Equivalent-circuit models for

frequency-selective surfaces at oblique

angles of incidence
C.K. Lee. B.Sc, and R.J. Langley, Ph.D.. C.Eng.. M.I.E.E.
Indexing terms: Antennas, Microwave circuits and networks

Abstract: The equivalent-circuit-modelling technique is useful for designing printed frequency-selective surfaces.
A correction term is derived which enables transmission responses to be modelled at oblique angles of incidence.
Models derived for arrays of gridded-square and double-square elements are compared with experimental
results for plane waves at normal incidence and at oblique angles for TE- and TM-incidence. Good agreement is
obtained for angles of incidence up to 45° in the principle planes. The importance of using the immittance
correction terms in the models is also discussed.

1 Introduction P d,
h •!" ••

•
Frequency-selective surfaces (FSS) are used in reflector i i i

HI®
antennas to provide dual or multiple frequency band oper-
ation. They can take the form of curved secondary mirrors —w
[1], flat diplexers [2] or a combination of the two and 1
usually consist of arrays of printed-circuit elements such as
Jerusalem crosses [3], concentric rings [4] or double
squares [5] supported by a dielectric substrate. The fre-
quency response of the arrays is required to be indepen-
dent of the angle of incidence, and they must introduce
•
minimal crosspolarisation. Several methods of analysis Fig. 1 Geometry of arrays
have been developed. The modal expansion method [6, 7] a gridded-square elements
provides a comprehensive analysis capable of giving details b double-square elements
of the frequency and polarisation responses, together with
a physical understanding of their operation. However, the high-frequency-reflection-band characteristic, whereas the
method often requires lengthy computing time which can double-square-array bands are reversed.
become prohibitive for more complex elements, such as The experimental results presented in the paper were
double squares, where large numbers of current modes are made on arrays 20 cm square, with 0.012 mm thick copper
required. The spectral-domain method [8, 9] also provides printed on a 0.027 mm polyester substrate to minimise the
a full wave analysis. As an alternative technique, effect of the dielectric. Correction for the substrate is made
equivalent-circuit modelling [5, 10, 11] allows rapid com- in the models by increasing the susceptance terms only by
putation of the frequency response for design work, an effective dielectric constant er = 1.12.
although at present crosspolarisation levels cannot be pre-
dicted. 2 Equivalent-circuit models
The basic circuit-modelling equations, originally pre-
sented by Marcuvitz [12], can only be applied at oblique The equivalent-circuit modelling technique was first
angles of incidence to an inductive component illuminated applied to frequency-selective circuits by Anderson [10].
by a TE-incident wave or a capacitive component illumi- In the models the interaction of incident waves with an
nated by a TM-incident wave. Modified equations are infinite periodic array of elements is represented as a wave
derived here for the other two cases, that is inductive TM- travelling down a transmission line, with shunt lumped
incidence and capacitive TE-incidence. Consequently, the circuit impedances representing the array. For an infinite
models can now be used to predict the transmission char- array of thin, continuous, perfectly conducting narrow
acteristics of FSS at oblique angles of incidence in all strips the shunt impedance is either inductive or capacitive,
planes. In this paper we discuss the modelling technique, depending on whether the incident wave is polarised paral-
the equivalent circuits obtained for gridded-square and lel or perpendicular to the strips.
double-square arrays, see Fig. 1, and compare the plane- In many practical arrays both inductive and capacitive
wave transmission responses calculated using the models components are present together, and their immittances
with experimental results at 0° and 45° incidence in the E- vary at oblique angles according to whether TE- or TM-
and H-planes. The correction term in these equations is of incidence, or a combination of the two, occurs. Referring
significance in practical FSS designs and is also discussed. to Fig. 2, for the inductive grating, TM-incidence occurs
Measured and predicted band centre frequencies are com- when the E-field is polarised parallel to the plane of inci-
pared for gridded-square arrays at normal incidence. Both dence, i.e. 6 = 0°, and TE-incidence when the E-field is per-
elements have transmission and reflection bands which are pendicular to the plane of incidence, i.e. 4> = 0°. Modelling
relatively insensitive to the angle of wave incidence. The arrays at oblique angles of incidence therefore requires
gridded array has a low-frequency-transmission-band, expressions for the immittances for both TE- and TM-
incidence. Using the geometry of Fig. 2, for conductors of
Paper 4086H (Ell), first received 24th January and in revised form 7th May 1985
periodicity p, width w and spaced a distance g apart, Mar-
The authors are with the Electronics Laboratory, University of Kent, Canterbury, cuvitz gave normalised immittance expressions for two
Kent CT2 7NT, United Kingdom cases:
395
IEE PROCEEDINGS, Vol. 132, Pt. H, No. 6, OCTOBER 1985
 Inductance, TE-incidence:
(6)
XTE = F(p, w, X) t • bn \
For TE-incidence
p cos 0 f fnw•J + G(p, w, X, 0)] (1)
In cosec — . p cos 6
t=J (7)
L \2p
metallic
strips
in
(8)

where the propagation constant yn is given by

yn = — j [ sin 9 -\ I — 1, therefore y0 = cos

VV PJ

From eqns. 7 and

p sin8 0 x2
-2

S-t =
and from eqn. 6

C TEn
_
P sin 0 V _ E ! ln'

therefore
Fig. 2 Plane wave incident on an inductive strip grating
For a capacitive strip grating exchange the incident electric field E for a magnetic
field H - 1 (9)
^ + i -?,
where X is the wavelength, and the correction term G(p, w,
X, 4>) is derived below. From Babinet's duality conditions
the susceptance for TM-incidence is given by This is identical to the factor given by Marcuvitz [12].
For TM-incidence
BTM = 4F(p, g, X)
p sec (10)
Ap cos t=J
In cosec I 2— ) + G{p, g, X, 4>) (2)
\ Pj J (11)
Eqns. 1 and 2 are valid for w <^ p, g <^ p and p < X. Expres-
sions for the two missing cases, i.e. TM-incidence for and
inductance XTM and TE-incidence for capacitance BTE,
n2X1 2
must be derived to complete the modelling equations. In yn = —j sin2 4> -\ 2 — 1' therefore y = cos
general, these immittances take the form [13]
V P
p sec
•A- T\4
TM — In cosec (j-\ + G(p, w, X, 0) (3) From eqns. 10 and 11
P s i n 4>Y . ..2
n22 - P*
and S-t = X J ' " X2
Ap sec 0 ng . and again substituting into eqn. 6 gives
BTE = cosec ( ^ ) + G(p, g, K 0)1 (4)
1 1
Archer, in Reference 14, has derived general expressions p sin 0 2 p^_
for the normalised wave reactances of inductive strip grat- n
ings. From these results we can obtain first-order correc- ~ x2
tion terms for both TE- and TM-wave incidences. The therefore
correction term is given as
1
C™ = -1 (12)
G(p, w, X, I) = p2 cos <{>
1 -
0.5(1 -
Incorporating Cf^f and C [ £ into eqn. 5 allows both induc-
tive and capacitive components of FSS array elements to
be calculated in all planes of incidence. 0 and <f> must be
interchanged in eqns. 9 and 12 for the capacitive cases (see
(5) Fig. 2).
The significance of this correction term G(p, w, X, £) in
where /? = sin nw/2p the calculation of practical immittances is discussed in the
In eqn. 5, the first-order coefficients C 1 ± are calculated next Section.
from The normalised admittance Y of an array at oblique
396 IEE PROCEEDINGS, Vol. 132, Pt. H, No. 6, OCTOBER 1985
incidence can now be calculated, and the transmission transmission/reflection band ratios of > 2.5:1 are
coefficient T determined from obtained. The presence of a grid modifies the low-
frequency transmission response of the simple squares,
Tl2 =
giving a more closely spaced frequency-band ratio of
4 + \Y\' between 1.3 and 2:1. Fig. 3a, which refers to array 3 in
Table 1, shows a typical gridded-square frequency
3 Gridded-square arrays response under plane-wave illumination at normal inci-
dence. The transmission band is centred at 15.3 GHz and
The transmission characteristic of an array of single the reflection band at 23.1 GHz, giving a band-spacing
squares [15] essentially consists of a single reflection band ratio of 1.51. These band centres are relatively insensitive
and a lower-frequency transmission band. Typically, to the angle of incidence, see Figs. 3b and 3c, moving by
about 0.7 GHz (7%) from 0° to 45° incidence. The band-
widths, common to all angles of incidence up to 45°, mea-
sured where the loss is 0.5 dB or less, are approximately
12% in both transmission and reflection. For a given
bandspacing, these bandwidths may be traded against one
another. For example, arrays 1 and 4 in Table 1 have the
same band ratio fJfT = 1.47, but transmission and reflec-
tion bandwidths of 16.4% and 7.1%, and 9.0% and 13.1%,
respectively. In general, as the band-spacing ratio fjjfr is
increased, the drift in the band centres, as the incident
angle varies, also increases, particularly in reflection, and
consequently the common 45° bandwidth may decrease to
about 6% for fjJfT = 2, see array 7 in Table 1.
The equivalent-circuit model for gridded squares is
shown in Fig. 4a and comprises an inductance L2 rep-
resenting the grid, shunting a series resonant circuit Lt and

Li Li

4= c.

Fig. 4 Equivalent-circuit models

10
a gridded-square array
b double-square array

20 Cx representing the squares. These components are given

by
30 X2 = coL2 = F(p, wl5 X)
10 20 30
frequency ( GHz Xx = coLi = 2{X2\\XZ) where X3 = F(p, 2w2, A) • -
P
Fig. 3 Plane-wave transmission response of gridded-square array 3
a normal incidence
B, =coCx =2erF(p,g,X)-
b 45° TE-incidence (tf> = 0°, 6 = 45°) P
c 45° TM-incidence W> = 45°, 6 = 0°) The frequency-band centres calculated using this model
measured
model with correction term are compared with experimental results in Table 1 for
x x x model without correction term normal incidence (see also Fig. 3a). Good agreement is

Table 1 : Measured and calculated band centres for gridded square arrays at normal incidence
Array Dimensions, mm Transmission Reflection Band ratio
number
P w
\ d w2 9 Band centre fT, GHz Bandwidth Band centre fR. GHz Bandwidth fR/fT
Measured Model % Measured Model %
1 4.0 0.38 2.5 0.2 0.56 26.8 27 16.4 39.5 39.1 7.1 1.47
2 4.5 0.33 3.47 0.17 0.35 17.2 17.1 8.2 24.3 24.7 13.2 1.42
3 5.05 0.15 3.7 0.15 0.6 15.3 15.3 13.1 23.1 23.1 12.6 1.51
4 5.02 0.15 4.09 0.15 0.39 13.3 13.0 9.0 19.6 19.4 13.1 1.47
5 5.02 0.09 4.19 0.2 0.37 12.5 12.1 8.3 19.4 19.0 14.4 1.55
6 5.48 0.27 4.41 0.7 0.4 14.0 14.1 14.3 25.9 24.1 13.9 •1.85
7 6.0 0.2 4.6 0.8 0.6 13.0 13.6 17.7 26.7 24.4 6.0 2.06
Note: Bandwidths are common to all angles up to 45° for TE- and TM-incidence

IEE PROCEEDINGS, Vol. 132, Pt. H, No. 6, OCTOBER 1985 397

obtained (within 5%) for band spacings, fn/fT, up to ~1.6. general, the band centres are least sensitive to changes in
However, for wider separations, where the widths of con- the angle of incidence for spacing ratios of < 1.6 :1. Band-
ductors and/or gaps between them are significant when widths of about 10% are obtained for 45° incidence in
compared with the periodicity, the model is, as expected
[12], somewhat less accurate. This is illustrated for array 7
where the conductor width w2 is 0.8 mm. The transmission
band is accurately predicted but the reflection band is in
error by ~9%. Close agreement between the modelled
and measured frequency response is also obtained at
oblique angles of incidence; for example, see Figs. 3b and
3c where the results for array 3 are compared for TE- and
TM-incidence at 45°.
The importance of the correction term G(p, w, X, #) (see
eqn. 5) in calculating the immittances is clearly shown in
Figs. 3a, b and c where the modelled response is also calcu-
lated without it for comparison. These terms are seen to be
important at higher frequencies, and significant differences
become apparent above 20 GHz where p > 0.25A. For
example, in Fig. 3 the reflection resonance is accurately
predicted using the model with correction terms, but
without these terms the prediction is — 1 GHz in error.

4 Double-square arrays
The equivalent circuit for double-square arrays is shown in
Fig. 4b and was first discussed, for normal incidence only,
in Reference 5. This array is essentially a double-resonant
structure, similar to the Jerusalem cross and concentric-
ring arrays, providing a stable low-frequency reflection
band coupled with a closely spaced high-frequency trans-
mission band, in contrast to the gridded-square array.
Somewhat wider frequency-band separations may be
accommodated by a double-square array, up to approx-
imately 2.2 :1. Fig. 5 shows the transmission characteristic
at normal incidence and TE- and TM-incidence at 45°.
This array has a band spacing fT/fR of 1.46. Common
bandwidths for 45° incidence are 8% in transmission and 20 30 40
14% in reflection, and again a bandwidth trade off frequency ,GHz

between the two is possible. These figures are essentially c

similar to those obtained for gridded-square elements with Fig. 5 Plane-wave transmission response of double-square array
a similar fjJfT ratio. The bandwidths are limited, as are p = 5.0 mm, wl = w2 = 0.15 mm, dl = 4.5 mm, d2 = 2.9 mm
those for the gridded squares, by the TM-incidence a normal incidence
b 45° TE-incidence ($ = 0°, 6 = 45°)
response in reflection and TE-incidence response in trans- c 45° TM-incidence (0 = 45°, 0 = 0°)
mission, see Figs. 5b and 5c. A major factor which deter- measured
model with correction term
mines the transmission bandwidth is the sensitivity of the x x x model without correction term
upper resonance to changes in the angle of incidence.
For comparison, the transmission characteristic calcu- transmission and reflection for band centre separations of
lated from the model, both with and without the correc- ~1.5. An important feature of the gridded-square array is
tion term G(p, w, X, £), is also plotted in Fig. 5. The its high-frequency reflection-band, low-frequency
corrected model is again in very good agreement with the transmission-band characteristic, which allows diffraction
experimental results, not only at normal incidence but also and blockage to be minimised at the FSS. The Babinet
at 45° incidence, up to the location of the second reson- complement of the double-square array has, of course, a
ance. The complex response near the upper resonance for similar frequency response.
45° TE- and TM-incidence is not predicted, nor is the The usefulness of the equivalent-circuit-modelling tech-
slight upward shift in the lower resonance (0.7 GHz). The nique for design work has been extended by the addition
corresponding results obtained for the model without the of the correction term. The inductance and capacitance of
correction term are again accurate at lower frequencies a strip grating can now be modelled for both TE- and TM-
< 12 GHz, but significantly in error near the upper reson- illumination. Consequently, the tranmission characteristic
ance at approximately 31 GHz. of FSS arrays in all planes can be calculated using the
model. It is also important to include the correction terms
5 Conclusions in the formulation to ensure the accuracy of these models
at high frequencies.
Gridded and double-square elements are useful additions
to the range of printed FSS arrays available to designers. 6 Acknowledgment
Both provide transmission and reflection bands which are
for the most part insensitive to the angle of incidence and This work was supported by a grant from the UK Science
can be modelled using equivalent-circuit techniques. In and Engineering Research Council.
398 IEE PROCEEDINGS, Vol. 132, Pt. H, No. 6, OCTOBER 1985
7 References 8 TSAO, C.H., and MITTRA, R.: 'Spectral-domain analysis of fre-
quency selective surfaces comprised of periodic arrays of cross dipoles
1 SCHENNUM, G.H.: 'Frequency-selective surfaces for multiple- and Jerusalem crosses', IEEE Trans., 1984, AP-32, pp. 478-486
frequency antennas', Microwave J., 1973,16, p. 55 9 OHTA, H.H., LANG, K.C., TSAO, C.H., and MITTRA, R.: 'Fre-
2 CAHILL, R., and PARKER, E.A.: 'Concentric ring and Jerusalem quency selective surface for satellite communications antenna applica-
cross arrays as frequency-selective surfaces for a 45° incidence diplex- tions'. AP-S International Symposium Digest, 1982, pp. 475-478
er\ Electron. Lett., 1982,18, pp. 313-314 10 ANDERSON, I.: 'On the theory of self-resonant grids', Bell Syst.
3 ARNAUD, J.A., and PELOW, F.A.: 'Resonant-grid quasi-optical Tech. J., 1975, 54, pp. 1725-1731
diplexers', Bell Syst. Tech. J., 1975, 54, pp. 263-283 11 OHTA, H.H.: 'Circuit representation of tripole element array for fre-
4 PARKER, E.A., HAMDY, S.M.A., and LANGLEY, R.J.: 'Arrays of quency selective surfaces'. AP-S International Symposium Digest,
concentric rings as frequency-selective surfaces', Electron. Lett., 1981, 1982, pp. 455-458
17, pp. 880-881 12 MARCUVITZ, N.: 'Waveguide handbook' (McGraw-Hill, 1951)
5 LANGLEY, R.J., and PARKER, E.A.: 'Double square frequency- 13 WAIT, J.R.: 'Reflection at arbitrary incidence from a parallel wire
selective surfaces and their equivalent circuit', ibid., 1983, 19, pp. grid', Applied Sci. Res., 1954, 4, pp. 393-400
675-677 14 ARCHER, M.J.: 'Wave reactance of thin planar strip gratings', Int. J.
6 MONTGOMERY, J.P.: 'Scattering by an infinite periodic array of Electron. (To be published)
thin conductors on a dielectric sheet', IEEE Trans., 1975, AP-23, pp. 15 LANGLEY, R.J., and PARKER, E.A.: 'Equivalent circuit model for
70-75 arrays of square loops', Electron. Lett., 1982,18, pp. 294-296
7 VARDAXOGLOU, J.C., and PARKER, E.A.: 'Performance of two
tripole arrays as frequency-selective surfaces', Electron. Lett., 1983, 19,
pp. 709-710

IEE PROCEEDINGS, Vol. 132, Pt. H, No. 6, OCTOBER 1985 399