Академический Документы
Профессиональный Документы
Культура Документы
arrays
A.J. Sangster
A.H.I. McCormick
c 300 F 300
£ 240 240
•= 180 = 180
_ 120
60
A 6 11 13 15 17 19
offset, mm slot length, mm
Fig. 3 Reflection coefficient phase as function of off-set from guide Fig. 5 Reflection coefficient phase as function of slot length, for longi-
centre Urn,for longitudinal dot in broadwall qfWG16 tudinal slot in broadwall of WGJ6
Frequency = 9J75 GHz; wall HwAn^s = L27 mm; number of basis Frequency = 9375 GHz; wall thiAnfiss = 121 mm; number of basis
functions = 1 functions = 1;
Slot length: Slot off-set:
16mm 8 nun
14 mm 5mm
12 mm 2mm
4 Program implementation
03 A flow diagram showing the basic steps in the synthesis
program is shown in Fig. 6. Most of the steps shown are
reasonably self-explanatory, but it is perhaps pertinent to
0.2 point out some aspects of the flow chart which have been
successful in promoting computational efficiency.
On the first pass through the program, and for the first
•o 0.1 slot in the array, improved values for lt and st are
deduced using the variational calculation. These are then
used as trial estimates in the moment method search
routine, to generate the desired first pass values for lt and
11 13 15 17 Sj. In subsequent passes trial values for the moment
slot length, mm method routine are obtained from the previous pass. At
Fig. 4 Reflection coefficient magnitude as function of dot length, for slot 2, on the first pass, lt and st are used as trial values
longitudinal slot in broadwall of WG16 in a variational search for s2, which is used with lx as
Frequency = 9375 GHz; wall thirirmess = L27 mm; number of basis suitable trial values for the moment method search for
functions = 1; first pass values for l2 and s2 - For slot n (n ^ 3) the pre-
Slot off-set:
.... 8mm vious value of slot length and an estimated slot-off-set,
5mm deduced by interpolation from the gradient of |p|, are
2mm used as trial values in the Gauss-Newton search for lm
and sa. The computation is repeated, as indicated on the
calculation, and consequently many of these modal com- flow-chart, until the values of lm and sn change from their
putations need be performed only once, for the first slot, values on the previous pass, by an increment which is set
values being stored for re-use in subsequent slot calcu- at just less than realistic fabrication tolerances. It has
lations. Computational speed in evaluating the integrals been found that 3 to 4 passes is usually enough to,
in eqns. 3-6 is also critically dependent on evolving effi- achieve this condition. *
cient convergence testing procedures. The final step in the program involves calculating the
Finally, these moment method equations are incorpo- small radiation phase and transmission phase deviations
rated into a corrected Gauss-Newton root-finding which occur at each slot, as a consequence of its finite
routine which generates improved values for /„ and sa. distributed length. These phase deviations are incorpo-
However, rather than search for slot self-admittance in rated into the design by making minor compensating
this routine, it has been found that searching for reflec- adjustments to the slot separation distances (see Fig. 7).
tion coefficient p is more effective. This is because with p 8.oil
it is easier to separate the effect of slot length from that of 5 Theoretical and experimental results
slot off-set, as a result of the differing ways in which these nqmoo
dimensions influence the magnitude and the phase of p, The array synthesis package described in the preoedfag
(Figs. 2-5). More particularly, the program makes use of Sections has been tested in two ways. Firstly, it
the observation that the magnitude of p is almost linearly used to design an array previously described
1EE PROCEEDINGS, VoL 136, PL H, No. i, FEBRUARY 1989
[7] and for which directly comparable results are from forward end-fire. The frequency of operation was
available. Secondly, it has been used to design a 19 specified as 9.375 GHz. The synthesis program predicts
element shunt slot array which, once constructed, was that a 21 element array is required, dimensioned as indi-
cated in Table 1.
pattern waveguide
specification dimensions IT
• frequency ex,
calculate roots
of array factor
Fig. 7 Schematic diagram of waveguide slots showing a and 5 phase
is specification increment number shifts
satisfied ? no. of roots
by 2 a = phase shift passed the slot
TyeT 5 = phase shift through the slot
determine coefficients
of array factor
Table 1: Elliott 21 element array
Stevenson estimate
for l n , s n No. Length, Offset, No. Length, Offset,
mm, mm mm mm
determine 1 15.374 1.689 12 15.167 1.794
z v 2 15.214 1.035 13 15.141 1.606
J
mn' mn 3 15.273 1.474 14 15.123 1.420
4 15.302 1.833 15 15.094 1.236
determine
5 15.306 2.131 16 15.066 1.044
6 15.313 2.315 17 15.040 0.866
7 15.308 2.375 18 14.996 0.715
8 15.296 2.353 19 14.984 0.551
9 15.270 2.273 20 15.013 0.374
10 15.243 2.144 21 14.930 0.562
11 15.212 1.978
trial values moment method
for Ij.s, using search for I ] . s {
variational method trial values from
previous pass
The slot length and slot off-set magnitudes shown in
moment method
this Table, are substantially in agreement with the corre-
search for I, , s. moment method sponding Elliott values, when due allowance is made for
search for l n , sn, n>1 the fact that Elliott's design is based on the use of mea-
trial I 2 =l,
trial l n = l n - i sured self-admittance values for round-ended slots and
trial s2 by trial s n = s n from also contains a constraint, which ensures that the result-
variational method previous pass
ant slot dimensions are consistent with data file values. In
Fig. 8, comparison is also made between theoretical and
moment method
search for l2, s 2 experimental results for this array. The Figure shows the
specified pattern (solid curve), a pattern deduced by
applying the method of Reference 7 to the slot distribu-
calculate tion of Table 1 (dotted curve), and a measured pattern
gradient of \p\
reconstructed from Elliott [7] (chain dotted curve). In
moment method
connection with this Figure, it is particularly interesting
search for l n , s n to note the discrepancy which occurs between the speci-
trial l n =l n _, fied pattern (solid curve) and the pattern associated with
trial s n from gradient the synthesised design. The discrepancy appears predomi-
of Ipl
nantly in the side lobe levels which are generally lower
than their 'ideal' value, except for those most remote
from the main beam. This error can largely be attributed
to the method used to evaluate mutual coupling, which
relies on analogy with dipole theory. This analogy neces-
sitates finding a relationship between slot voltages and
mode voltages, and currently this involves the use of the
} yes
approximate formula of Stevenson [12]. A more direct
radiation and transmission method of calculating the mutual impedances in slot
phase adjustments
arrays is clearly indicated, if this weakness in the synthe-
sis procedure is to be eliminated. However, when the syn-
thesised results are compared with Elliott's measurements
Fig. 6 Array synthesis program flow diagram (based on 144 data points; chain linked line, Fig. 8) the
degree of agreement indicated is not untypical of synthe-
comprehensively evaluated on an accurate near-field sised slot array designs.
measurement system. A nineteen element shunt slot array with staggered
The Elliott array is a Dolph-Chebyshev design with a slots has also been designed using the synthesis pro-
— 30 dB sidelobe specification with a mainbeam at 45° cedure described above. The specification for this array
42 IEE PROCEEDINGS, Vol. 136, Pt. H, No. 1, FEBRUARY 1989
required the main beam to be 9° from broadside, with coupling a staggered array suffers. This in turn means
— 24 dB sidelobes and in AT-band waveguide (WG16) at a that the Stevenson approximation which has been identi-
centre frequency of 9.375 GHz. The resultant design fied as the primary cause of these theoretical discrep-
which uses square-ended 1.6 mm wide slots is presented ancies is much less significant for this array. A measured
in Table 2. pattern for this array is compared with the 'ideal' pattern
-45
Fig. 8 Comparison of theoretical and measured far-field patterns for 21 element linear slotted-waveguide array
Frequency = 9.375 GHz; wall thickness = 1.27 mm; waveguide = WG16; slot width = 1.6 mm
ideal pattern
predicted pattern
• • — • • measured pattern
in Fig. 10, and shows that while the main beam position
Table 2: Shunt slot array and width are in good agreement with theory, the far-out
No. Length, mm Offset, mm Separation, mm side-lobes in particular are considerably in error. The
reasons for this error are predominantly due to machin-
1 15.171 +2.034 18.36
2 15.122 -1.485 18.34
ing inaccuracies, observed using an electron microscope,
3 15.109 +1.701 18.33 of the square-ended slots. These inaccuracies are high-
4 15.121 -2.046 18.31 lighted by near-field amplitude and phase patterns for the
5 15.157 +2.280 18.30 array (Figs. 11 and 12), which show that slots 10, 15, 18
6 15.167 -2.467 18.29 and 19 are particularly poorly formed, and when they are
7 15.189 +2.564 18.29
8 15.191 -2.595 18.29 computer corrected to eliminate the gross near-field
9 15.185 +2.549 18.30 pattern deviations, indicated in Figs. 11 and 12, the far-
10 15.178 -2.447 18.31 field pattern of the array is modified as shown in Fig. 13.
11 15.159 +2.301 18.32 The adjusted pattern is represented by the solid line
12 15.151 -2.122 18.34
13 15.150 +1.918 18.35
curve, and shows that the far-out side lobes which were
14 15.146 -1.702 18.36 particularly in error, are significantly improved by this
15 15.133 + 1.466 18.37 simple re-alignment.
16 15.136 -1.226 18.38
17 15.158 +1.006 18.36
18 15.013 -0.826 18.38 6 Conclusions
19 15.206 + 1.086
A computer aided design package for the synthesis of rec-
tangular waveguide broadwall slot arrays has been
The predicted far field pattern for this array (dotted described. The package differs from preceding contribu-
curve) has been compared with the 'ideal' or specified tions in this area in that the self-admittances of the slots
pattern (solid line) in Fig. 9. The differences between the are computed, rather than taken from a data bank of
two patterns are less than those observed for the Elliott measured values. The self-admittance computations rely
array, and this is due to the lower level of the mutual on the use of the moment method, which is known to be
IEE PROCEEDINGS, Vol. 136, Pt. H, No. 1, FEBRUARY 1989 43
- 10-
- 15 -
-20 -
-25 -
-30 -
-35 -
-40 -
-45
30 60 90 120 150
degrees
Fig. 9 Comparison of theoretical far-field patterns for 19 element linear slotted waveguide array
Frequency = 9J75 GHz; wall thickaess = 1.27 mm; waveguide = WG16
ideal pattern
predicted pattern
-30 -
-35 -
-40-
-45
30 60 90 120 150 180
degrees
Fig. 10 Comparison of theoretical and measured far-field patterns for 19 element linear slotted-waveguide array
Frequency = 9375 GHz; wal thickness = 1.27 mm; waveguide = WG16
measured patters
idea) pattern
-10 -
-15 -
-20 -
-25 -
-30 -
-35 -
-40 -
-45
60 90 120 150 180
degrees
Fig. 1 3 Comparison of measured and theoretically adjustedfar-field pattern for 19 element slotted-waveguide array
Frequency = 9.375 GHz; wall thickness = L27mm; waveguide = WG16
adjusted pattern
measured pattern