is designed to match 50 W and thus when passing through it, only the transmission phase is considered. To ensure isolation between ports 1 and 2 and between either two of ports 3 to 6 of the network, the internal ports with port numbering 1 ! and 2 ! or 3 ! and 4 ! are the isolated pair of the proposed tunable coupler and this port numbering convention will be kept consistent with that used in Section III. For the excitation from coupler port 1 ! or 2 ! , the resulting phase difference f1 or f2, respectively, of the tunable coupler is defined as

A. SCHEMATIC DIAGRAM AND OPERATING PRINCIPLES
(1b) This feeding network has two operating modes. When port 1 is the input port, the SPDTs 1 and 2 are switched to PS 1 and PS 2, respectively, i.e., to the upper paths for both SPDTs, and thus the phase differences between adjacent outputs Dj34, Dj45, and Dj56 are given by where Djps1 and Djps2 are the introduced phase shifts of PS 1 and PS 2, respectively. Thus, a constant phase progression will require the phase shifts Djps1 and Djps2 to satisfy: (3) On the other hand, when port 2 is the input port, the SPDTs 1 and 2 are both switched to the lower paths, PS 2 and PS 1, respectively. The phase progression Dj34, Dj45, or Dj56 is Similarly, the required phase relation can be found as (3) and (5), it can be shown that, if the differential phase Djps1-Djps2 between PS 1 and PS 2 can be synchronously tuned with the coupler's phase difference f1,2, the main beam of the antenna array can be continuously steered; furthermore, since the phase progression between antenna ports has the form of f1 or -f2, the antenna array has greater potential to scan its main beam in either forward and backward direction by simply switching the excitation port. The latter is an important advantage for simplifying the coupler design. The structural analysis and design of the proposed coupler will be comprehensively detailed in Section III.

B. VARACTOR-TUNED LUMPED-ELEMENT BASED PHASE SHIFTERS
To support dual-band beam steering, the phase shifters PS 1 and PS 2 are also necessary to provide tunable phase shifts at both frequencies of interest. To simplify circuit complexity, the lumped-element based lowpass network is paired with a highpass network to implement the tunable PS 1 and PS 2, respectively, as shown in where w is the angular frequency. Thus, the overall phase shift Djps1 or Djps2 is equal to the sum of the phase shift of each corresponding unit cell or the incremental phase shift multiplied by the number of unit cells. By conversion between ABCD and S network parameters, the Djps1, Djps2, and the impedance matching conditions are given by where N is the number of unit cells in a network and Z0=50 W. The required differential phase between lowpass and highpass networks, as dictated by (3) and (5), can therefore be obtained by simultaneously controlling Cl and Ch in PS 1 and PS 2, respectively. To validate the applicability of the paired lowpass and highpass networks in the proposed dual-band beam-scanning network, the varactor-tuned lumped-element based PS 1 and PS 2 were prototyped for dual operating frequencies at 0.9 GHz and 1.8 GHz, and the measured/simulated magnitude and phase responses are shown in Fig. 3   the prescribed phase differences at dual frequencies, the required inductances and capacitances can be preliminarily decided by (7). Furthermore, to maintain good impedance matching across a wide range of phase differences, the fixed inductances L1 to L4 were properly optimized. In this design, N=3 and the inductances L1= 2.2 nH, L2= 4.3 nH, L3= 15 nH, and L4= 7.2 nH. Measured and simulated results are in good agreement. In Fig. 3 (b), the measured phase difference between two paths was synchronously controlled at 0.9/1.8 GHz and continuously tuned from 60 o to 120 o . In Fig. 3 (c), the measured return loss is greater than 15 dB at either frequency for both PS 1 and PS 2, thus ensuring good power transmission when applied to the proposed feeding network.

III. PROPOSED DUAL-BAND HYBRID COUPLER WITH SYNCHRONOUSLY TUNED PHASE DIFFERENCES
As aforementioned, the proper operation of the proposed dual-band beam-scanning/switching network needs a wellperformed coupler that enables synchronous phase engineering at two flexible frequencies. To this end, a varactor-based dual-band hybrid coupler with two synchronously tunable phase differences is proposed for the first time to offer an adaptive solution without the need of incorporating all functional hardware with substantial increase in complexity, size, and loss.
Electronically tunable couplers with tunable operating frequency [18], [19], tunable power-dividing ratio [20]- [22], or both [23]- [25] were extensively reported. Owing to the adaptability to providing flexible phase compensation in a beamformer, such as in the switched-beam Butler and Nolen matrices, and the contribution to optimum performance of reflectometers [26], [27], a coupler with tunable phase difference between through and coupled ports has received increasing attention and several tunable or switchable solutions [28]- [33] were proposed in the recent literature. In [29], the switchable coupler can realize three discrete phase differences. To extend the phase coverage, varactor-tuned couplers were studied to fulfill the continuously controllable differential phase at a fixed frequency [30], [32]. The frequency agility was also demonstrated in addition to the phase control [28], [31], [33]. To allow for a greater degree of flexibility and applicability in the existing and emerging multiband applications, however, the simultaneous dualband and continuous phase engineering is increasingly useful and is not explored till now.

A. COUPLER ANALYSIS
The schematic of proposed tunable coupler with dualfrequency phase control is shown in Fig. 4. This coupler consists of two host microstrip lines of characteristic impedance " and electrical length q1, two horizontal microstrip lines of characteristic impedance % and electrical length q2, and two groups of stub networks with the respective input impedances of jZs1 or jZs2. Note that to carry out tunable phase differences, varactor diodes will be used in the stub networks. When port 1 is the input port, port 2 is the isolated port while ports 3 and 4 will be two outputs. Therefore, to ensure ideal impedance match and port isolation, this proposed coupler has to hold the condition in (8) at either operating frequency: The phase difference f1 and f2, and power division ratio K are defined as where f1 and f2 are the seen phase differences between output ports 3 and 4 with the input ports 1 and 2, respectively.
It can be shown that f1 and f2 are supplementary angles, i.e., f1+f2=180 o , and thus only f1 will be used in the mathematical derivations that follow. As seen, this coupler is symmetric with respect to the plane of symmetry PP' and thus the even-odd mode decomposition technique is used for coupler analysis. By applying the open-circuited and shortcircuited boundary conditions along the PP' plane, the resulting half circuits are shown in Fig. 4 (b) and (c). The transmission (ABCD) matrix elements of the even-mode or odd-mode half circuit are given as where " 8,: is the input admittance seen toward the bisected host line under even-or odd-mode excitation, and they are given by: The coupler's frequency response can be reconstructed from the reflection coefficients Γ 8,: and transmission coefficients Τ 8,: of the half circuits as (13d) By using (8) and (13), and the conversion between ABCD and S network parameters, the following relations yield: C .
; ( In terms of the coupler's electrical parameters, the power division ratio K is given by From (15), the phase difference f1, and the relation between K and f1 can be obtained as From (16a), it is noted that the phase difference of this coupler can be tuned by controlling input impedances of the stub networks. For the coupler designed with equal power division ratios (Kf1=Kf2=K) and equal phase differences (ff1=ff2= f1) at dual frequencies f1 and f2 (f1 < f2), the electrical length q2 of these microstrip lines can be solved from (16b), where M is frequency ratio and the smallest value of q2 is chosen for implementation. Thus, the required impedance Z2 will then be determined by (16b) for the given f1, f2, K (=1), and f1. By using (11) and (14), the results in (18) yield: With (16a) and (18b), the following result can be derived: Again, to realize equal K and f1 in dual bands with (17), the microstrip length q1 in (19) is given by and the characteristic impedance Z1 can readily be obtained by (19). Finally, the input impedances of the stub networks at f1 and f2, i.e., Zs1(f1), Zs1(f2), Zs2(f1), and Zs2(f2), can be deduced from (18) and they are shown at the bottom of this page. By inspection of (21), it is seen that Zs1(f1)=-Zs2(f2) and Zs1(f2)=-Zs2(f1). Given two operational frequencies f1 and f2, power division ratio K=1, and the phase difference f1, the microstrip impedance Z1 or Z2 and electrical length q1 or q2 can be computed by (16b), (17), (19), and (20); moreover, the required Zs1 and Zs2 at each frequency can be computed by (21). Note that q1 and q2 are decided by f1 and f2 only. On the other hand, the calculated characteristic impedances Z1 and Z2 of the microstrip lines and the stub's input impedance Zs2(f1)=-Zs1(f2) and Zs2(f2)=-Zs1(f1), only Zs1(f1) and Zs1(f2) are shown here. It is seen that Z1 or Z2 shows a symmetric distribution about 90 o and reaches its maximum at f1=90 o .
The value of Z2 is larger than Z1. Besides, with increasing M, the required Z1 or Z2 increases while the variation of Zs1 against f1 decreases. Practical limitations of arbitrary dualband realization and the attainable phase tuning range based on the proposed coupler structure will be examined in Section III-C. To accommodate a wide range of stub's input impedances at two flexible frequencies, a highly tunable stub structure is proposed and detailed in Section III-B.

B. PROPOSED VARACTOR-LOADED STUB
As a means of enabling synchronously tuned phase differences at dual frequencies, a simple varactor-tuned microstrip stub with input impedance jZsi (i = 1, 2) is proposed in Fig. 6. In particular, the varactor capacitance C1 where Zin is the seen input impedance just to the left of the capacitive load C1, i.e., equal to the parallel effective impedance of the varactor capacitance C1 and the terminated stub (by C2). Fig. 7 shows the calculated Zsi as a function of C2 for C1= 0.37, 1, and 1.5 pF. When qs1 is small (< 10 o ) and C1/C2 are in the range of 0.3 pF to 0.8 pF, the Zsi (f2) is mainly controlled by C2. Thus, to carry out the required Zsi (f1) and Zsi (f2), the tuning capacitance C2 is essentially determined by Zsi (f2) and C1 is then additionally decided by Zsi (f1). This property will simplify the control of tuning varactors and will be exploited in the fabricated prototype. Besides, C1 can be used to enlarge the tunable range of Zsi (f1) to adapt to the required variation across phase difference f1, and thus the resulting capacitance tuning range (of C1) will be larger than that of C2, as observed in simulation and measurement.
For an arbitrary differential phase f1, five electrical parameters (Zs, qs1, qs2, C1, and C2) are used to realize Zsi (f1) and Zsi (f2). To maximize the tuning range, it is important to properly determine the non-tunable parameters Zs, qs1, and qs2 in accordance with the available varactor capacitances.
Since both of Zsi (f1) and Zsi (f2) increase with C1/C2 and thus Zs, qs1, and qs2 can be solved by using the minimum   capacitance Cmin in place of C1 and C2 to realize the minimum Zsi (f1) and Zsi (f2) in the attainable or required phase tuning range. As a result, larger Zsi (f1) and Zsi (f2), corresponding to other differential phases, can be easily fulfilled with available diode capacitances. Note that to allow parasitic effects, Cmin is chosen slightly larger than the minimum available diode capacitance provided by the manufacturer's data sheet. According to (22), Fig. 8 shows the calculated qs1, qs2, and Zs for Zsi (f1)=-84.99 W and Zsi (f2)=23.60 W. Here, only the value ranging from 0 o to 90 o is concerned for qs1 or qs2. Two important conclusions can be drawn. For a given Cmin, there may exist up to two solutions of (qs1, qs2) for a Zs. The optimal choice will depend on the potential for realizing a maximal tuning range and will be exemplified in Section III-C. Furthermore, it is noted that as Cmin increases, the required Zs increases. Thus, the impedance Zs can be used to accommodate the practical minimum capacitance influenced by the parasitic effects or manufacturing tolerance.

C. EXAMPLES AND DESIGN GUIDELINES
As illustrated in Fig. 5 (a), the ideal impedance Z1 or Z2 of the host microstrip lines varies with f1. In order to comply with this requirement, additional tuning varactors are generally necessary for host lines, thus substantially increasing structural and design complexity. If a smaller phase tuning range is allowed or further optimization is applied to the tunable stubs, four simple microstrip lines can be conveniently used given that a reasonable approximation to the ideal impedances is fulfilled. Because the impedance Z1 or Z2 is symmetric with respect to f1=90 o and a quadrature coupler, i.e., f1=90 o , is widely used in many components and networks, the values of Z1,f1=90 o and Z2,f1=90 o are chosen to realize Z1 and Z2, respectively. To examine the resulting impact and estimate the achievable tuning range, the reflection coefficient of a microstrip line of characteristic impedance Zf1=90 o and length qM with the reference impedance Zf1 was calculated in Fig. 9. Note that Zf1 represents the ideal impedance required for the phase difference f1 and it can be Z1 or Z2 in Fig. 5 (a). The length qM is a function of the frequency ratio M only and can be obtained by (17) or (20). Here, a normalized impedance P (i=1, 2), defined as the impedance ratio of the characteristic impedance Zi,f1 to Zi,f1=90 o , is derived based on (16b) and (19): where it is shown that " = % (= K ) and thus in Fig. 9 (b), the calculated results are plotted against K for a wide range of frequency ratio M. When f1=90 o , K =1 and this corresponds to perfect match in Fig. 9 (b) regardless of M (or qM). As f1 varies from 90 o , K becomes smaller and the degree of impedance mismatch increases, implying that a feasible tuning range can be preliminarily estimated according to the acceptable return loss of this study. For example, for M=2, the range for greater than 15-dB return loss is from K =0.81 to K =1, corresponding to the phase tuning range of 54 o~1 26 o , and the phase tuning range for 20-dB return loss is from about 63 o to 117 o for K ≥ 0.89. As M becomes larger, the attainable tuning range is improved as a result of a smaller qM. Nevertheless, the maximum feasible M will be limited by the characteristic impedance Z2, which increases to ~150 W for M=8.24 at f1=90 o and approaches to upper bound of the realizable line impedance based on standard PCB manufacturing process.
As mentioned in Section III-B, a proper choice of the stub's parameters Zs, qs1, and qs2 is crucial to the achievable phase tuning range. Based on the results in Fig. 8, Fig. 10 illustrates the calculated varactor capacitances C1 and C2 at pF to C2~0.56 pF, implying that precise capacitance control is requisite across different phase differences and will considerably increase implementation difficulty. On the other hand, Fig. 10 (b) shows the result when the value qs1 is comparable to qs2. In this case, a relatively irregular variation in either C1 or C2 between neighboring differential phases is obtained and may result in some impractical values (C2<0.3 pF at f1=70 o and 80 o given that 0.3 pF is the minimum available diode capacitance) and thus a reduced phase tunable range. With qs1=9.5 o and qs2=24.6 o in Fig. 10 (a), the phase differences can be synchronously tuned by a steady increase in tuning capacitances C1 and C2, thus contributing to an extended tuning range. The design procedure for the proposed dual-band coupler with synchronously tuned phase differences is summarized as follows.
Step 3) Use (22) to determine the parameters qs1, qs2, and Zs of the varactor-loaded stub with a given Cmin and the minimum Zs1 (f1) and Zs1 (f2) in the phase tuning range. The attainable phase tuning range can be estimated by the results in Fig. 9 (b). In practice, the value Zs is limited to the highest realizable line impedance and from the parametric study, it is seen that qs1=~10 o is a proper choice.
Step 4) With the chosen qs1, qs2, and Zs from Step 3), the required varactor capacitances C1 and C2 for each differential phase f1 can be calculated by the results from Step 2) and (22).
Step 5) Choose commercially available tuning varactors to cover the required capacitances across the tuning range.

D. MEASURED AND SIMULATED S-PARAMETERS OF THE FABRICATED COUPLER
A hybrid coupler with operating frequencies at 0.9 GHz and 1.8 GHz, i.e., M=2, was experimentally developed to perform synchronously tuned phase differences. Fig. 11 shows the physical layout and photograph. The physical dimensions of the fabricated coupler are given in Table 1. The fabricated coupler was built on a RT/Duroid 5880 substrate of thickness 0.508 mm. Loaded with the tuning varactors C1 and C2, two microstrip stubs on the left side are tuned by control voltages V1 and V2, respectively, to realize tunable input impedance jZs1; the voltages V3 and V4 are used to control varactors C3 and C4, respectively, in right-sided microstrip stubs of tunable input impedance jZs2. The DC block capacitance Cdc of 20 pF and the bias resistance Rrf of 100 KW were used. In addition, four DC blocks were connected to each port to prevent the flow of DC frequencies to RF signals and prevent damage to the network analyzer. The surface-mount GaAs tuning varactors MA46H201 (0.3 pF−2.3 pF) were used to implement C1 and C3, while the varactors MA46H070 (0.3 pF − 1.3 pF) were used to implement C2 and C4. Note that the relatively small capacitance tuning range of the varactor diode MA46H070 is used to accommodate a small variation of C2 and C4 as predicted in Section III-C for Zs=140 W, qs1 =9.5 o , and qs2 =24.6 o . Figs. 12-14 show measured and simulated Sparameters of the fabricated prototype with dual-frequency phase differences of 60 o , 90 o , and 120 o , respectively. As seen, measured results agree well with the simulated counterparts, and dual-band equal power division and equal/synchronous phase control are experimentally validated. The bias voltages V1 − V4 for each tuning state are given and it is seen that the voltages V1/V2 and V3/V4 monotonically increase and decrease, respectively, with increasing f1, as expected from the result in Fig. 10 (a). In the phase tuning range from 60 o  V1=2.82 V, V2=3.2 V, V3=20 V, and V4=20 V. The  simulated junction capacitances C1=1.414 pF, C2=0.649 pF, C3=0.547 pF,  and C4=0.324 pF. At 0.9/1.8 Fig. 15, the differential phases at dual frequencies can be synchronously and continuously tuned from 60 o to 120 o and meanwhile, they can be well controlled with phase error less than 2.5 o across the entire tuning range. Note that the achieved phase tuning range (60 o -120 o ) agrees very well with the prediction result in Fig. 9 (b) for M=2 and 15-dB return loss. Table 2 summarizes measured bandwidths of several key performances. Note that the phase bandwidth is calculated as the fractional bandwidth with less than 5 o phase deviation  from each measured/desired f1. It is shown that the obtained bandwidth at the first frequency is larger than the corresponding bandwidth at the second frequency, as a result of the larger wavelength in the first band. Moreover, the minimum bandwidth comes from either end of the tuning range, i.e., 60 o or 120 o . In other words, the asymmetry of coupler structure to realize a nonstandard differential phase (other than 90 o ) will lead to bandwidth reduction. The linearity of proposed tunable coupler was also studied and for the entire phase tuning range, the measured IIP3 is from 23.6 dBm to 36.3 dBm at 0.9 GHz and is from 23.7 dBm to 32.9 dBm at 1.8 GHz.

IV. MEASURED AND SIMULATED RESULTS OF THE DUAL-BAND FEEDING NETWORK AND THE BEAM-SCANNING/SWITCHING RADIATION PATTERNS
In Section III, the measured phase tuning range of the fabricated coupler is from 60 o to 120 o for f1 (=∠S41-∠S31) at 0.9/1.8 GHz and it can also be experimentally shown that the corresponding f2 (=∠S32-∠S42) ranges from 120 o to 60 o . Thus, based on the previous discussions in Section II-A, the continuous beam scanning in the switchable forward or backward direction can be carried out with the proposed dual-band feeding network. Figs. 16 and 17 show the photograph and measured/simulated S-parameters of the fabricated dual-band beam-scanning/switching network, respectively. This network was again developed on a RT/Duroid 5880 substrate of thickness 0.508 mm. Between neighboring antenna ports, the measured phase differences of 60 o , 90 o , and 120 o for port 1 as the input are illustrated in Fig. 17 (a). In Fig. 17 (b), the measured return loss |S11| is greater than 20 dB at 0.9/1.8 GHz and the measured insertion loss |S31|, |S41|, |S51|, or |S61| is about 8.2 dB on average. The amplitude imbalance between output ports is less than 1.5 dB. Note that the Analog Devices HMC284A SPDT switch was used in the dual-band feeding network and at these frequencies, the measured insertion loss is about 1 dB. In Fig.  17 (c), the measured isolation between antenna ports is greater than 15 dB for all tuning states. Based on this beamscanning/switching network, a 1 × 4 antenna array was further developed for integration in the proposed dual-band beam-scanning/switching module. Fig. 18 shows the measurement setup for far-field radiation patterns in the xy (horizontal)-plane, i.e., the plane with this antenna array. The array is connected to the network by cables and its radiation patterns were measured in the anechoic chamber. Fig. 19 shows the measured beam-scanning radiation patterns when the coupler's phase difference was tuned to f1=60 o , 90 o , and 120 o and the input was switched between ports 1 and 2 of the module. The fabricated 1×4 array has measured return loss greater than 15 dB at 0.9/1.8 GHz and the inter-element spacing is 130 mm, corresponding to 0.39l0 at 0.9 GHz (or 0.78l0 at 1.8 GHz). When the module is fed through port 1, it is seen that the main beam is steered from 336 o (-24 o ) to 318 o (-42 o ) with the measured peak gains of 1.3-3.2 dBi at 0.9 GHz. At the same frequency, when feeding the module through port 2, the main beam is steered from 24 o to 34 o with the measured peak gains of 1-3.3 dBi. Moreover, at 1.8 GHz, the main beam is steered from 348 o (-12 o ) to 336 o (-24 o ) with measured gains of 5.8-6.8 dBi and from 12 o to 22 o with measured gains of 4.7-6.6 dBi for input excitation at port 1 and port 2, respectively. Thus, continuous beam scanning in the forward or backward direction is validated by experimental results obtained at 0.9 GHz and 1.8 GHz, and the slight asymmetry in beam-scanning angles in two opposite directions is mainly attributed to the measurement error and the output amplitude/phase imbalance of the fabricated dual-band beam-scanning/switching network.
In Table 3, careful comparisons are made between this fabricated prototype and the prior art in terms of the beamscanning/switching technique, the number of applicable operating band(s), the experimentally achieved phase progression Dj between antenna ports, beam-scanning range, and the array gain. By exploiting the proposed dual-band coupler with synchronously tuned phase differences, it is seen that the proposed beam-scanning/switching network features the remarkable advantage of flexible dual-frequency applications (and beam scanning at two fixed frequencies) whereas the prior work is mainly developed with singlefrequency scheme. Moreover, in spite of its simple configuration, i.e., consisting of merely passive tunable couplers and phase shifters, this proposed beamscanning/switching network provides a cost-effective solution to continuous beam-scanning, the feeding network based, and several-GHz applications without extra complex and active control circuitry [3]. Also note that the practically attainable beam-scanning range is dependent on the antenna's element pattern. Especially, at 0.9 GHz the beamscanning range is reduced because the main beam resulted from Dj=120 o is shifted towards boresight by the directional element pattern.

V. CONCLUSION
A novel and simple dual-band beam-scanning/switching network is proposed to contribute to dual-band beam scanning in both of the forward and backward directions. To support the beam-scanning/switching capabilities at two flexible frequencies, for the first time, a varactor-tuned microstrip coupler with continuously and synchronously tunable phase differences is proposed for this network realization. As a proof-of-concept, a 0.9/1.8-GHz beamscanning/switching network and the array module were prototyped and achieved the measured beam-scanning ranges of -   Differential phase between antenna ports. 2 Beam-scanning range. 3 Port 1 or port 2 as the input. 4 Butler matrix. 5 Predicted results based on the measured singlechannel S-parameters.