Optimization of the elementary cell of a broadband reflectarray for spatial application Etienne GIRARD1, Afshin ZIAEI2, Raphaël GILLARD1, Michel CHARRIER2, Hervé LEGAY3 et Béatrice PINTE3 1 IETR/INSA de Rennes – 20 avenue des Buttes de Cœsmes - 35043 Rennes Cedex – France. {etienne.girard, raphael.gillard}@insa-rennes.fr 2 Thales Research and Technology France, Domaine de Corbeville, 91404 Orsay Cedex, France. {afshin.ziaei, michel.charrier}@thalesgroup.com 3 Alcatel Space Industrie, 26 avenue JF Champollion, BP 1187, 31037 Cedex, Toulouse, France. {herve.legay,beatrice.pinte}@space.alcatel.fr Abstract. This paper describes the optimization of a reflectarray unit cell using the Finite Differences in Time Domain ( FDTD ). The studied reflectarray consists of printed radiating elements embedded in metallic cavities illuminated by a circularly polarized wave. The re-radiated wave must have a similar polarization and a given angle of redirection. In the first part of this work, the studied structure is a passive one and two ways of optimization were successively used, with or without the insertion of an air gap. In the second part, the analysis focuses on the presentation of an active cell that permits to choose the angle of the re-radiated wave thanks to the use of MEMS commutators. 1. Introduction. The RNRT project ARRESAT ( REflectarray Antenna for SATellite ) aims at using a reflectarray as a spatial antenna. Reflectarray antennas with scanning ability are low cost compared to mechanically steerable antennas and exhibit low losses compared to direct radiating arrays. This paper describes the analysis and optimization of the elementary cell of a reflectarray using the Finite Differences in Time Domain method (FDTD). The studied reflectarray has to re-radiate an incoming circularly polarized wave in a chosen direction without inverting its polarization. The unit cell of such a reflectarray must therefore re-radiate an identical circularly polarized wave as the incoming wave while adding a phase shift depending on the position of the radiator in the cell [1,2]. A rigorous electromagnetic analysis is needed to study such a structure. FDTD was chosen as it can deal with complex 3D structures and provides a large bandwidth characterization in a single simulation. The present paper will first analyze a passive structure operating in C band. The influence of geometrical parameters will be studied. In a second part, an active cell operating in Ku band will be shown and the phase shift will be achieved by using MEMS switches developed for this work. Details on the technological process are given in paragraph 4. 2. Operating mode of the structure. The studied passive cell is displayed in figure 1. The radiating element is embedded in a metallic cavity in order to reduce mutual coupling between adjacent cells. In the final version, the metallic cavity will be covered with a dielectric sheet to minimize the importance of incidence angle [3]. For a circularly polarized incoming wave, two degenerated modes will be induced in the cavity. The radiating element has to re-radiate those two modes with a phase difference of 180°. To do so dipoles above a ground plane are used. A dipole reflects the mode whose electric field is parallel to it (along Y in the following) while the ground plane reflects the orthogonal mode. The distance between dipole and ground (e+h) is used to control the phase shift between the two reflected modes. The ability to control the phase of the resultant re-radiated circularly polarized wave is due to the combination of several dipoles (6 in the present work) with different rotation angles [1,2]. For a given phase, only one of the six dipoles is active (i.e. its two arms are connected together through the central disc, while the others are not connected to this central disc). MEMS switches are used to select one dipole among six as explained in paragraph 4.
3. Optimization parameters. The influences of geometrical parameters used to optimize the structure are studied in this paragraph. This study uses a passive cell embedded in a square metallic cavity operating in C band. A switch “on” is simulated by a perfect electrical connection (like the dipole along Oy in figure 1) whereas a switch “off” is simulated as a gap of 0.325 mm (other dipoles). The optimization process aims at a phase shift of 180° between the reflection coefficient Γyy (excitation of the TE10 guided mode, parallel to the active dipole) and Γxx (excitation of the TE01 guided mode, perpendicular to the active dipole). All simulations present cross reflection coefficients Γxy (reflected mode in TE01 for an excitation mode in TE10) and Γyx (reflected mode in TE10 for an excitation mode in TE01) less than -80dB due to the structure’s symmetry. More details on the numerical process can be found in [4]. The main parameters that are used to optimize the structure are the dipole’s length L, the external width w1 (this parameter controls the flare of the dipole), the substrate thickness e and its distance from the short-circuit plane h. The diameter of the central disc (d =2.94 mm) and internal width of the dipole (w2=0.325 mm) are fixed so as to be able to put twelve switches around the metallic central disc. The dielectric substrate has a 2.94 permittivity. A first structure (e=7.1mm,L=16.93mm) has been studied and results are presented in figure 2. In this simulation, the dielectric lay on the bottom of the cavity (h=0mm). As can be seen, the dipole’s flare (selected by the external dipole’s width w1) is an interesting tuning parameter that can lower the phase variation over the frequency range [5.45-5.95] GHz and leads to the desired broadband behaviour. Nevertheless, such a thick substrate (e=7.1mm) is irrelevant due to cost and loss. Moreover, a first round of measurement using this thickness showed a great dependency to the dielectric’s quality (anisotropy seems to be a key point). A solution to this problem was to use thin substrate of normalized thickness (e=1.524mm,εr=2.5). In this case the dielectric must be hung in the cavity in order to have a sufficient height between the dipole and the short-circuit plane (h>0) so as to reach a phase shift of 180°. For a given length of L=18.54mm, the height must be h=6.858 mm (figure 3). Figure 4 displays the two optimal lengths (16.59 mm or 21.14 mm) that are suitable for a height of 6.858mm. After the optimization process, figure 5 shows the phase over the whole frequency range. The best case, that is to say the more stable, is for the shortest dipole (h=6.858 mm, L=16.59 mm, w1=2.275 mm). 4. Operating mode of the active structure. The active cell was designed with MEMS switches that enable to choose one active dipole among six. The very orientation of the selected dipole determines the additional phase of the reflected wave and therefore permits to steer the re-radiated field (assuming several cells are used with a linear phase weighting) [4]. The FDTD method takes into account switches and polarization lines (figure 6) by using lumped elements. The optimization process uses a cylindrical cavity, this structure is much more representative of the practical measurement device (an hexagonal cavity). All six dipoles and MEMS switches were built monolithically on a glass substrate. The central disc diameter (d=0.625mm) and the internal width of the dipoles (w2=0.150mm) are made so that twelve MEMS switches can be put around the central disc. As this structure operates at a much higher frequency range than the passive one, the PTFE substrate thickness (e1=1.270 mm, εr1= 2.94) made it possible to use a structure put in the bottom of the cavity (h=0mm). The optimization process uses therefore the length of the dipole and its wide to tune the studied structure. The first MEMS developed at the THALES Research and Technology Center were “membrane” switch type. This choice was made to have a suspended structure so as to lower the electrostatic sticking risks. Compared to a “cantilever” switch type, this “membrane” type is more rigid, that means shorter commutation time but as a drawback, higher actuation voltage. Glass was chosen for the substrate as it is of a moderate price and high resistivity. MEMS switches are made in six micro-lithography and deposit steps. After realizing the membrane on an expendable slab (sacrificial layer), the last step is to get rid of this slab to release the membrane. Between two operating modes, a wet one and a dry one, we chose to use O2 plasma that doesn’t require to be dried by CO2. The wet etching mode leads to a problem, that is the released “membrane” sticks to the substrate. One of the advantages of the RIE 02 tank etching is to have a cooling substrate device, but the etching is anisotropic. That is why the “membrane” must be drilled regularly. These holes are made every 8 µm and are of 4 µm diameter. MEMS switches were made in aluminium. This material is a good compromise as it has a good electrical conductivity and provides a good rigidity of the final structure. Results of the first round of tests on MEMS switches (embedded figure 7) in “parallel” configuration show an isolation of –18dB and insertion loss of –0.1dB at 20 GHz.
Figure 1 : Studied structure Γyy - Γxx
Γyy - Γxx
200
Phase ( ° )
Phase ( °)
210
190 180 170 160 5,45
5,53
5,61
5,69
5,78
5,86
5,94
h = 6.858 mm Phase = 162°
350 300 250 200 150 100 50 0 0
0,762 1,857 3,81 5,334 6,858 8,382 13,72 22,1 28,96
Frequency ( GHz ) w1 = 1,62
w1 = 2,28
h ( mm ) w1 = 2,92
Figure 2 : Large bandwith operation by using the dipole’s width as a tuning parameter.
Figure 3 : Use of the air slab thickness to optimize a structure with a given length.
a= 34.8 mm, L = 16.93 mm, d = 2.94 mm, w2 = 0.325 mm, e = 7.1 mm, h = 0 mm,εr=2.94
a= 34.8 mm, L = 18.54 mm, d = 2.94 mm, w2 = 0.325 mm, w1 =2.275 mm, e = 1.524 mm,εr=2.5
Γyy - Γxx
Γyy - Γxx
350 Phase ( °)
330 310 Phase ( °)
290 270
L=16.59 mm
250
L=21.14 mm
230 210 190
190 185 180 175 170 165 160 5,45
5,53
5,61
5,69
5,78
5,86
5,94
Frequency ( Ghz )
180°
170
Length 16,59 mm
Length 21,14 mm
150 11,39 15,29 16,59 17,89 18,54 19,84 21,14 22,44 23,09
Dipole length
Figure 4 : Use of the dipole length to optimize a structure with a given air slab thickness.
Figure 5 : Optimized structure, Broadband : L = 16.59 mm, narrower Band : L = 21.14 mm.
a= 34.8 mm, h=6.858 mm, d = 2.94 mm, w2 = 0.325 mm, w1 = 2.275 mm, e = 1.524 mm,εr=2.5
a= 34.8 mm, h=6.858 mm, d = 2.94 mm, w2 = 0.325 mm, w1 = 2.275 mm, e = 1.524 mm,εr=2.5
Figure 6: Simplified Active cell
Figure 7 : MEMS embedded in the structure.
Γyy - Γxx
Axial Ratio (dB)
Phase ( °)
195
2
190
1,5
185
1 180
0,5
175 170 17,8
18,0
18,2
18,4
18,6
18,7
Frequency ( GHz )
18,9
19,1
0 17,80
17,99
18,18
18,37
18,56
18,75
18,94
19,13
Frequency ( GHz )
( a ) Phase shift ( b ) Axial Ratio Figure 8 : Optimal case for an active cell. The technology used here was validated first on parallel switches. After this a more complex system that uses several serial switches was built. The elementary cell is displayed in figure 7. The optimal case for this active cell is displayed in figure 8. Phase variation of this cell (figure 9a) is stronger than the passive case. The insertion of MEMS switches and polarization lines disturbs the operation of the structure. Nevertheless, the axial ratio (figure 9b) is less than 2dB on the whole frequency range [17.8-19.3] GHz. 5. Conclusion. The optimization process of a passive cell of a reflectarray operating in C band has been presented. Those first simulations were used to tune an active cell operating in Ku band. This active cell reflects a wave in a chosen direction by using MEMS switches developed for this spatial application. Measurements and performance analysis of an active antenna will be presented during the symposium. 6. References. [1] J. Huang, R.J. Pogorzelski, « A Ka-band microstrip reflectarray with elements having variable rotation angle », IEEE transactions on antennas and propagation, Vol. 46, N° 5, May 1998 [2] R.D. Janor, X-D Wu, K. Chang, « Design and performance of a microstrip reflectarry antenna », IEEE transactions on antennas and propagation, Vol. 43, N° 9, September 1995. [3] G.H.Knittel, « Wide-angle impedance matching of phased array antennas, a survey of theory and practice », in Proceedings of the 1970 phased array antenna symposium, p157-171. [4] E.Girard, R.Moulinet, R.Gillard, « An FDTD Optimization of a Circularly Polarized Reflectarray Unit Cell », IEEE Antennas and Propagation. Soc. Int. Symposium, San Antonio, TX, vol. 3, 136-139, June, 2002. [5] A.Taflove, Computationnal Electrodynamics, The FDTD Method, ch. 6.5, p. 111, Artech House.
