University of Pretoria etd – Mayhew-Ridgers, G (2004

)

CHAPTER

4

Numerical and Experimental Results

4.1

INTRODUCTORY REMARKS

The spectral-domain moment-method (SDMM) formulation of Chapter 3 was implemented in the
C programming language and, as such, this implementation can now be used to analyse any antenna configuration consisting of microstrip patches with capacitive feed probes. In this chapter,
it is shown how this implementation can be applied to a variety of applications. Throughout most
most of this chapter, the results of the SDMM are compared to those of two other commercial
codes, as well as some measured results. The two commercial codes are IE3D from Zeland Software [186] and FEKO from EM Software and Systems [187], while the circuit analysis module of
Sonnet, from Sonnet Software [188], was also used. The measurements were all performed at the
Centre for Electromagnetism at the University of Pretoria. This facility houses a Scientific Atlanta
compact antenna test range, integrated with a dedicated Hewlett-Packard 8510C vector network
analyser. For more general S-parameter measurements, a Hewlett-Packard 8510B vector network
analyser is also used. Antenna gain in this facility is typically measured by using the gain-transfer
method.
The SDMM implementation is validated in Section 4.2 for isolated parts of the antenna structure.
In Section 4.3, the new antenna elements are characterised in order to determine the effect of
the various geometrical parameters on the behaviour of the different elements. In Section 4.4, a
number of applications are presented. These include vertically- and horizontally-polarised arrays,
as well as ±45◦ slant-polarised arrays. The latter are often required for cellular base-stations
antennas. Finally, in Section 4.5, it is shown how the capacitive feed probes can be used with
alternatively-shaped resonant patches.

103

This then would be the equivalent of a single antenna element. is that far-field radiation towards the horizon (i.2 VALIDATION OF THE SPECTRAL-DOMAIN MOMENT-METHOD IMPLEMENTATION In order to validate the accuracy of the SDMM implementation. between two and seven basis functions were used on the probe. except under special conditions when the University of Pretoria—Electrical. Thereafter. G (2004) Chapter 4 Numerical and Experimental Results z θ 2a h L εr . A single probe that is embedded within a grounded substrate. θ = ±90◦ ) should always go down to zero. is probably the most basic part of the structure and is therefore firstly considered. it is advisable to first analyse isolated parts of the antenna structure.2. with length L and radius a.1 Geometry of a probe that is embedded within a grounded substrate. A detailed investigation into the capabilities of the various attachment modes also forms part of this validation. Electronic and Computer Engineering 104 . while the outer conductor of the coaxial cable is connected to the ground plane. 4. As such.1. which is embedded within a grounded substrate of thickness h. Figure 4. 4. The input impedance is usually a sensitive parameter and can give a good indication of how well the code performs. In doing so.University of Pretoria etd – Mayhew-Ridgers. The probe can be realised by extending the inner conductor of a coaxial cable through the ground plane into the substrate. and finally a resonant patch. one can obtain a better understanding of the capabilities of the formulation. 190].1 Single Probe in a Grounded Substrate Consider the probe of Figure 4. This problem has been addressed before and published results for the input impedance of such a probe are available within the open literature [189. For the SDMM formulation. It can be seen that the agreement between all the codes is very good and that it also compares exceptionally well with the published results. tan δε Connector Figure 4. it is used as the basis for comparison throughout most of the validation process. the structure is extended by adding a capacitor patch to the probe.2 shows simulated and published results for the input impedance versus normalised length of a probe that is totally embedded within a grounded substrate.e. Another interesting phenomena that can be validated here.

6 (a) IE3D Published FEKO SDMM 60 40 Input reactance (Ohm) 20 0 -20 -40 -60 -80 -100 -120 0.1 1. (b) Input reactance.5 mm.4 (b) Figure 4.8 0.8 0.4 1.7 0.9 1. Electronic and Computer Engineering 105 .0 k0 L 1.2 1.7 0.6 0. εr = 3.University of Pretoria etd – Mayhew-Ridgers.1 1. G (2004) Chapter 4 Numerical and Experimental Results IE3D Published FEKO SDMM 160 140 Input resistance (Ohm) 120 100 80 60 40 20 0 0.5 1.5 0.5 0.4 0.6 1.9 and tan δε = 0. h = 0. University of Pretoria—Electrical.2 1.5 1. Parameters: a = 1.6 0. (a) Input resistance.3 1.3λ0 .9 1.0 k0 L 1.4 0.3 1.2 Input impedance of a probe that is embedded within a grounded substrate.

both of them are used University of Pretoria—Electrical. Figure 4. 2. it can indeed be seen that there is far-field radiation towards the horizon and that the results of the SDMM are virtually indistinguishable from those of the two commercial codes.2. As discussed in Chapter 3. εr = 3. the performance of these attachment modes under a wide variety of geometrical parameters. Parameters: L = 48 mm. Electronic and Computer Engineering 106 . 4. 191]. 3 . four basis functions were used on the probe. can model probes in multilayered substrates very successfully. h= √ 2 εr − 1 i = 1. there are two common approaches: the rectangular attachment mode and the circular attachment mode.University of Pretoria etd – Mayhew-Ridgers. it is clear that the SDMM. In order to get a better understanding of the abilities of these two approaches.3 shows the radiation pattern for the case where i = 1. This is due to the stationary-phase evaluation of the far fields. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 0 Relative gain (dB) -5 -10 -15 -20 -25 -30 -90 -60 -30 0 θ (degrees) 30 60 90 Figure 4. . (4.5 mm. all the codes give zero far fields at exactly θ = ±90◦ . h = 90 mm. tan δε = 0 and f = 1 GHz. However. From the results that have been presented here. From this figure.2 Circular Versus Rectangular Attachment Modes In order to model the connection between a probe and a patch.3 Radiation pattern of a probe that is embedded within a grounded substrate. also known as an attachment mode.1) is satisfied [189. is not well documented. a special basis function. In this case. .78. a = 1. relation iλ0 . Note that although there is radiation towards the horizon. is required. as implemented here.

together with two sets of numerical results that were calculated by using the two different attachment-mode approaches.0). as measured by them.1) modes were used for the y-directed current. (b) Side view of the multilayered substrate.0). Figure 4.133]. the (1.5 shows the input impedance of this antenna. Electronic and Computer Engineering 107 .4. as well as Aberle et al. As can be seen from the results in Figure 4. The best results for the circular attachment mode were achieved by extending the radius of the attachment mode to the closest edge of the patch.4 Geometry of a rectangular probe-fed patch. in order to verify that both approaches yield the correct results as claimed in the literature. The simulated results are first compared to published results for a conventional narrowband microstrip antenna on a thin substrate. Shaubert et al.4 GHz and 2.0). and the other set having much smaller patches (below University of Pretoria—Electrical. which have either been published or measured specifically for this study. in this section to analyse various rectangular probe-fed patch configurations. G (2004) Chapter 4 Numerical and Experimental Results y z θ h(1) d φ Layer 1 x W ø = 2a h(2) L Layer 2 Connector (a) (b) Figure 4. Both approaches are then tested for small and resonant probe-fed patches that reside on multilayered substrates of varying thickness. It is not clear from the literature how the attachment modes would perform for much smaller patch sizes and thick substrates. (1. (7. two sets of patches were constructed.0). the one set having resonant patches (at frequencies between 1. To gain a better understanding of how they would perform. This is specifically for the case of a resonant patch on a relatively thin substrate.2). have published input-impedance results for a narrowband probe-fed microstrip patch antenna on a thin substrate. both approaches are capable of generating results that agree very well with measurements.1 GHz) of 50 mm × 50 mm.0) and (9. [132. The simulated results are then compared to experimental results. The general geometry of such a patch is shown in Figure 4. Both sets of calculated results were obtained by using nine entire-domain sinusoidal basis functions on the patch and one piecewise sinusoidal (PWS) basis function (excluding the one associated with the attachment mode) on the probe.University of Pretoria etd – Mayhew-Ridgers. For the entire-domain sinusoidal basis functions. (3.5. (a) Top view of the patch. [192]. (2.0) modes were used for the x-directed current. while the (2. (5.0) and (2.

Once again. Parameters: L = 9. a = 0. For all of the substrate thicknesses.0).2 GHz Me Measured Sub Circular Rectangular Ent 4.University of Pretoria etd – Mayhew-Ridgers.4 GHz  4. (2. (5.5 4. (3.5 mm. εr(1) = 10.6 GHz 4.7 shows the measured and calculated input-impedance results for the set of small patches.0).27 mm. for the entiredomain sinusoidal basis functions. while the (2. h(1) = 1. but that the rectangular attachment mode has difficulty in modelling the small patches.2 mm. In this case it is clear that the circular attachment mode still yields results that compare favourably with the measurements.2). W = 12. tan δε(1) = 0. the (1.8 GHz  ¢¡  4. A radius of 10 mm provided the best results for the circular attachment mode. Figure 4. Figure 4. G (2004) Chapter 4 Numerical and Experimental Results  ©¨ § 4. The two sets of patches would therefore represent a combination of different sizes and substrate thicknesses. Three of the four patches contained air gaps. depending on the substrate thickness.45 mm.0). and one to seven piecewise sinusoidal basis functions (excluding the one associated with the attachment mode) on the probe.5 GHz 1 2 5 ¥ £ ¦¤ !       Figure 4.002 and h(2) = 0. between the FR-4 layer and the ground plane. both sets of calculated results were obtained by using forty-five subdomain rooftop basis functions on the patches. In this case.6 shows the measured and calculated input-impedance results for the set of resonant patches. ranging from more or less 5 mm to 15 mm. d = 3.1) modes were used for the y-directed current.6 mm layer of FR-4.0) and (2.5 Input impedance (normalised to 50 Ω) of a rectangular probe-fed patch as modelled with the circular and rectangular attachment modes.0) modes were used for the x-directed current. Both sets of calculated results were obtained by using nine entire-domain basis functions on the patches. it would appear as though the rectangular University of Pretoria—Electrical. Both sets contained four patches and were etched on a 1. and one to seven PWS basis functions (excluding the one associated with the attachment mode) on the probe. (7. Electronic and Computer Engineering 108 .0).2 0.5 mm. resonance) at 5 mm × 10 mm.3 GHz 4. It would therefore appear that both approaches can handle thick multilayered substrates with very good accuracy.53. It is apparent that both attachment-mode approaches yield results that compare very well with the measured results.7 GHz 00 0. provided that the patch is of a resonant size.0) and (9. (1. depending on the substrate thickness.

8 1.02 and tan δε(2) = 0. University of Pretoria—Electrical.48 1. a = 0.0 Frequency (GHz) 2.45 mm.8 mm.45 mm.44 1.25.1 2. (a) h(2) = 0.6 mm. (b) h(2) = 5.1 mm.7 1. d = 15 mm. h(1) = 1.2 (b) 100 Reactance (Ohm) Rectangular 80 50 100 Circular 2. a = 0.6 1.9 2. (c) h(2) = 11.7 1.64 mm.9 2.4 mm.University of Pretoria etd – Mayhew-Ridgers. tan δε(1) = 0. a = 0.1 2. a = 0.9 2.8 1. εr(1) = 4.2 (d) Figure 4.0 Frequency (GHz) 2.6 Input impedance of a square resonant probe-fed patch as modelled with the circular and rectangular attachment modes.7 (a) Meas Circular Meas Rectangular Resistance (Ohm) Resistance (Ohm) 80 60 40 20 Reactance (Ohm) 0 160 140 120 1. Parameters: L = 50 mm.40 Frequency (GHz) 80 60 40 20 1.8 1. G (2004) Chapter 4 Numerical and Experimental Results Circular Meas Meas Rectangular Resistance (Ohm) Resistance (Ohm) 40 30 20 10 30 Reactance (Ohm) Reactance (Ohm) 60 40 20 0 100 0 40 20 10 0 -10 -20 1.52 1. W = 50 mm.6 1. (d) h(2) = 15.45 mm.36 1. Electronic and Computer Engineering 109 .0 Frequency (GHz) 1.1 2.2 Circular Rectangular 140 120 100 80 60 40 20 0 280 260 240 220 200 180 160 1.6 (c) 1.

0 Frequency (GHz) 2.7 (a) Meas Circular Rectangular Meas Resistance (Ohm) Resistance (Ohm) 2.9 2. G (2004) Chapter 4 Numerical and Experimental Results Meas Circular Rectangular Meas 8 6 4 2 20 15 10 5 0 -40 Reactance (Ohm) Reactance (Ohm) 0 -20 -40 -60 -80 -100 -120 1.0 Frequency (GHz) 2.0 Frequency (GHz) 2. (b) h(2) = 5.1 -80 -120 -160 -200 -240 2.45 mm.8 1.45 mm.University of Pretoria etd – Mayhew-Ridgers.6 mm. tan δε(1) = 0. (a) h(2) = 0. W = 10 mm. d = 2. h(1) = 1.6 1.6 (c) 1. Electronic and Computer Engineering 110 .7 1.6 1.1 2.2 1.9 2. a = 0.2 0 -40 -80 -120 -160 1. a = 0.8 1.02 and tan δε(2) = 0.6 1. (c) h(2) = 10.9 2.25.7 1.3 mm.3 mm.1 2.5 mm. εr(1) = 4.64 mm.1 2.45 mm.8 1.7 1. a = 0.2 Circular Rectangular 80 40 30 20 10 60 40 20 0 40 Reactance (Ohm) 0 0 Reactance (Ohm) 1. a = 0. Parameters: L = 5 mm. University of Pretoria—Electrical.7 Input impedance of a small rectangular probe-fed patch as modelled with the circular and rectangular attachment modes.2 (d) Figure 4.2 mm.9 2. (d) h(2) = 15.8 1.0 Frequency (GHz) (b) 50 -40 -80 -120 -160 -200 Rectangular 25 Resistance (Ohm) Resistance (Ohm) 10 Circular 1.

the radius of the circular attachment mode was extended to the two closest edges for the best results.8 to 4. if these basis functions are also decreased in size. as it is more versatile than the rectangular one. however.3 Capacitor Patches Following from the investigation into the attachment modes. taking into account all of the observations. In this case. was modelled with four PWS basis functions (excluding the one associated with the attachment mode). will now be used to analyse both the circular and rectangular probe-fed capacitor patches. It has been stated by Taboada [108] that for triangular Rao. but that the rectangular attachment mode only works well for resonant patches. as well as a higher-order circular attachment mode that was developed in Chapter 3. Wilton and Glisson (RWG) basis functions [104]. is that it can be used with any shape of patch and that it is faster to evaluate. Here. It also seems to be true in this case. attachment modes. One uncertainty related to the circular attachment mode. It will now be used to further validate the SDMM implementation in terms of its ability to model both the circular and rectangular probe-fed capacitor patches as shown in Figures 4. while the current density on the probe. upon which the rectangular attachment mode is based. having a radius of a. Furthermore. SDMM refers to the implementation using the normal University of Pretoria—Electrical. The electric current density on the patch was modelled with only the circular attachment mode. is not a good approximation when dealing with very small patches. The circular attachment mode. having a radius of b.10. It would appear that the magnetic-wall cavity model with a uniform current filament source. Finally. This investigation was an attempt to gain a better understanding of the abilities of two widelyused attachment modes for a SDMM analysis of probe-fed microstrip patch antennas. implying that it should be possible to use a smaller radius for the attachment mode together with underlying subdomain basis functions. Figure 4. thereby resulting in the inaccuracies that have been observed. as a function of the probe radius. that the circular attachment mode treats all patch sizes and substrate thicknesses with good accuracy. is the choice of the radius over which the current spreads on the surface of the patch. albeit it on thin and thick substrates. In all cases that were considered. which is based on the eigenmodes of the magnetic-wall cavity model. must be derived from scratch for other shapes of patches. It is clear from the results that have been presented here. analogous to the rectangular attachment mode. 4. the accuracy increased as the radius of the disk was increased.University of Pretoria etd – Mayhew-Ridgers. G (2004) Chapter 4 Numerical and Experimental Results attachment mode mainly yields an offset in the reactive part of the input impedance.8. Consider the circular capacitor patch as shown in Figure 4.2. Electronic and Computer Engineering 111 . it makes sense to retain only the circular attachment mode. it would appear that the circular attachment mode is the more versatile of the two.11 shows the simulated input impedance of the circular capacitor patch. it is also shown for two different radii of the capacitor patch. Additionally. the support of the attachment mode should be larger than the underlying subdomain basis functions. Some other benefits of the circular attachment mode over the rectangular attachment mode.

The validation will now continue by adding a square resonant patch to the structure. however.9. there is not much difference between the results of the normal circular attachment mode and the higher-order version.University of Pretoria etd – Mayhew-Ridgers. The geometry is shown in Figure 4. It can be seen that both the magnitude and the phase of the coupling. was modelled with subdomain rooftop basis functions.10. For the patch size of 5 mm × 5 mm. In order to make sure that the SDMM implementation can also calculate the coupling between two capacitor patches accurately. apart from a small difference in the input reactance for the larger patch. as predicted by the SDMM implementation. having a size of l × w. as far as the input reactance goes. forty-seven were used. having a radius of a. while SDMM (HO) refers to the implementation using the higher-order circular attachment mode. The same number of basis functions on each patch and probe was used as for the previous example.13 shows the simulated input impedance of the rectangular capacitor patch. In this case. In general. resulting in a single patch antenna element with a capacitive feed probe. it is also shown for two sizes of the rectangular patch. was modelled with four piecewise sinusoidal basis functions (excluding the one associated with the attachment mode).12 shows the simulated coupling between the two capacitor patches. Once again. while for the patch size of 5 mm × 20 mm. two probe-fed circular capacitor patches in fairly close proximity. G (2004) Chapter 4 Numerical and Experimental Results circular attachment mode. the higher-order circular attachment mode was used (the normal one basically gives the same results). Electronic and Computer Engineering 112 . Furthermore. Consider the rectangular capacitor patch as shown in Figure 4. Additionally. In this case it can be seen that. while the current density on the probe. It can be seen that the higher-order circular attachment mode predicts a slightly higher input reactance than the normal circular attachment mode. were analysed. For the small patch. SDMM refers to the implementation using the normal circular attachment mode. agrees very well with the results of both IE3D and FEKO. Figure 4. as a function of the probe radius. a radius of 2. while SDMM (HO) refers to the implementation using the higher-order circular attachment mode.5 mm was used for the circular attachment mode. University of Pretoria—Electrical. but for the larger patch. The results in this section have shown that the SDMM implementation can model the probe-fed capacitor patches quite successfully. twelve subdomain rooftop basis were used on the patch. Overall though. In both cases. there is good agreement between the various codes. there seem to be more variation in the results. Figure 4. it would appear that the agreement is not far off. while the results of the normal circular attachment mode seem to be closer to the FEKO results. The electric current density on the patch. all the results agree fairly well. the results for the higher-order circular attachment mode seem to follow the IE3D results rather closely.

(b) Side view of the multilayered substrate with εr = 4.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer. (b) Side view of the multilayered substrate with εr = 4. (b) Side view of the multilayered substrate with εr = 4. y z θ 1. y z θ 1. Electronic and Computer Engineering 113 .8 Geometry of the circular capacitor patch. University of Pretoria—Electrical.6 mm φ FR-4 2b x 15 mm Air ø = 2a Connector (a) (b) Figure 4. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1.University of Pretoria etd – Mayhew-Ridgers.8 mm 30 mm Connector (a) (b) Figure 4.25 and tan δε = 0. (a) Top view of the patch. (a) Top view of the patch.25 and tan δε = 0.10 Geometry of the rectangular capacitor patch.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer.6 mm 8 mm Port 2 φ Port 1 FR-4 x 15 mm Air ø = 0. (a) Top view of the patches.6 mm φ w FR-4 x l 15 mm Air ø = 2a Connector (a) (b) Figure 4.25 and tan δε = 0.9 Geometry of two circular capacitor patches in close proximity.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer.

(b) Input reactance.0 (a) SDMM (HO) SDMM FEKO IE3D 40 Input reactance (Ohm) 20 0 -20 b = 4 mm -40 -60 b = 2.9 1.7 0.1 0.5 a (mm) 0. (a) Input resistance.3 0.1 0.9 1.5 a (mm) 0. G (2004) Chapter 4 Numerical and Experimental Results SDMM (HO) SDMM FEKO IE3D 20 Input resistance (Ohm) 15 b = 4 mm 10 b = 2.2 0.University of Pretoria etd – Mayhew-Ridgers.4 0.6 0. University of Pretoria—Electrical.8 0.7 0.5 mm -80 -100 0.2 0.8 0.0 (b) Figure 4.4 0.11 Input impedance (at 1.6 0.3 0.5 mm 5 0 0.8 GHz) of the circular capacitor patch. Electronic and Computer Engineering 114 .

University of Pretoria—Electrical.2 (a) IE3D FEKO SDMM (HO) 80 60 40 P ∠S1.8 Frequency (GHz) 2.0 2. G (2004) Chapter 4 Numerical and Experimental Results IE3D FEKO SDMM (HO) 0 -5 P |S1.6 1.9 1.12 Coupling between the two circular capacitor patches.4 1.9 1.8 Frequency (GHz) 2. (b) Phase.0 (b) Figure 4.5 1. Electronic and Computer Engineering 115 .6 1.1 2.University of Pretoria etd – Mayhew-Ridgers.1 2.2 (degrees) 20 0 -20 -40 -60 -80 -100 1.7 1.7 1.4 1. (a) Magnitude.5 1.2 2.2 | (dB) -10 -15 -20 -25 -30 1.

University of Pretoria—Electrical. w = 5 mm 5 0 0. w = 20 mm 15 10 l = 5 mm.8 GHz) of the rectangular capacitor patch.University of Pretoria etd – Mayhew-Ridgers.7 0.1 0. w = 20 mm 25 0 -25 -50 l = 5 mm.4 0.0 (a) SDMM (HO) SDMM FEKO IE3D 100 75 Input reactance (Ohm) 50 l = 5 mm.8 0.2 0.0 (b) Figure 4.3 0.5 a (mm) 0.8 0.9 1. G (2004) Chapter 4 Numerical and Experimental Results SDMM (HO) SDMM FEKO IE3D 25 Input resistance (Ohm) 20 l = 5 mm.6 0. w = 5 mm -75 -100 0.9 1.5 a (mm) 0.1 0.13 Input impedance (at 1.7 0. (b) Input reactance.3 0.6 0.4 0.2 0. Electronic and Computer Engineering 116 . (a) Input resistance.

the electric current density on the structure was finally modelled with eleven entire-domain sinusoidal basis functions on the resonant patch. in the two graphs. Each curve in Figures 4. the single circular University of Pretoria—Electrical. The first curve assumes that there is no y-directed mode. both configurations will now be considered. Here too. Electronic and Computer Engineering 117 . (b) Side view of the multilayered substrate with εr = 4. as shown in Figure 4. Furthermore. after which y-directed modes are added accumulatively for the rest of the curves.4 Antenna Elements with Capacitive Feed Probes The new antenna element is excited by making use of capacitive coupling through either a circular probe-fed capacitor patch or a rectangular probe-fed capacitor patch.15 show how the input impedance of the antenna varies as a function of the number of sinusoidal entire-domain modes on the resonant patch when keeping the number of segments on the probe fixed.2.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer. Based on the convergence tests. First consider the antenna element with a circular capacitor patch.University of Pretoria etd – Mayhew-Ridgers.25 and tan δε = 0. it can be seen that the results converge as the number of segments is increased. each curve corresponds with a specific number of y-directed modes. after which other x-directed modes are added accumulatively as one moves towards the right-hand side of the x axis (i. The results in Figure 4.16 provide some convergence tests.14 Geometry of the antenna element with a circular capacitor patch. The results in Figure 4. 4.15 and 4.16 show how the input impedance of the antenna varies as a function of the number of segments on the probe (each full PWS basis function spans two adjacent segments) when keeping the number of sinusoidal entire-domain basis functions on the resonant patch fixed.14. (a) Top view of the antenna element.0) xdirected mode on the left-hand side of the x axis. In order to obtain a better understanding in terms of the choice and number of basis functions that are required to model the antenna element. all the modes towards its left are also included).e.6 mm 52 mm 2b φ FR-4 x h Air ø = 0. From these results it can definitely be seen that the results converge as the number of x.and y-directed modes is increased. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1. In terms of validating the SDMM implementation. for a specific x-directed mode. Figures 4.15(a) and (b) starts with the (1.8 mm 52 mm Connector d (a) (b) Figure 4.

0 (1 d ad ) .0 (9 d ad x-directed sinusoidal entire-domain modes ad ) .2 (1 ad d ) . (1 ad d 0) 3.2) add (2.2 (3 d ad ) .0 (3 d ad ) .0 (5 d ad ) .0) none Input resistance (Ohm) 70 60 50 40 30 ) .0 (7 d ) .5 -5.0 11 ( ad d ) .2 (1 ad d ) . (a) Input resistance.0 (7 d ) . but as a function of the entire-domain sinusoidal basis functions on the resonant patch.University of Pretoria etd – Mayhew-Ridgers.0 ) . Electronic and Computer Engineering 118 .0 Input reactance (Ohm) 7.2 (2 d ad ) .0 (2 d ad ) . (b) Input reactance.2) add (2.15 Input impedance (at 1.0 -2.5 mm and d = 4 mm.1) add (2.8 GHz and with four segments on the probe) of the antenna element with a circular capacitor patch.0 13 ( d ad ) .2 (3 d ad ) .0 (1 d ad ) . (1 d ad 0) 5.0 15 ( (b) Figure 4.5 0.0) none 10.0 2.0 (3 d ad ) .1) add (2.0 (9 d ad x-directed sinusoidal entire-domain modes ad 0) 1.0 (5 d ad ) .2 (2 d ad ) .5 5. (1 (a) y-directed sinusoidal entire-domain modes add (2. University of Pretoria—Electrical.0 (2 d ad ) . Parameters: b = 4. G (2004) Chapter 4 Numerical and Experimental Results y-directed sinusoidal entire-domain modes add (2.

2). (a) Input resistance.8 54. (3.0) and (9.6 55. (b) Input reactance. while the (2.0 2 3 4 7 6 5 Number of segments on the probe (b) Figure 4.2). (5. Parameters: b = 4. (2.0) and (2. the (1.0) modes were used for the x-directed current. University of Pretoria—Electrical.4 Input resistance (Ohm) 56.8 -1.University of Pretoria etd – Mayhew-Ridgers. G (2004) Chapter 4 Numerical and Experimental Results 56. For the entire-domain sinusoidal basis functions on the resonant patch.8 GHz) of the antenna element with a circular capacitor patch.0).0). (7. (3. (2.2 54. Electronic and Computer Engineering 119 . (1.5 mm and d = 4 mm.4 2 3 4 7 6 5 Number of segments on the probe 8 9 8 9 (a) Input reactance (Ohm) -0.0).2 -1.2).16 Input impedance (at 1.6 -2. but as a function of the number of segments on the probe.0 55.1) modes were used for the y-directed current.0).

17 shows the simulated and measured gain of the antenna element.1 2.2).2).17 Gain of the patch antenna with a circular capacitor patch. Electronic and Computer Engineering 120 .0) and (2. (7.5 mm and d = 4 mm.2 Figure 4. (3.18 shows the simulated and measured input impedance of the antenna element.0).0). (3. Parameters: b = 4. SDMM refers to the implementation using the normal circular attachment mode. and four PWS basis functions on the probe (excluding the one associated with the attachment mode).8 Frequency (GHz) 2. while there is practically no difference in the results obtained with the two circular attachment modes. The electric current density on the resonant patch is primarily directed along the x direction. Figure 4. For the entire-domain sinusoidal basis functions. (1. (5. once again.5 1. As can be seen. Figure 4.6 1. G (2004) Chapter 4 Numerical and Experimental Results IE3D Measured FEKO SDMM (HO) SDMM 10 Gain (dBi) 8 6 4 2 0 1. while SDMM (HO) refers to the implementation using the higher-order circular attachment mode.University of Pretoria etd – Mayhew-Ridgers.4 1.0). From these it can be seen that.2). especially for the reactive part of the input impedance. attachment mode on the capacitor patch. is fairly good. (2.1) modes were used for the y-directed current. The simulation with the higher-order circular attachment mode appears to provide results that agree better with those of the two commercial codes and the measurements than that of the normal circular attachment mode.19 shows the simulated radiation patterns in the E-plane and the H-plane of the antenna element.7 1. The cross-polarised pattern in the E-plane of the antenna element is very low and therefore has not University of Pretoria—Electrical.0 2. there is practically no difference in the results obtained with the two circular attachment modes.9 1. the (1.0) and (9. Here.0) modes were used for the x-directed current. the agreement between the three codes is fairly good. Figure 4. (2. is very good. while the (2.0). from which it can also be seen that the agreement between the three codes and the measurements. from which it can be seen that the agreement between the three codes and the measurements.

4 1.0 2. (a) Input resistance.7 1.1 2.4 1.8 Frequency (GHz) 2. University of Pretoria—Electrical.5 mm and d = 4 mm.6 1.8 Frequency (GHz) 2.2 (b) Figure 4.5 1.University of Pretoria etd – Mayhew-Ridgers. (b) Input reactance. Electronic and Computer Engineering 121 .1 2.2 (a) IE3D Measured FEKO SDMM (HO) SDMM 40 Input reactance (Ohm) 20 0 -20 -40 -60 1.7 1. G (2004) Chapter 4 Numerical and Experimental Results IE3D Measured FEKO SDMM (HO) SDMM 70 Input resistance (Ohm) 60 50 40 30 20 10 0 1.0 2.18 Input impedance of the antenna element with a circular capacitor patch.9 1.9 1.6 1.5 1. Parameters: b = 4.

(a) Radiation pattern in the E-plane (φ = 0◦ ).University of Pretoria etd – Mayhew-Ridgers.5 mm and d = 4 mm. Electronic and Computer Engineering 122 . Parameters: b = 4. G (2004) Chapter 4 Numerical and Experimental Results SDMM (HO) SDMM FEKO IE3D 0 Relative gain (dB) -5 -10 -15 -20 -90 -30 -60 0 θ (degrees) 30 60 90 (a) SDMM (HO) SDMM FEKO IE3D 0 Relative gain (dB) -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 (b) Figure 4. (b) Radiation pattern in the H-plane (φ = 90◦ ). University of Pretoria—Electrical.8 GHz) of the antenna element with a circular capacitor patch.19 Radiation patterns (at 1.

the SDMM results lie in between those of the other two codes. Four PWS basis functions were used on the probe (excluding the one associated with the attachment mode). (5.5 mm was used for the attachment mode. A radius of 2.20. Also. However. Here too. the agreement between the three codes is fairly good and that there is also practically no difference in the results obtained with the two circular attachment modes. (2.0) modes were used for the x-directed current. The electric current density on the structure was modelled with eleven entire-domain sinusoidal basis functions on the resonant patch. it can be seen that the cross-polarised pattern in the H-plane is somewhat higher. and twenty-two subdomain rooftop basis functions on the capacitor patch. from which it can be seen that the agreement between the three codes and the measurements.8 mm 52 mm Connector d (a) (b) Figure 4.6 mm φ 52 mm l w FR-4 x h Air ø = 0. is very good. From these it can be seen that.2). once again. The cross-polarised pattern in the H-plane shows some spread between the results of the three codes. there is practically no difference in the results obtained with the two circular attachment modes.23 shows the simulated radiation patterns in the E-plane and the H-plane of the antenna element. as shown in Figure 4. (7. Figure 4.25 and tan δε = 0. University of Pretoria—Electrical.0).0). as θ approaches ±90◦ ). (a) Top view of the antenna element. while the (2.University of Pretoria etd – Mayhew-Ridgers.0).0) and (2. there is not so much difference between the results obtained with the two circular attachment modes.e. seems to be slightly lower when compared to the rest. the circular attachment mode. This behaviour is due to the symmetry in one plane of the structure. Electronic and Computer Engineering 123 . but the asymmetry in the orthogonal plane. consider the patch antenna element with a rectangular capacitor patch. Now.20 Geometry of the antenna element with a rectangular capacitor patch. as compared to the antenna element with the circular capacitor patch. been shown.1) modes were used for the y-directed current.2). especially towards the horizon (i.2). For the entire-domain sinusoidal basis functions. Also. (3. The input reactance as simulated by the SDMM implementation. (1.0) and (9.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer.22 shows the simulated and measured input impedance of the antenna element. (3. (b) Side view of the multilayered substrate with εr = 4. from which it can also be seen that the agreement between the three codes and the measurements. is fairly good. On the other hand. the (1.21 shows the simulated and measured gain of the antenna element. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1.0). Figure 4. Figure 4. (2.

it can be concluded that the SDMM implementation compares very well with measurements and also with the two commercial codes. Following from all of the results that have been presented in this section.6 1. University of Pretoria—Electrical. it will now be used to characterise antenna elements with capacitive feed probes.8 Frequency (GHz) 2. In the next section.4 1.2 Figure 4.University of Pretoria etd – Mayhew-Ridgers. The higher-order circular attachment mode will be used henceforth. w = 10 mm and d = 8 mm. Parameters: l = 5 mm.9 1. as it appears to provide results that agree well with those of the two commercial codes and the measurements when considering all the configurations that were used for the validation.0 2.1 2.5 1. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D Measured SDMM (HO) SDMM 10 Gain (dBi) 8 6 4 2 0 1.7 1.21 Gain of the antenna element with a rectangular capacitor patch. the cross-polarised pattern of this antenna element appears to be slightly higher than the one for the element with the circular capacitor patch. Electronic and Computer Engineering 124 .

22 Input impedance of the antenna element with a rectangular capacitor patch.0 2.7 1.2 (a) IE3D Measured FEKO SDMM (HO) SDMM 40 Input reactance (Ohm) 20 0 -20 -40 -60 -80 -100 1. University of Pretoria—Electrical.7 1.4 1.2 (b) Figure 4. Parameters: l = 5 mm. (a) Input resistance. G (2004) Chapter 4 Numerical and Experimental Results IE3D Measured FEKO SDMM (HO) SDMM 60 Input resistance (Ohm) 50 40 30 20 10 0 1.6 1.4 1.5 1.9 1. w = 10 mm and d = 8 mm.0 2.8 Frequency (GHz) 2. (b) Input reactance.1 2.University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 125 .1 2.6 1.8 Frequency (GHz) 2.5 1.9 1.

Electronic and Computer Engineering 126 . (a) Radiation pattern in the E-plane (φ = 0◦ ). G (2004) Chapter 4 Numerical and Experimental Results SDMM (HO) SDMM FEKO IE3D 0 Relative gain (dB) -5 -10 -15 -20 -90 -30 -60 0 θ (degrees) 30 60 90 (a) SDMM (HO) SDMM FEKO IE3D 0 Relative gain (dB) -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 (b) Figure 4. Parameters: l = 5 mm.University of Pretoria etd – Mayhew-Ridgers.23 Radiation patterns (at 1.8 GHz) of the antenna element with a rectangular capacitor patch. w = 10 mm and d = 8 mm. (b) Radiation pattern in the H-plane (φ = 90◦ ). University of Pretoria—Electrical.

2. From the results it can be seen that the gap width once again affects both the input resistance and the input reactance.4 were used for the analysis. as shown in Figure 4.24(a) shows the effect of the gap width on the antenna input impedance. The effect of these parameters will now be addressed in more detail. Figure 4. and the one with the rectangular capacitor patch. The input reactance becomes more inductive as the radius (and therefore the size) of the capacitor patch is increased.University of Pretoria etd – Mayhew-Ridgers. Furthermore. mainly determines the input impedance of the antenna element. 4.4 were used for the analysis.24(b) shows the effect of the radius of the capacitor patch on the antenna’s input impedance. In both cases. as shown in Figure 4. G (2004) Chapter 4 4. Both the length and the width of the capacitor patch mainly affects the input reactance. while they only have a very small effect on the input resistance. University of Pretoria—Electrical. From the results it can be seen that the gap width affects both the input resistance and the input reactance. the size of the capacitor patch. The input reactance becomes more inductive as the dimensions (and therefore the size) of the capacitor patch are increased. the same basis functions as in Section 4. The length of the capacitor patch is denoted by l and its width by w. as shown in Figure 4. but that its effect on the input resistance is more significant than on the input reactance. Once again. The same basis functions as in Section 4.25(c) shows the effect of the width of the capacitor patch on the antenna’s input impedance. Figure 4. while the gap width between the resonant patch and the capacitor patch is denoted by d.1 Input Impedance of the Antenna Element Consider the antenna element with the circular capacitor patch. as well as the gap-width between the resonant patch and the capacitor patch. while the gap width between the resonant patch and the capacitor patch is denoted by d. but that its effect on the input resistance is more significant than on the input reactance.20.25(b) shows the effect of the length of the capacitor patch on the antenna’s input impedance.2. The radius of the capacitor patch is denoted by b. there are some key geometrical parameters that can be used to control the antenna’s input impedance and bandwidth. Figure 4. while it has hardly any effect on the input resistance. Both the input resistance and input reactance decrease as the gap is widened.3 Numerical and Experimental Results CHARACTERISATION OF ANTENNA ELEMENTS WITH CAPACITIVE FEED PROBES For both the antenna element with the circular capacitor patch. The bandwidth of the antenna element is mainly determined by the thickness of the substrate.14. consider the antenna element with the rectangular capacitor patch.3. Now. it can be seen that the radius of the capacitor patch mainly affects the input reactance.25(a) shows the effect of the gap width on the antenna’s input impedance.20. Electronic and Computer Engineering 127 . Both the input resistance and input reactance decrease as the gap is widened. while Figure 4.14. as shown in Figure 4. while Figure 4.

8 GHz) of some geometrical parameters on the input impedance of the antenna element with the rectangular capacitor patch.0 -40 15. G (2004) Chapter 4 Numerical and Experimental Results Imag Real 100 Imag Real 100 Input impedance (Ohm) 80 50 60 0 40 20 -50 0 -100 -20 0. (b) Radius of the capacitor patch (keep d = 4 mm). (c) Width of the capacitor patch (keep d = 8 mm and w = 10 mm).0 l (mm) 10.8 GHz) of some geometrical parameters on the input impedance of the antenna element with the circular capacitor patch.0 d (mm) 10.0 d (mm) (a) -40 10. Real Imag 100 80 Input impedance (Ohm) Real Imag Real Imag 60 60 40 40 60 20 20 40 0 0 20 -20 -20 0 -20 10.5 mm).24 Effect (at 1.0 (b) (a) Figure 4.0 10.0 (c) Figure 4.0 5. (b) Length of the capacitor patch (keep d = 8 mm and l = 5 mm).0 (b) -60 0.University of Pretoria etd – Mayhew-Ridgers.0 b (mm) 0.0 5. (a) Gap width between rectangular resonant patch and capacitor patch (keep l = 5 mm and w = 10 mm).0 5. Electronic and Computer Engineering 128 .0 w (mm) 20.0 5. (a) Gap width between rectangular resonant patch and capacitor patch (keep b = 4.25 Effect (at 1. University of Pretoria—Electrical.

At every thickness of the air layer. an attempt was made to maximise the impedance bandwidth of the antenna element (i. the capacitor patch cannot be placed close enough to the resonant patch and would probably have to be positioned below the resonant patch. was varied from 12. and therefore also the achievable bandwidths that can be obtained. For both sets of results. the optimum bandwidth is achieved for air layers that are between 15 mm and 20 mm thick. The bandwidths that correspond to voltage standing-wave ratios (VSWRs) of 1.20. the thickness of the air layer. It can be seen that for these particular antenna elements and specific operating frequency. for the antenna elements with the circular and rectangular capacitor patches respectively. G (2004) Chapter 4 4. especially in an array University of Pretoria—Electrical. after which it decreases again.e. A possible explanation for this behaviour is that the coupling mechanism between the various parts of the antenna element is quite complex and also that an attempt was made to match the antenna to 50 Ω at each substrate thickness that was evaluated. of which parameters are responsible for which behaviour. the capacitor patch has to be bigger in size and closer to the resonant patch for the thinner air layers. is very important. In this context. at various substrate thicknesses. Here.2. while for thicker air layers. the authors indeed show that the impedance bandwidth of probe-fed microstrip patch antennas increases with substrate thickness up to a certain point. Such a knowledge. Figure 4. For example. As the substrate thickness is increased.26 and 4. the loci exhibits a rather wide loop. For an impedance match on the thinner substrates. the capacitor patch has to be relatively large and closely spaced to the resonant patch. to have as much as possible of the input impedance loci inside the relevant constant VSWR circles on the Smith chart). most probably explains the unusual trend (i. As has already been pointed out. From the results that have been presented in this section. if the substrate is too thin.3.28 shows the behaviour of the impedance loci on a Smith chart for the antenna element with the circular capacitor patch.5:1 and 2:1. it has to be smaller in size and farther away from the resonant patch. it is worthwhile to also refer to some results in [193]. Electronic and Computer Engineering 129 . it is apparent that there are a number of parameters that can be used to control the behaviour of these antenna elements. the relation between impedance bandwidth and substrate thickness does not remain linear).2 Numerical and Experimental Results Impedance Bandwidth of the Antenna Element Figures 4. the fact that the overall size of the structure and the spacing between the different parts change with substrate thickness.University of Pretoria etd – Mayhew-Ridgers. In [193]. denoted by h in Figures 4.4 were also used for the analysis.e.27 show the impedance bandwidth. These relationships clearly place a limit on the range of substrate thicknesses that can be used. are shown.5 mm to 25 mm. but that for thicker substrates. for a thinner substrate thickness. the capacitor patch has to become smaller and also spaced farther apart from the resonant patch. up to a certain point where it more or less seems to taper off and decrease again. Therefore. concerning probe-fed microstrip patch antennas on electrically thick substrates.14 and 4. the loop becomes tighter. the same basis functions as in Section 4. It can be seen that the impedance bandwidth increases with increasing substrate thickness. It can be seen that.

0 20.0 h (mm) Figure 4.0 20.5:1 VSWR = 2:1 40 Bandwidth (%) 30 20 10 0 12. G (2004) Chapter 4 Numerical and Experimental Results VSWR = 1.5:1 VSWR = 2:1 40 Bandwidth (%) 30 20 10 0 12.5 22. VSWR = 1.8 GHz) versus substrate thickness for the antenna element with the rectangular capacitor patch.8 GHz) versus substrate thickness for the antenna element with the circular capacitor patch.5 15.27 Bandwidth (more or less around 1.0 17.5 22. University of Pretoria—Electrical.5 15.0 h (mm) Figure 4.5 25.0 17.University of Pretoria etd – Mayhew-Ridgers.26 Bandwidth (more or less around 1.5 25. Electronic and Computer Engineering 130 .

29.2.4. 4.32 GHz $#" A 0 0. a number of applications are addressed where patch antennas with capacitive feed probes can be very useful. G (2004) Chapter 4 Numerical and Experimental Results h 12.3.University of Pretoria etd – Mayhew-Ridgers.4 APPLICATIONS Microstrip patch antennas are very often used in array configurations where there might be specific requirements in terms of antenna gain.4.1. In Sections 4. The characteristics of such a configuration are addressed in Section 4. vertically.e. In this section. Finally.28 Impedance loci (normalised to 50 Ω) of the antenna element with the circular capacitor patch. a 180◦ phase difference) with the first capacitor patch.8 are devoted to various ±45◦ slant- polarised arrays.4.4 and 4. in Section 4. it will now be shown how these antenna elements can be used in typical antenna applications.5 1 2 5 % ' (& CB @ 9 1.= 17. beamwidth and polarisation. it is important that the second capacitor patch gets excited exactly out of phase (i.4. This is then followed by a discussion on dual-polarised elements in Section 4. In Section 4. The only University of Pretoria—Electrical.4.2. 4.6.and horizontally-polarised arrays are presented.5 mm 6 10 ) 8 2.4.4. as illustrated in Figure 4. 4. However.5 mm h 22.4. it is first shown how the cross-polarisation levels can be reduced for a single antenna element.4. It is possible to reduce these levels by making the structure more symmetric.1 Antenna Element with Reduced Cross-Polarisation Levels As is evident from the results in Section 4.4. In the next section. This can be done by adding another probe-fed capacitor patch on the opposite side of the resonant patch. an antenna element with a single probe-fed capacitor patch might have cross-polarisation levels in the H-plane that are unacceptably high for some applications.7 and 4. environment where there are even more parameters that have an influence on the behaviour of each element.= 12.4 GHz 67 2 43 5 Figure 4.5 mm h 17. The building block for some of the arrays that are investigated.5. consists of two elements that are positioned in a back-to-back configuration. Electronic and Computer Engineering 131 . it is shown how the capacitive feed can be used with resonant patches that are not rectangular in shape. while Sections 4.2 0.= 22.5.

2 .1 P Z2. as it requires a feed network anyway. it then follows that the input impedance at port 1.2) where V1 and V2 are the voltages at ports 1 and 2 respectively. the input impedance at a specific port is a function of the self-impedance at that port as well as the mutual impedance between the two ports. (b) Side view of the multilayered substrate with εr = 4.29 Geometry of the antenna element with two capacitor patches for reduced crosspolarisation levels.8 mm 52 mm Connector 5.2 I2 .3).1 + Z1. (a) Top view of the antenna element. (4. due to the 180◦ phase difference between I1 and I2 .3) in From (4. However.1 Z1.University of Pretoria etd – Mayhew-Ridgers. The Z-parameters relate the port voltages to the port currents through    V1       V  2   = P P Z1.5 mm (b) (a) Figure 4. can be expressed as in Z1 = V1 I1 P P = Z1. drawback of such an element is that it requires some sort of feed network below the ground plane in order to split the signal and to provide the phase shift.1 I1 + Z1. The voltage at port 1 can therefore be expressed as P P V1 = Z1. ·    I  2 (4. Due to the fact that the antenna element is driven by two ports. I2 = −I1 .6 mm Port 2 52 mm 9m m φ FR-4 x Port 1 15 mm Air ø = 0. I1 (4. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1.1 − Z1. and I1 and I2 are the currents at these ports. Z1 .2     I1     .2 P Z2.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer. this is usually not a problem in an antenna array. University of Pretoria—Electrical. It is possible to express the input impedance at any of the ports by making use of the Z-parameters associated with the two ports.4) Now.2 I2 .25 and tan δε = 0. The input impedance at port 1 then reduces to P P in Z1 = Z1. Electronic and Computer Engineering (4.5) 132 .

2 Reactance (Ohm) 20 0 -20 P ℑm Z1.30 Z-parameters of the two ports of the antenna element with two capacitor patches for reduced cross-polarisation levels. (a) Resistive part of the Z-parameters. University of Pretoria—Electrical.6 1.8 Frequency (GHz) 2.7 1.5 1.1 2.9 1.6 1.4 1.1 2.7 2.2 (a) SDMM FEKO IE3D 40 P ℑm Z1.2 -20 -40 1.1 -40 -60 -80 1.5 1.University of Pretoria etd – Mayhew-Ridgers. (b) Reactive part of the Z-parameters.2 2.4 1.1 0 P ℜe Z1.9 1.0 (b) Figure 4.0 1. Electronic and Computer Engineering 133 . G (2004) Chapter 4 Numerical and Experimental Results SDMM FEKO IE3D 60 Resistance (Ohm) 40 20 P ℜe Z1.8 Frequency (GHz) 2.

G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 10 Relative gain (dB) 5 0 -5 -10 -90 -60 -30 0 θ (degrees) 30 60 90 60 90 (a) FEKO IE3D SDMM 10 Relative gain (dB) 5 0 -5 -10 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4.8 GHz) of the antenna element with two capacitor patches for reduced cross-polarisation levels. University of Pretoria—Electrical.University of Pretoria etd – Mayhew-Ridgers.31 Radiation patterns (at 1. (a) Radiation pattern in the E-plane (φ = 0◦ ). Electronic and Computer Engineering 134 . (b) Radiation pattern in the H-plane (φ = 90◦ ).

Figure 4. the expression for the input impedance at port 2 follows in a similar way.4 were used for the SDMM analysis.1 and Z2. the single higherorder circular attachment mode on each capacitor patch.8) and P P ℑm Z2.1 . (4. Electronic and Computer Engineering 135 .1 and Z2. This element was designed for an input impedance of 100 Ω at each port. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on the resonant patch. The cross-polar levels in the H-plane have now effectively been reduced to levels that fall well below the range of the graph. which can then transform the overall input impedance down to 50 Ω.32. it is necessary to have antenna elements with two orthogonal polarisations.6) P P ℑm Z1.University of Pretoria etd – Mayhew-Ridgers.6) to (4.2 − ℜe Z2. As in Section 4.4.4.2 = ℑm Z2.2. will primarily be directed along the y direction.2 .30 shows the simulated Z-parameters of the antenna element in Figure 4. the element will usually be implemented with a power splitter below the ground plane.2 can then be changed until (4.1 = Z0 (4. For a Z0 input impedance at port 2. as well as the separation distance between each P P capacitor patch and the resonant patch.1 (4.3. For the entire-domain sinusoidal University of Pretoria—Electrical. Once again. The antenna element can be designed through an iterative process. the size of the capacitor patches. For specific values of P P P P Z1. The dimensions of the dual-polarised element of Figure 4.29. The same basis functions as in Section 4. P P ℜe Z2. the three codes show good agreement.2. G (2004) Chapter 4 Numerical and Experimental Results If one now wants to set the input impedance at port 1 equal to Z0 .31 shows the simulated radiation patterns in the E-plane and the H-plane of the antenna element.1 = ℑm Z1. while that for port 2. which should be equal by the way.32 are similar to those of the singlepolarised element of Section 4.2 = Z0 (4. From the results it can be seen that the agreement between all three codes is quite good. and four PWS basis functions on each probe (excluding the one associated with the attachment mode).9) are satisfied. The electric current density on the resonant patch will primarily be directed along the x direction for port 1. one port is used to generate the one polarisation. Z1. This can be achieved with the antenna element depicted in Figure 4.2 and Z2. Here. can be used to control Z1.2 Dual-Polarised Antenna Element In many applications. As stated before. Figure 4. 4. while the other port is used to generate the orthogonal polarisation.9) have to be satisfied. it implies that P P ℜe Z1.7) and that Of course.1 − ℜe Z1.2 .

the coupling between the two ports of the dual-polarised antenna element might be too high. It might. This can be improved by using symmetric feeds. is quite good. while the same set of modes were also used for the y-directed current. (2.2). the (1. It can also be seen that the coupling between the two ports vary between approximately −15 dB and −20 dB over the operating frequency band.6 mm φ Port 1 FR-4 x 15 mm Air ø = 0. basis functions. For some applications.1).33(a) shows the simulated and measured reflection coefficient at port 1 of the antenna element. In this case. although a second port was introduced (note that the dimensions of the capacitor patches. as well as those of the SDMM.University of Pretoria etd – Mayhew-Ridgers.0). (1. similar to those in Section 4. as well as the distance between each capacitor patch and the resonant patch. while Figure 4.1. Figure 4.25 and tan δε = 0. require a more complex feed network.33(b) shows the simulated and measured coupling between the two ports of the antenna element.4. (a) Top view of the antenna element. however.8 mm 52 mm Connector 4 mm (a) (b) Figure 4.32 Geometry of the dual-polarised antenna element. From these it can be seen that the element remains well matched over a wide bandwidth. Electronic and Computer Engineering 136 . (7. In some of the following sections. it will also be shown how various array configurations can be used to realise dual-polarised antennas with less coupling between the two ports. University of Pretoria—Electrical. (3.0).0).2.2). the agreement between the measurements and the simulations. (3.0) modes were used for the x-directed current. the results of FEKO. In both cases.0). (2.4 for the single-polarised antenna element). G (2004) Chapter 4 Numerical and Experimental Results y z Port 2 θ 9m 52 mm m 1. (5. (b) Side view of the multilayered substrate with εr = 4. were kept equal to those of Section 4.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer.0) and (9. seems to compare slightly better to the measurements than the results of IE3D do.

6 1.1 | (dB) -15 -20 -25 -30 -35 -40 1. (b) Coupling between the two ports.9 1. (a) Reflection coefficient at port 1.7 1.4 1.6 1. Electronic and Computer Engineering 137 . University of Pretoria—Electrical.1 2.33 S-parameters of the dual-polarised antenna element.4 1.0 (b) Figure 4. G (2004) Chapter 4 Numerical and Experimental Results IE3D Measured SDMM FEKO 0 -5 -10 P |S1.8 Frequency (GHz) 2.2 (a) IE3D Measured SDMM FEKO 0 -5 P |S1.0 2.2 | (dB) -10 -15 -20 -25 -30 1.University of Pretoria etd – Mayhew-Ridgers.9 1.7 1.8 Frequency (GHz) 2.1 2.5 1.2 2.5 1.

it is important to know how the separation distance between the two elements affects both the selfimpedance at each port and the mutual impedances between the two ports.2 . (7. than on the selfimpedance at each port. In order to design such a two-element array.35(a) shows P how the self-impedance at port 1.2 and Z2. while the (2. cancel out.36 shows the simulated Z-parameters of the two-element antenna array in Figure 4. For the entire-domain sinusoidal basis functions. (2.2).34.1 and Z2.1 and Z2.0) and (2. the size of the capacitor patches.9) are satisfied.2 can then be changed until (4. of 30 mm). is quite good.4. a 180◦ phase difference). varies with the separation distance. Figure 4. It can be seen that separation distance has a more profound effect on the mutual impedance between the two ports. Electronic and Computer Engineering 138 .35(b) shows how the mutual impedance. as well as the separation distance P P between them and the resonant patches.1) modes were used for the y-directed current.6) to (4. which was designed for an input impedance of 50 Ω at each port (this corresponds to a separation distance. is to use two elements that are positioned in a back-to-back configuration. University of Pretoria—Electrical. and four PWS basis functions on each probe (excluding the one associated with the attachment mode). Z1. Once again. as is the case in Section 4.1. The cross-polar levels in both planes fall well below the range of the graph. Z1. The resulting effect is that the co-polarised currents on the two resonant patches are aligned. It can be seen that the beamwidth in the E-plane is narrowed due to the alignment of the two elements.e.0). From the results it can be seen that the agreement between the measurements and all three codes.University of Pretoria etd – Mayhew-Ridgers. the antenna element can be designed through an iterative process. and therefore the input impedance at each port is a function of the self-impedance at that port as well as the mutual impedance between the two ports.4. the three codes show good agreement.3. (3. the two ports have to be excited exactly out of phase (i. The electric current density on the structure was modelled with eleven entire-domain sinusoidal basis functions on each resonant patch. between the two P patches.34. (3.2 .1 .2). which should be equal.0). the (1. Figure 4. Both these have been simulated with the SDMM.0).2). Z1. For specific values P P P P of Z1.1. As in Section 4. similar to the one in Section 4. Here.3 Numerical and Experimental Results Two-Element Linearly-Polarised Array An alternative way to reduce the cross-polarisation levels of a probe-fed patch antenna.0) and (9. (5.37 shows the simulated radiation patterns in the E-plane and the H-plane of the two-element antenna array.1 .4.0) modes were used for the x-directed current. but that the fields radiated by the cross-polarised currents. as shown in Figure 4. varies with the separation distance. The two-element array has to be excited by a feed network. s. (2. can be used to control Z1. Once again. G (2004) Chapter 4 4.0). (1. while Figure 4. s. Figure 4. the single higher-order circular attachment mode on each capacitor patch.

G (2004) Chapter 4 Numerical and Experimental Results y z θ 7 mm 1. (a) Top view of the array.35 Effect (at 1. Real Imag Real 100 Imag 60 80 40 P Z1.5 mm (a) (b) Figure 4. (b) Side view of the multilayered substrate with εr = 4.University of Pretoria etd – Mayhew-Ridgers.8 GHz) of the separation distance on the Z-parameters of the two-element linearly-polarised array.8 mm Connector 1. Electronic and Computer Engineering 139 .0 20.2 (Ohm) P Z1. (a) Self-impedance at port 1. (b) Mutual impedance between the two ports.0 (a) 40.0 (b) Figure 4.1 (Ohm) 60 40 20 0 0 -20 -40 20 -20 20.0 s (mm) 60.34 Geometry of the two-element linearly-polarised array.6 mm 52 mm φ s 52 mm FR-4 x 15 mm Air ø = 0.0 s (mm) 60. University of Pretoria—Electrical.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer.0 40.25 and tan δε = 0.

36 Z-parameters of the two-element linearly-polarised array (s = 30 mm).5 1.1 -60 -80 1.8 Frequency (GHz) 2.7 1.9 1.0 (b) Figure 4.2 0 1.9 1. University of Pretoria—Electrical.1 2. G (2004) Chapter 4 Numerical and Experimental Results IE3D Measured SDMM FEKO 100 80 Resistance (Ohm) P ℜe Z1.4 1.0 (a) IE3D Measured SDMM FEKO 20 Reactance (Ohm) 0 P ℑm Z1.2 2.1 60 40 20 P ℜe Z1.7 1.1 2. Electronic and Computer Engineering 140 .5 1.8 Frequency (GHz) 2.6 1. (b) Reactive part of the Z-parameters. (a) Resistive part of the Z-parameters.2 -20 -40 P ℑm Z1.University of Pretoria etd – Mayhew-Ridgers.2 2.6 1.4 1.

University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 141 .8 GHz) of the two-element linearly-polarised array (s = 30 mm).37 Radiation patterns (at 1. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 15 Relative gain (dB) 10 5 0 -5 -10 -15 -90 -60 -30 0 θ (degrees) 30 60 90 60 90 (a) FEKO IE3D SDMM 15 Relative gain (dB) 10 5 0 -5 -10 -15 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4. (a) Radiation pattern in the E-plane (φ = 0◦ ). University of Pretoria—Electrical. (b) Radiation pattern in the H-plane (φ = 90◦ ).

The feed network was modelled with Sonnet’s circuit analysis module. (7. G (2004) Chapter 4 4. (5.0). while it is also valid for FEKO when the default settings are used.1 Computer memory required for the storage of the interaction matrix associated with the four-element verticallypolarised antenna array. and for IE3D when the full matrix solver is used.0) and (2. The electric current density on the structure was modelled with eleven entire-domain sinusoidal basis functions on each resonant patch. the amount of duplicate entries in the interaction matrix equates to 67.1 shows how the computer-memory requirements. It can be seen that there is close agreement between the results of the FEKO and SDMM codes.88%.2). However. but that the reflection coefficient.1 It can be seen that the memory requirements of the SDMM are substantially lower then that of the other two codes.2). two of the probes are excited exactly out of phase with respect to the other two probes. An example of such an array is shown in Figure 4. (3.0). the (1. as shown in Table 4. It only gives an indication of the amount of matrix elements that are not calculated directly. is somewhat lower.e. it is assumed that all the codes store the entire interaction matrix (i. This assumption is always valid for the SDMM code.3 0.38. while the (2.40 shows the simulated radiation patterns in the E-plane and the H-plane of the antenna array. varies among the three codes. (1.39 shows the simulated reflection coefficient at the input port of the feed network. Note that the resonant patches are all equally separated.0 101.4 Numerical and Experimental Results Four-Element Vertically-Polarised Array Many applications require the use of linear vertically-polarised antenna arrays.2. This is done to reduce cross-polarisation levels. For the entire-domain sinusoidal basis functions. Table 4. Method Unknowns Memory (MB) IE3D FEKO SDMM 2560 2576 64 100. the 10 dB return-loss bandwidths for all three codes seem to be comparable.University of Pretoria etd – Mayhew-Ridgers.0) and (9. Also. University of Pretoria—Electrical. 2 The amount of duplicate entries within the interaction matrix makes no difference in the amount of matrix entries that are actually stored within the computer’s memory (the entire matrix is always stored).1) modes were used for the x-directed current.4. for the SDMM. but that are merely copied from other entries (i. for the storage of the interaction matrix. Figure 4. N 2 values) in the computer’s memory. The array was designed so that the input impedance at each port is equal to 50 Ω. Electronic and Computer Engineering 142 . The 1 For comparison. As can be seen from Figure 4. (2. the speedup in filling the matrix).e.38. (2. the single higher-order circular attachment mode on each capacitor patch and four PWS basis functions on each probe (excluding the one associated with the attachment mode). This particular one consists of a pair of two-element subarrays.0).0).2).0) modes were used for the y-directed current.063 Figure 4. as simulated by IE3D. (3.2 Table 4.

(c) Schematic of the feed network. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1.5 mm (b) 30 mm 52 mm 1∠180◦ 1∠0◦ 52 mm φ x Port 7 mm ø = 0.University of Pretoria etd – Mayhew-Ridgers.38 Geometry of the four-element vertically-polarised array.6 mm FR-4 Air 15 mm Connector 1.8 mm (a) (c) Figure 4.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer. Electronic and Computer Engineering 143 .25 and tan δε = 0. (a) Top view of the array. University of Pretoria—Electrical. (b) Side view of the multilayered substrate with εr = 4.

9 21. In terms of the radiation patterns.University of Pretoria etd – Mayhew-Ridgers.2 10 dB Return-loss bandwidth at the input port of the fourelement vertically-polarised antenna array. G (2004) Chapter 4 Numerical and Experimental Results IE3D SDMM FEKO 0 -5 P |S1.2 Figure 4. Table 4. Electronic and Computer Engineering 144 .1 | (dB) -10 -15 -20 -25 1.1 2. It can be seen that the beamwidth in the E-plane is narrowed due to the vertical alignment of the antenna elements.0 2.7 1. Method Bandwidth (%) IE3D FEKO SDMM 25. the three codes show very good agreement.6 1.4 1.5 1.39 Reflection coefficient at the input port of the feed network for the four-element vertically-polarised array.7 21.8 Frequency (GHz) 2.3 University of Pretoria—Electrical.9 1. cross-polar levels in both planes fall well below the range of the graph.

Electronic and Computer Engineering 145 . (b) Radiation pattern in the H-plane (φ = 0◦ ). (a) Radiation pattern in the E-plane (φ = 90◦ ).University of Pretoria etd – Mayhew-Ridgers.8 GHz) of the four-element vertically-polarised array. University of Pretoria—Electrical. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 15 10 Gain (dBi) 5 0 -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 60 90 (a) FEKO IE3D SDMM 15 10 Gain (dBi) 5 0 -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4.40 Radiation patterns (at 1.

Also. Figure 4. (7. (3.41. the single higher-order circular attachment mode on each capacitor patch.0) and (9. for the SDMM. It can be seen that the memory requirements of the SDMM are once again substantially lower then that of the other two codes. (2. the amount of duplicate entries in the interaction matrix equates to 83. from which it can be seen that the agreement is also quite good.4. Method Unknowns Memory (MB) IE3D FEKO SDMM 4216 2944 84 271. (3. for the storage of the interaction matrix.2). appears to be a fraction higher than that of the other two codes. The electric current density on the structure was modelled with eleven entire-domain sinusoidal basis functions on each resonant patch. four of the probes are excited exactly out of phase with respect to the other four probes. as simulated by IE3D. The antenna gain. An example of such an array is shown in Figure 4. the three codes show very good agreement.42 shows the simulated reflection coefficient at the input port of the feed network. In terms of the radiation patterns. The feed network was modelled with Sonnet’s circuit analysis module.2).43 shows the simulated radiation patterns in the E-plane and the H-plane of the antenna array. As can be seen from Figure 4. University of Pretoria—Electrical. each with two capacitor patches to reduce cross-polarisation levels. (2.2 132.0).11%. Electronic and Computer Engineering 146 .3 0. It can be seen that there is good agreement between the results of all three codes. This particular one consists of four elements.5 Numerical and Experimental Results Four-Element Horizontally-Polarised Array For some applications. varies among the three codes. (5.4. Table 4. and four PWS basis functions on each probe (excluding the one associated with the attachment mode).0).3 shows how the computer-memory requirements. G (2004) Chapter 4 4. For the entire-domain sinusoidal basis functions. It can also be seen that the beamwidth in the H-plane is narrowed due to the vertical alignment of the antenna elements. while the (2.1) modes were used for the y-directed current. a linear horizontally-polarised antenna array is more advantageous than vertically polarised ones. however.University of Pretoria etd – Mayhew-Ridgers.0) modes were used for the x-directed current.0).108 Figure 4. The array was designed so that the input impedance at each port is equal to 50 Ω.41.3 Computer memory required for the storage of the interaction matrix associated with the four-element horizontallypolarised antenna array. The cross-polar levels in both planes fall well below the range of the graph. Table 4. the (1.0).0) and (2. The simulated 10 dB return-loss bandwidths are shown in Table 4. (1.2).

University of Pretoria—Electrical. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1.41 Geometry of the four-element horizontally-polarised array. Electronic and Computer Engineering 147 .25 and tan δε = 0.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer.University of Pretoria etd – Mayhew-Ridgers.6 mm FR-4 Air 15 mm Connector 6. (b) Side view of the multilayered substrate with εr = 4.8 mm 30 mm (a) (c) Figure 4. (a) Top view of the array.5 mm 52 mm (b) φ 52 mm x 9 m m Port 1∠180◦ 1∠0◦ ø = 0. (c) Schematic of the feed network.

6 38.2 2.4 1. Method Bandwidth (%) IE3D FEKO SDMM 41.0 2. G (2004) Chapter 4 Numerical and Experimental Results IE3D SDMM FEKO 0 -5 P |S1.University of Pretoria etd – Mayhew-Ridgers.1 38.42 Reflection coefficient at the input port of the feed network for the four-element horizontally-polarised array. Table 4.7 1.4 10 dB Return-loss bandwidth at the input port of the fourelement horizontally-polarised antenna array.5 1.1 2.9 1.8 Frequency (GHz) 2. Electronic and Computer Engineering 148 .1 | (dB) -10 -15 -20 -25 1.5 University of Pretoria—Electrical.3 Figure 4.6 1.

University of Pretoria—Electrical. (a) Radiation pattern in the E-plane (φ = 0◦ ). (b) Radiation pattern in the H-plane (φ = 90◦ ).8 GHz) of the four-element horizontally-polarised array. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 15 10 Gain (dBi) 5 0 -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 60 90 (a) FEKO IE3D SDMM 15 10 Gain (dBi) 5 0 -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4.43 Radiation patterns (at 1.University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 149 .

To achieve these requirements. Table 4. The array was designed so that the input impedance at each port is equal to 50 Ω.4.5 129. G (2004) Chapter 4 4. Figure 4.0). (2. Like other dual-polarised antennas. while the same set of modes were also used for the −45◦ -directed current. For the entire-domain sinusoidal basis functions. Method Unknowns Memory (MB) IE3D FEKO SDMM 4076 2912 112 253. varies among the three codes. (5. a 180◦ phase difference) so that the copolarised fields add constructively and the cross-polarised fields cancel out. The input impedance at a specific probe is a function of the self-impedance at that probe as well as the mutual impedance between that probe and all the other probes.0) and (9.191 University of Pretoria—Electrical. and four PWS basis functions on each probe (excluding the one associated with the attachment mode). They are usually required to have a wide beamwidth in the azimuth plane and a narrow beamwidth in the elevation plane.e. It can be seen that the memory requirements of the SDMM are once again substantially lower then that of the other two codes. while the other two probes provide the orthogonal polarisation. Electronic and Computer Engineering 150 .0) modes were used for the +45◦ -directed current.0). the single higher-order circular attachment mode on each capacitor patch. the probes that are diagonally opposite to each other. Also. as well as the separation distance between them and the resonant patches.5 shows how the computer-memory requirements. (1. they are therefore usually also constructed as linear arrays. the (1. The array can be viewed as two subarrays.6) to (4.6 Numerical and Experimental Results Four-Element ±45◦ Slant-Polarised Array Linear antenna arrays with ±45◦ slant polarisation are nowadays very popular for use on cellular base stations. provide the one polarisation. these antennas also have two ports.4 0. (3. are excited exactly out of phase (i. (3. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on each resonant patch. This array was designed by taking into account only the mutual impedance between two diagonally opposite probes in each subarray.44 shows the geometry of such an array that can be realised by means of patch antenna elements with capacitive feed probes. The size of the capacitor patches. the amount of duplicate entries in the interaction matrix equates to 83. for the SDMM. one for the +45◦ polarisation and one for the −45◦ polarisation. (2.University of Pretoria etd – Mayhew-Ridgers.0).41%. for the storage of the interaction matrix. (7. Table 4. The feed network was modelled with Sonnet’s circuit analysis module.1).5 Computer memory required for the storage of the interaction matrix associated with the four-element ±45◦ slantpolarised antenna array.0). was then altered until (4.2). In each subarray.2).9) were satisfied for the two diagonally opposite probes. The two probes that are diagonally opposite to each other. each consisting of two dual-polarised elements.

5 m m Connector ø = 0.44 Geometry of the four-element ±45◦ slant-polarised array. G (2004) Chapter 4 Numerical and Experimental Results z y θ 1.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer.8 mm 8 mm (c) 1∠180◦ 1∠180◦ 1∠0◦ ◦ 1∠0 φ x m m 52 m 52 m 100 mm Port 2 Port 1 (a) (b) Figure 4. (a) Top view of the array. University of Pretoria—Electrical. Electronic and Computer Engineering 151 .University of Pretoria etd – Mayhew-Ridgers. (b) Side view of the multilayered substrate with εr = 4. (c) Schematic of the feed network.25 and tan δε = 0.6 mm FR-4 Air 15 mm 4.

φ) = respectively. φ) √ 2 (4. Overall though. The coupling. 196].46 shows the simulated co-polarised and cross-polarised radiation patterns in the azimuth and elevation planes of the antenna array.0 The gain associated with the co-polarised radiation patterns. This will also increase the gain of the array. Electronic and Computer Engineering 152 . φ) scat scat Eθ (θ.10) 2η(0) Pin and Gxpol (θ. as well as the gain associated with the cross-polarised radiation pattern. φ) = scat 4π Ecopol (θ. The elevation beamwidth can be decreased by adding more elements to the array.8 27. the impedance bandwidth of the array seems to be satisfactory for most cellular applications (the array was designed to roughly operate within the GSM 1800 frequency band. For the reflection coefficient.45(a) shows the simulated reflection coefficient at port 1 of the feed network. The simulated 10 dB return-loss bandwidths for port 1 are shown in Table 4. Ecopol and Expol . are given by scat Ecopol (θ. Table 4. φ) + Eφ (θ. (4. while all three codes compare fairly well in terms of the coupling results. University of Pretoria—Electrical.6 10 dB Return-loss bandwidth at port 1 of the four-element ±45◦ slant-polarised antenna array. φ) = 2 scat 4π Expol (θ. From these it can be seen that the agreement between the three codes is very good. but was not optimised as such). φ) − Eφ (θ. Gxpol . G (2004) Chapter 4 Numerical and Experimental Results Figure 4.12) scat scat Eθ (θ. while Figure 4. Figure 4. This should suffice for most cellular applications [195. φ) √ = 2 (4. can be calculated as Gcopol (θ.13) and scat Expol (θ. as simulated by IE3D. φ) (4. and that the array has a cross-polar discrimination factor of about 18 dB to 20 dB.11) scat scat where the absolute values of the co-polarised and cross-polarised electric fields.6. Gcopol . φ) 2η(0) Pin 2 .8 27. 195]. Method Bandwidth (%) IE3D FEKO SDMM 29. is slightly higher than as simulated by the other two codes.University of Pretoria etd – Mayhew-Ridgers. while the decoupling of the two ports is close to the 25 dB to 30 dB that is normally used for cellular applications [194.45(b) shows the simulated coupling between the two ports of the feed network. the results of the SDMM lies between those of IE3D and FEKO.

(a) Reflection coefficient at port 1 of the feed network.6 1. University of Pretoria—Electrical.0 2.University of Pretoria etd – Mayhew-Ridgers.8 Frequency (GHz) 2.4 1.2 2.2 | (dB) -20 -25 -30 -35 -40 1.1 2.7 1.45 S-parameters of the four-element ±45◦ slant-polarised array.0 (b) Figure 4. Electronic and Computer Engineering 153 .8 Frequency (GHz) 2.7 1.1 | (dB) -10 -15 -20 -25 -30 -35 1.5 1. (b) Coupling between the two ports of the feed network.9 1. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 0 -5 P |S1.1 2.4 1.5 1.9 1.6 1.2 (a) FEKO IE3D SDMM -10 -15 P |S1.

46 Radiation patterns (at 1. (a) Radiation pattern in the azimuth plane (φ = 0◦ ).8 GHz) of the four-element ±45◦ slant-polarised array. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 15 10 Co-polar Gain (dBi) 5 0 Cross-polar -5 -10 -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 (a) FEKO IE3D SDMM 15 10 5 Gain (dBi) Co-polar 0 -5 -10 Cross-polar -15 -20 -90 -60 -30 0 θ (degrees) 30 60 90 (b) Figure 4.University of Pretoria etd – Mayhew-Ridgers. (b) Radiation pattern in the elevation plane (φ = 0◦ ). University of Pretoria—Electrical. Electronic and Computer Engineering 154 .

Of the five resonant patches. Also. It consists of a zig-zag type of arrangement where adjacent antenna elements are positioned in a back-to-back configuration. from which it can be seen that the agreement between the three codes is very good.48. (2. the single higher-order circular attachment mode on each capacitor patch. (5.3. varies among the three full-wave codes. University of Pretoria—Electrical. are associated with the orthogonal polarisation. are associated with the one polarisation.4. An physical model of this antenna.0) and (9. but that the three patches in the middle provide both polarisations. the (1. It can therefore be seen that the topmost and bottommost patches only provide one polarisation each.2). and four PWS basis functions on each probe (excluding the one associated with the attachment mode).9 0. Table 4. while those connected to port 2. while the top four patches provide the orthogonal polarisation. All of the probes connected to port 1. Electronic and Computer Engineering 155 . It can be seen that the memory requirements of the SDMM are once again substantially lower then that of the other two codes.4. (3.0).0). for the storage of the interaction matrix associated with the array. G (2004) Chapter 4 4. (2.2).7 Numerical and Experimental Results Five-Element ±45◦ Slant-Polarised Array An alternative form of the ±45◦ slant-polarised antenna array is shown in Figures 4.0).1).6 183. The feed network was modelled with IE3D.47.42%. the bottom four patches provide the one polarisation. (7. (1. Table 4.49 shows the simulated and measured gain (all cases including losses in the feed network) of the antenna.7 shows how the computer-memory requirements. for the SDMM. was also constructed in order to compare the simulations to measurements. By following the same approach as in Section 4. (3. Method Unknowns Memory (MB) IE3D FEKO SDMM 5682 3472 130 492. This can be ascribed to the rather small ground plane that was used for the physical antenna model. but that the measured gain is only slightly lower than the simulated gain. For the entire-domain sinusoidal basis functions. the input impedance at each port was designed to be equal to 50 Ω.University of Pretoria etd – Mayhew-Ridgers.258 Figure 4.0). as shown in Figure 4. after which the S-parameters of the array and the feed network were cascaded by means of Sonnet’s circuit analysis module. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on each resonant patch. the amount of duplicate entries in the interaction matrix equates to 82. while the same set of modes were also used for the −45◦ -directed current. The antenna array was designed by only taking into account the mutual coupling between two adjacent probes.0) modes were used for the +45◦ -directed current.7 Computer memory required for the storage of the interaction matrix associated with the five-element ±45◦ slantpolarised antenna array.

while Figure 4.50(a) shows the measured and simulated reflection coefficient at port 1 of the feed network. As far as the coupling between the two ports goes. Electronic and Computer Engineering 156 . There is a small difference between the measured/IE3D results and the SDMM/FEKO results. Method Measured IE3D FEKO SDMM Bandwidth (%) 28.University of Pretoria etd – Mayhew-Ridgers.8 10 dB Return-loss bandwidth at port 1 of the five-element ±45◦ slant-polarised antenna array. In this case it can be seen that the beamwidth of the co-polarised pattern in the azimuth plane is slightly narrower than that of the array in Section 4. The cross-polar discrimination of this array is better than 20 dB over most of the usable beamwidth in the two planes.4. From these it can be seen that the agreement between the three codes is quite good and that the simulations compare fairly well with the measurements. G (2004) Chapter 4 Numerical and Experimental Results Figure 4. The simulated 10 dB return-loss bandwidths for port 1 are shown in Table 4. University of Pretoria—Electrical. however.6.7 Figure 4. The overall bandwidth. while Figure 4.4 30.0 28. the measured results and simulated results all agree fairly well. In terms of the reflection coefficient.52 shows the measured and simulated co-polarised and cross-polarised radiation patterns in the elevation plane of the antenna array. given the finite size of the ground plane that was used for the physical antenna model.4 32.50(b) shows the measured and simulated coupling between the two ports of the feed network. Table 4. This is due to the somewhat larger horizontal dimensions of the array. From these results it can be seen that the impedance bandwidth of the antenna is sufficiently wide for most cellular applications and that the decoupling between the two ports seem to be slightly better than that of the array in Section 4.4. appears to be more or less the same for all four sets of results.8.51 shows the measured and simulated co-polarised and cross-polarised radiation patterns in the azimuth plane of the antenna array.6. as well as close agreement between the SDMM results and the FEKO results. there is close agreement between the measured results and the IE3D results.

University of Pretoria—Electrical.6 mm FR-4 15 mm Air FR-4 5 Feed line m m (b) Z0 = 50 Ω 1∠180◦ m 52 m m 30 m Z0 = 35 Ω 1∠0◦ φ Z0 = 50 Ω x Z0 = 35 Ω ø = 0.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer.University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 157 . (b) Side view of the multilayered substrate with εr = 4. (c) Feed network for the array.25 and tan δε = 0. G (2004) Chapter 4 Numerical and Experimental Results z y θ 1. (a) Top view of the array.47 Geometry of the five-element ±45◦ slant-polarised array.8 mm 52 m m Port 2 Z0 = 50 Ω 7 m m Port 1 (a) (c) Figure 4.

University of Pretoria etd – Mayhew-Ridgers.6 1.4 1. (a) Antenna elements.5 1.0 2.2 Figure 4. IE3D Measured SDMM FEKO 16 14 12 Gain (dBi) 10 8 6 4 2 0 1.8 Frequency (GHz) 2.48 Photographs of the five-element ±45◦ slant-polarised array. Electronic and Computer Engineering 158 .1 2.9 1.7 1. G (2004) Chapter 4 Numerical and Experimental Results (a) (b) Figure 4. (b) Feed network.49 Gain of the five-element ±45◦ slant-polarised array. University of Pretoria—Electrical.

6 1.7 1.1 2.1 2.1 | (dB) -10 -15 -20 -25 -30 1.8 Frequency (GHz) 2.8 Frequency (GHz) 2. G (2004) Chapter 4 Numerical and Experimental Results Measured IE3D FEKO SDMM 0 -5 P |S1.6 1.5 1.4 1. (a) Reflection coefficient at port 1 of the feed network. University of Pretoria—Electrical.50 S-parameters of the five-element ±45◦ slant-polarised array.3 (a) Measured IE3D FEKO SDMM -10 -15 -20 P |S1.0 2.9 1.9 1.0 2.2 2. (b) Coupling between the two ports of the feed network. Electronic and Computer Engineering 159 .3 (b) Figure 4.7 1.2 | (dB) -25 -30 -35 -40 -45 -50 1.4 1.University of Pretoria etd – Mayhew-Ridgers.5 1.2 2.

University of Pretoria—Electrical. (a) Co-polarised pattern.51 Radiation patterns (at 1.8 GHz) in the azimuth plane (φ = 0◦ ) of the five-element ±45◦ slant-polarised array. Electronic and Computer Engineering 160 . (b) Cross-polarised pattern. G (2004) Chapter 4 Numerical and Experimental Results Measured IE3D FEKO SDMM 0 Relative gain (dB) -5 -10 -15 -20 -25 -30 -90 -60 -30 30 0 θ (degrees) 60 90 (a) Measured IE3D FEKO SDMM -10 Relative gain (dB) -15 -20 -25 -30 -35 -40 -90 -60 -30 0 θ (degrees) 30 60 90 (b) Figure 4.University of Pretoria etd – Mayhew-Ridgers.

52 Radiation patterns (at 1. University of Pretoria—Electrical.8 GHz) in the elevation plane (φ = 90◦ ) of the five-element ±45◦ slant-polarised array. (b) Cross-polarised pattern. Electronic and Computer Engineering 161 . G (2004) Chapter 4 Numerical and Experimental Results Measured IE3D FEKO SDMM 0 Relative gain (dB) -5 -10 -15 -20 -25 -30 -90 -60 -30 30 0 θ (degrees) 60 90 (a) Measured IE3D FEKO SDMM -10 Relative gain (dB) -15 -20 -25 -30 -35 -40 -90 -60 -30 0 θ (degrees) 30 60 90 (b) Figure 4.University of Pretoria etd – Mayhew-Ridgers. (a) Co-polarised pattern.

while Figure 4. the (1.0). (2. (7.1). (3.0) and (9.2). As far as the coupling between the two ports goes. It is a somewhat larger version of the one in Section 4.9 Computer memory required for the storage of the interaction matrix associated with the nine-element ±45◦ slantpolarised antenna array. (2.07%. and four PWS basis functions on each probe (excluding the one associated with the attachment mode). For the entire-domain sinusoidal basis functions. This can be achieved by adding more elements to the antenna array. In terms of the reflection coefficient. an antenna array is presented that is more or less comparable in size to that of a typical high-gain cellular base station antenna.4. (1. the single higher-order circular attachment mode on each capacitor patch.0). The simulated 10 dB return-loss bandwidths for port 1 are shown in Table 4. the amount of duplicate entries in the interaction matrix equates to 90. For the SDMM. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on each resonant patch. University of Pretoria—Electrical. From these results it can be seen that the impedance bandwidth of the antenna is sufficiently wide for most cellular applications and that the decoupling between the two ports is also quite good. (3.9 shows how the computer-memory requirements. the simulated results all agree fairly well. G (2004) Chapter 4 4. In this section.8 582. (5.10. but that those for the other two codes are greater than the available memory on most personal computers. while the same set of modes were also used for the −45◦ -directed current.54(a) shows the simulated reflection coefficient at port 1 of the feed network.0). thereby extending the length of the array. there is close agreement between the SDMM results and the FEKO results.4. for the storage of the interaction matrix associated with the array. The overall bandwidth. but there is a slight frequency shift when compared to the IE3D results. appears to be more or less the same for all three sets of results.0). It can be seen that the memory requirements of the SDMM are basically insignificant.8 Numerical and Experimental Results Nine-Element ±45◦ Slant-Polarised Array Many cellular base station installations require antennas with higher gains and narrower elevation beamwidths than those that have been presented to far. The feed network was modelled with Sonnet’s circuit analysis module. however. Table 4.0 0.7. varies among the three full-wave codes.University of Pretoria etd – Mayhew-Ridgers.54(b) shows the simulated coupling between the two ports of the feed network.53.2). Electronic and Computer Engineering 162 .0) modes were used for the +45◦ -directed current. Consider the antenna array in Figure 4. Table 4.894 Figure 4. Method Unknowns Memory (MB) IE3D FEKO SDMM 7747 6176 242 915.

University of Pretoria etd – Mayhew-Ridgers.4 Figure 4.5 19. Electronic and Computer Engineering 163 . G (2004) Chapter 4 Numerical and Experimental Results Table 4.10 10 dB Return-loss bandwidth at port 1 of the nineelement ±45◦ slant-polarised antenna array. From these it can be seen that the agreement between the three codes is quite good.8 20. University of Pretoria—Electrical. Method Bandwidth (%) IE3D FEKO SDMM 20. The cross-polar discrimination of this array is also better than 20 dB over most of the usable beamwidth in the two principal planes. It can also be seen that the extra elements have raised the gain of the antenna and that the beamwidth in the elevation plane has been narrowed.55 shows the simulated co-polarised and cross-polarised radiation patterns in the azimuth and elevation planes of the antenna array.

25 and tan δε = 0.53 Geometry of the nine-element ±45◦ slant-polarised array.02 for the FR-4 layers and with εr = 1 and tan δε = 0 for the air layer. (c) Schematic of the feed network. Electronic and Computer Engineering 164 . (b) Side view of the multilayered substrate with εr = 4.8 mm 52 m m 7 m m φ x 1∠180◦ Port 2 1∠0◦ Port 1 (a) (c) Figure 4.6 mm 5 m m FR-4 Air 15 mm Connector 52 m m (b) m m 30 ø = 0. G (2004) Chapter 4 Numerical and Experimental Results y z θ 1. University of Pretoria—Electrical.University of Pretoria etd – Mayhew-Ridgers. (a) Top view of the array.

0 2.University of Pretoria etd – Mayhew-Ridgers.8 Frequency (GHz) 2.2 (a) FEKO IE3D SDMM -10 -15 -20 P |S1. (a) Reflection coefficient at port 1 of the feed network.9 1.1 2.1 | (dB) -10 -15 -20 -25 1.4 1. Electronic and Computer Engineering 165 .1 2.54 S-parameters of the nine-element ±45◦ slant-polarised array.7 1.6 1.8 Frequency (GHz) 2.0 (b) Figure 4.5 1.4 1.9 1.2 | (dB) -25 -30 -35 -40 -45 -50 1.5 1. G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 0 -5 P |S1.2 2. (b) Coupling between the two ports of the feed network. University of Pretoria—Electrical.6 1.7 1.

(b) Radiation pattern in the elevation plane (φ = 0◦ ). G (2004) Chapter 4 Numerical and Experimental Results FEKO IE3D SDMM 20 10 Gain (dBi) Co-polar 0 -10 Cross-polar -20 -30 -90 -60 -30 0 θ (degrees) 30 60 90 60 90 (a) FEKO IE3D SDMM 20 10 Gain (dBi) Co-polar 0 Cross-polar -10 -20 -30 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4. University of Pretoria—Electrical. Electronic and Computer Engineering 166 .55 Radiation patterns (at 1.8 GHz) of the nine-element ±45◦ slant-polarised array.University of Pretoria etd – Mayhew-Ridgers. (a) Radiation pattern in the azimuth plane (φ = 0◦ ).

1 GB (estimated) 14. are positioned next to each other in order to form an array that is suitable for adaptive beamforming in the azimuth plane. As such.53.11 illustrates how the computer-memory requirements of the SDMM would compare to that of IE3D and FEKO. Method Unknowns Memory IE3D FEKO SDMM 30988 24704 968 14. The geometry of the array is shown in Figure 4. (1. the single higher-order circular attachment mode on each capacitor patch. and to connect the input/output ports of the individual arrays by means of a suitable beamforming network [194.1). Table 4. G (2004) Chapter 4 4.0) modes were used for the +45◦ -directed current. In this section. (3. Essentially.56.0) and (9. Due to the size of the antenna array. it was only modelled with the SDMM.4.0).3 MB Figure 4. For the entire-domain sinusoidal basis functions. A common way to realise an adaptive antenna. for a non-iterative solution. (3.57 shows the coupling between the different subarrays. the antenna effectively suppresses interference that comes from other directions. (2. and four PWS basis functions on each probe (excluding the one associated with the attachment mode).0). similar to the one in Figure 4. (5. Electronic and Computer Engineering 167 .11 Computer memory required for the storage of the interaction matrix associated with the thirty six-element ±45◦ slant-polarised antenna array. also called smart antennas. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on each resonant patch. while the same set of modes were also used for the −45◦ -directed current. is to place a number of linear antenna arrays next to each other. it is an antenna array of which a relatively narrow beam can be steered towards the mobile subscriber.53. the (1. (2.4.9 Numerical and Experimental Results Thirty Six-Element ±45◦ Slant-Polarised Array The concept of adaptive antenna systems. which makes extensive use of entire-domain basis functions. as presented in Section 4. It can be used to increase the capacity of cellular communications systems.3 GB (estimated) 9. (7. It can be seen that. 197].University of Pretoria etd – Mayhew-Ridgers. has lately been a topic of extensive research.0). are exactly the same as those in Figure 4.55%. Each subarray also has a feed network with two ports. This includes coupling between University of Pretoria—Electrical.2). For the SDMM. The feed networks were modelled with Sonnet’s circuit analysis module. four of the nine-element ±45◦ slant-polarised arrays. The dimensions of the antenna elements. Table 4.0).8. it is required that these arrays should only be able to scan in the azimuth plane [198]. as well as the spacings between them.2). Usually. the amount of duplicate entries in the interaction matrix equates to 97. the memory requirements of the two commercial codes would be greater than the available memory on most personal computers.

As shown in Figure 4. it can be seen that the coupling between all ports in the array is relatively low. as well as coupling between cross-polar ports (e.g.57(a) shows the coupling between subarray 1 and all of the other subarrays. University of Pretoria—Electrical. It can be seen that the horizontal distribution of the subarrays effectively narrows the beamwidth in the azimuth plane. In most cases the coupling is below 30 dB. while subarray 2 represents a subarray in the middle of the array. Figure 4. two +45◦ ports). Electronic and Computer Engineering 168 .University of Pretoria etd – Mayhew-Ridgers. while Figure 4. subarray 1 represents a subarray at the edge of the array. a +45◦ port and a −45◦ port).57(b) shows the coupling between subarray 2 and all of the other subarrays. From these two figures.g.56. The cross-polar discrimination of this array is in the order of 25 dB over most of the usable beamwidth in the two principal planes. G (2004) Chapter 4 Numerical and Experimental Results co-polar ports (e.58 shows the simulated co-polarised and cross-polarised radiation patterns in the azimuth and elevation planes of the antenna array when all the co-polar ports are driven with equal amplitude and phase. Figure 4.

University of Pretoria etd – Mayhew-Ridgers. University of Pretoria—Electrical. G (2004) Chapter 4 Numerical and Experimental Results y φ x 1 2 3 4 Figure 4. This array consists of four linear subarrays that are positioned next to each other.56 Geometry of the thirty six-element ±45◦ slant-polarised array. Electronic and Computer Engineering 169 .

University of Pretoria etd – Mayhew-Ridgers. (b) Coupling to subarray 2. G (2004) Chapter 4 Numerical and Experimental Results Co-polar Cross-polar 0 -10 P |S1. (a) Coupling to subarray 1.57 Co-polar and cross-polar coupling (at 1.j | (dB) -20 -30 -40 -50 1 3 2 4 Subarray j (b) Figure 4.j | (dB) -20 -30 -40 -50 1 3 2 4 Subarray j (a) Co-polar Cross-polar 0 -10 P |S2. Electronic and Computer Engineering 170 . University of Pretoria—Electrical.8 GHz) between the linear subarrays of the thirty six-element ±45◦ slant-polarised array.

8 GHz) of the thirty six-element ±45◦ slant-polarised array when the co-polar ports are excited with equal amplitude and phase.University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 171 .58 Radiation patterns (at 1. University of Pretoria—Electrical. (a) Radiation pattern in the azimuth plane (φ = 0◦ ). G (2004) Chapter 4 Numerical and Experimental Results Co-polar Cross-polar 30 20 Gain (dBi) 10 0 -10 -20 -30 -90 -60 -30 30 0 θ (degrees) 60 90 60 90 (a) Co-polar Cross-polar 30 20 Gain (dBi) 10 0 -10 -20 -30 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4. (b) Radiation pattern in the elevation plane (φ = 0◦ ).

two +45◦ ports). This includes coupling between co-polar ports (e. and four PWS basis functions on each probe (excluding the one associated with the attachment mode). Figure 4. for a non-iterative solution.0). In order to illustrate the computational benefits of the SDMM for such large problems. each one being ±45◦ slant polarised. (3.59. (2.5 GB (estimated) 76. Method Unknowns Memory IE3D FEKO SDMM 81520 58240 2240 99. Table 4.56. For the entire-domain sinusoidal basis functions.0 GB (estimated) 50. Electronic and Computer Engineering 172 . one for each polarisation.2). while Figure 4.6 MB Figure 4.2).60(a) shows the coupling between subarray 1 and all of the other subarrays. due to the size of the antenna array. it was only modelled with the SDMM. are exactly the same as those in Figure 4. (7. The feed networks were modelled with Sonnet’s circuit analysis module. (2. For the SDMM. subarray 1 represents a subarray at the edge of the array. the amount of duplicate entries in the interaction matrix equates to 98. the (1.4. the single higher-order circular attachment mode on each capacitor patch. The electric current density on the structure was modelled with eighteen entire-domain sinusoidal basis functions on each resonant patch.University of Pretoria etd – Mayhew-Ridgers.0) modes were used for the +45◦ -directed current.g. As shown in Figure 4.g. as well as coupling between cross-polar ports (e. as well as the spacings between them. Table 4. (5. The geometry of the array is shown in Figure 4. while subarray 5 represents a subarray in the middle of the array. Once again.59.44.12 illustrates how the computer-memory requirements of the SDMM would compare to that of IE3D and FEKO. It can be seen that. G (2004) Chapter 4 4. From these two figures.12 Computer memory required for the storage of the interaction matrix associated with the eighty-element ±45◦ slant-polarised antenna array. The dimensions of the antenna elements. a +45◦ port and a −45◦ port).0).0). it can be seen that the coupling between the subarrays is substantially higher than the coupling for the configuration in Figure 4.0). Each subarray also has a feed network with two ports. University of Pretoria—Electrical.1).60(b) shows the coupling between subarray 5 and all of the other subarrays.60 shows the coupling between some of the subarrays. while the same set of modes were also used for the −45◦ -directed current. (3. (1.0) and (9.10 Numerical and Experimental Results Eighty-Element ±45◦ Slant-Polarised Array Lindmark [194] has shown how up to twelve subarrays can be used to realise an adaptive antenna. this section presents an antenna array that consists of ten linear subarrays. the memory requirements of the two commercial codes would far outweigh the available memory on any personal computer.89%.

Electronic and Computer Engineering 173 . G (2004) Chapter 4 Numerical and Experimental Results y φ x 1 2 3 4 5 6 7 8 9 10 Figure 4.University of Pretoria etd – Mayhew-Ridgers.59 Geometry of the eighty-element ±45◦ slant-polarised array. the cross-polar discrimination of this array is not good at all. University of Pretoria—Electrical. the zig-zag configuration of Figure 4. It appears to be only about 10 dB.61 shows the simulated co-polarised and cross-polarised radiation patterns in the azimuth and elevation planes of the antenna array when all the co-polar ports are driven with equal amplitude and phase. In terms of these requirements. Figure 4. as well as low cross-polarisation levels. is required. It can be seen that the ten subarrays narrows the beamwidth in the azimuth plane significantly. This array consists of ten linear subarrays that are positioned next to each other. However. The results in this section and the previous section show that the layout of the antenna elements is very important when good isolation between subarrays.56 is definitely superior.

(b) Coupling to subarray 5. University of Pretoria—Electrical. (a) Coupling to subarray 1.8 GHz) between the linear subarrays of the eighty-element ±45◦ slant-polarised array. Electronic and Computer Engineering 174 .60 Co-polar and cross-polar coupling (at 1.j | (dB) -20 -30 -40 -50 1 2 3 4 7 6 5 Subarray j 8 9 10 8 9 10 (a) Co-polar Cross-polar 0 -10 P |S5.j | (dB) -20 -30 -40 -50 1 2 3 4 6 5 Subarray j 7 (b) Figure 4. G (2004) Chapter 4 Numerical and Experimental Results Co-polar Cross-polar 0 -10 P |S1.University of Pretoria etd – Mayhew-Ridgers.

University of Pretoria etd – Mayhew-Ridgers. Electronic and Computer Engineering 175 . (b) Radiation pattern in the elevation plane (φ = 0◦ ). University of Pretoria—Electrical. (a) Radiation pattern in the azimuth plane (φ = 0◦ ).8 GHz) of the eighty-element ±45◦ slant-polarised array when the co-polar ports are excited with equal amplitude and phase.61 Radiation patterns (at 1. G (2004) Chapter 4 Numerical and Experimental Results Co-polar Cross-polar 30 20 Gain (dBi) 10 0 -10 -20 -30 -90 -60 -30 30 0 θ (degrees) 60 90 60 90 (a) Co-polar Cross-polar 30 20 Gain (dBi) 10 0 -10 -20 -30 -90 -60 -30 0 θ (degrees) 30 (b) Figure 4.

Table 4. Table 4. The antenna elements in Figures 4.64 shows the measured and simulated radiation patterns in the E-plane and the H-plane of the antenna element with a circular resonant patch. it is possible to use alternative shapes for the resonant patches. Circular Measured IE3D Annular ring 27.13 shows the measured and simulated 10 dB return-loss bandwidths of the two antenna elements. only antenna elements with rectangular resonant patches have been investigated.0 dBi University of Pretoria—Electrical.65 shows the same for the the antenna element with a annular-ring resonant patch.5 Numerical and Experimental Results ALTERNATIVELY-SHAPED PATCHES So far. Circular Measured IE3D Annular ring 8. Electronic and Computer Engineering 176 . are the circular patch and annular-ring patch. as the SDMM implementation in this thesis only applies to antenna elements with rectangular resonant patches.14 Measured and simulated gain values of the antenna elements with alternatively-shaped patches.6 dBi 8. G (2004) Chapter 4 4.63 depict the geometry of two antenna elements that were designed to operate at a centre frequency of more or less 1.9% 26. This happens to be the case even when using thin substrates.14. Both these antenna elements have impedance loci that are very similar to that of the antenna element with the rectangular resonant patch.9% Figure 4. while Figure 4.University of Pretoria etd – Mayhew-Ridgers. The corresponding gain values are shown in Table 4. Figures 4.8% 26.8 GHz.63 were analysed with IE3D.1% 25. a physical model of each antenna element was constructed on a 150 mm × 150 mm ground plane. For verification purposes.5 dBi 8. Table 4. while they are also comparable to that of the antenna element with a rectangular resonant patch.62 and 4. The annular ring was designed to operate in its TM11 mode.13 Measured and simulated 10 dB return-loss bandwidths of the antenna elements with alternatively-shaped patches.8 dBi 8. All the measured and simulated results compare favourably. while the radiation patterns and gain values of the two elements are also very similar. Two shapes that have been found to be useful. The capacitive feed probe is an ideal feeding mechanism for the annular ring as it is impossible to obtain an impedance match for the annular ring in its TM11 mode when using a direct feed to the ring. However. The measured and simulated bandwidth values compare very well.62 and 4.

G (2004) Chapter 4 Numerical and Experimental Results PSfrag y z θ 5 × 12 mm 32 mm 1.6 mm φ FR-4 x 15 mm Air ø = 0.25 and tan δε = 0. y z θ 30 mm 5 × 10 mm 1.63 Geometry of the antenna element with an annular-ring resonant patch. (b) Side view of the multilayered substrate with εr = 4.25 and tan δε = 0. (a) Top view of the antenna element.6 mm φ FR-4 x 15 mm Air ø = 0. Electronic and Computer Engineering 177 . (b) Side view of the multilayered substrate with εr = 4.9 mm Connector 8 mm (b) (a) Figure 4.62 Geometry of the antenna element with a circular resonant patch.9 mm 11 mm Connector 9 mm (a) (b) Figure 4.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer.University of Pretoria etd – Mayhew-Ridgers.02 for the FR-4 layer and with εr = 1 and tan δε = 0 for the air layer. University of Pretoria—Electrical. (a) Top view of the antenna element.

University of Pretoria etd – Mayhew-Ridgers. (b) Co-polar and cross-polar radiation patterns in the H-plane (φ = 90◦ ).8 GHz) of the antenna element with a circular resonant patch.64 Radiation patterns (at 1. G (2004) Chapter 4 Numerical and Experimental Results Measured IE3D 0 Relative gain (dB) -10 -20 -30 -40 -180 -120 -60 60 0 θ (degrees) 120 180 (a) Measured IE3D 0 Co-polar Cross-polar Relative gain (dB) -10 -20 -30 -40 -180 -120 -60 0 θ (degrees) 60 120 180 (b) Figure 4. University of Pretoria—Electrical. (a) Co-polar radiation pattern in the E-plane (φ = 0◦ ). Electronic and Computer Engineering 178 .

University of Pretoria—Electrical.8 GHz) of the antenna element with an annular-ring resonant patch. Electronic and Computer Engineering 179 .65 Radiation patterns (at 1.University of Pretoria etd – Mayhew-Ridgers. (b) Co-polar and cross-polar radiation patterns in the H-plane (φ = 90◦ ). (a) Co-polar radiation pattern in the E-plane (φ = 0◦ ). G (2004) Chapter 4 Numerical and Experimental Results Measured IE3D 0 Relative gain (dB) -10 -20 -30 -40 -180 -120 -60 60 0 θ (degrees) 120 180 (a) Measured IE3D 0 Co-polar Cross-polar Relative gain (dB) -10 -20 -30 -40 -180 -120 -60 0 θ (degrees) 60 120 180 (b) Figure 4.

The rectangular attachment mode was unable to model the electric current density on very small probe-fed patches. The circular and rectangular attachment-mode approaches were weighed up against each other in terms of their abilities to model rectangular probe-fed patches of various sizes and on various substrate thicknesses. are the two important parameters for controlling the input impedance of the antenna element. it was shown that impedance bandwidths of well over 30% can be achieved for a VSWR of 2:1. The size of the capacitor patch has a more profound effect on the input reactance. measurements and the results of other commercial codes. G (2004) Chapter 4 4. It was also shown how the antenna element can be used in a number of different applications. Some of the arrays that were analysed. Also. The higher-order circular attachment mode proved to be quite successful for the modelling of the electric current density on the circular capacitor patches. for a rather thin air substrate. after which it would appear that the gap width between the capacitor patch and the resonant element becomes too large. but only to a certain point. are electrically large and served as good examples to show how efficient the SDMM implementation is in terms of computer-memory requirements. the SDMM implementation can be used University of Pretoria—Electrical. linear horizontally-polarised arrays. while the gap width largely affects the input resistance. The SDMM implementation was verified by first analysing isolated parts of the antenna elements. consisting of a resonant patch that is excited by a small probe-fed capacitor patch. On the high end. This knowledge is very important in the design of such elements or antenna arrays consisting of such elements. in order to verify that the SDMM implementation is accurate. the capacitor patch cannot be placed close enough to the resonant patch. For a fixed amount of computer memory. therefore also placing a lower limit on the impedance bandwidth. It was shown that the size of the capacitor and the gap width between the capacitor patch and the resonant patch. These are often required in applications such as cel- lular communications. Electronic and Computer Engineering 180 .University of Pretoria etd – Mayhew-Ridgers. Most importantly. as well as a number of ±45◦ slant-polarised arrays. Throughout. It turned out that the circular attachment mode is more versatile than the rectangular attachment mode. The examples included linear vertically-polarised arrays. to characterise the new antenna elements and to show how they can be used for different applications. the SDMM results compared well with published results. the SDMM implementation also proved to be accurate for the analysis of complete antenna elements. The bandwidth of the antenna element increases together with the thickness of the air substrate. The new antenna elements were characterised in order to show how the various dimensions of the structure affects the input impedance and impedance bandwidth of the antenna element. such as the probes and probe-fed capacitor patches. such as those that are used for the capacitor patches.6 Numerical and Experimental Results CONCLUDING REMARKS This chapter presented a number of numerical and experimental results. The results compares favourably to published data.

it turned out that the time required to evaluate the relevant entries in the interaction matrix. also has a linear relation with the number of unknowns.67. Once again. this is based on the problems that were analysed in this chapter. One feature of the SDMM implementation that proved to be very successful. University of Pretoria—Electrical. The relation in Figure 4. G (2004) Chapter 4 Numerical and Experimental Results to analyse much larger arrays than what is possible with the commercial codes. it can be seen that.66 is based on the problems that were solved in this chapter and is of course very dependant on the antenna geometry and types of basis functions that are used. Figure 4. Without this feature.66 shows a typical relation between the number of unknowns and the number of entries in the interaction matrix that has to be evaluated after all the duplicate entries have been identified. in general. it would appear that there is a linear relation between the number of entries to be evaluated and the number of unknowns.University of Pretoria etd – Mayhew-Ridgers. Throughout all the applications that were presented. the number of entries to be evaluated would be proportional to the square of the number of unknowns. is the identification of duplicate entries within the interaction matrix. the simulation results of the SDMM also compared well with measurements and the results of the two commercial codes that were used. As shown in Figure 4. However. Electronic and Computer Engineering 181 . This is done so that these entries do not need to be evaluated more than once.

University of Pretoria etd – Mayhew-Ridgers.67 Time required to fill the interaction matrix of the SDMM on a 2. Time (s) to fill interaction matrix 50000 40000 30000 20000 10000 0 500 1500 1000 Number of unknowns 2000 2500 Figure 4. University of Pretoria—Electrical. Electronic and Computer Engineering 182 . G (2004) Chapter 4 Numerical and Experimental Results 30000 Entries to evaluate 25000 20000 15000 10000 5000 0 500 1500 1000 Number of unknowns 2000 2500 Figure 4.66 Number of entries in the interaction matrix of the SDMM that has to be evaluated after all duplicate entries have been identified.6 GHz Pentium 4 processor at one frequency point.

Sign up to vote on this title
UsefulNot useful