V&V Article 6: Using EM.Cube And NeoScan System Together For Antenna Design
<script src="http://www.emagtech.com/themes/business/js/imag_box.js" type="text/javascript"></script>
Contents
Introduction
In this article, we will describe the process of designing an S-band (2.4GHz) rectangular microstrip patch antenna with a recessed feed as shown in Figure 1 using <a href="/content/emcube">EM.Cube</a>. The designed antenna is fabricated using a rapid prototyping milling machine. After being connectorized, the antenna is characterized using a vector network analyzer, and its far field radiation patterns are measured in EMAG Technologies' in-house anechoic chamber. The patch antenna is initially designed and optimzied using EM.Cube's Planar Module, also known as <a href="/content/empicasso">EM.Picasso</a>. This module assumes a dielectric substrate and an underlying perfect electric conductor (PEC) ground plane of infinite lateral dimensions. The same patch design is also analyzed using EM.Cube's FDTD Module, also known as <a href="/content/emtempo">EM.Tempo</a>. However, this time a finite substrate and a finite underlying PEC ground plane. The simulated and measured data for the return loss of the antenna, as well as its radiation pattern, are compared. We also compute the near-zone electric and magnetic field distributions on a horizontal plane placed slightly above the surface of the patch antenna. Both the Planar and FDTD modules of EM.Cube are used for these computations. <a href="/content/neoscan-turnkey-field-measurement-system">NeoScan</a>, EMAG Technologies' turnkey field measurement system is used to measure that actual electric field distribution map above the surface of the fabricated patch antenna. Both magnitude and phase of all the three field components are scanned using NeoScan's tangential and normal probes. The tangential field data are then used to estimate the far field radiation pattern of the patch antenna using a near-to-far-field transformation. The estimated far-field data are compared with the actual measured patterns in the anechoic chamber as well as simulation data. Interesting conclusions are drawn from comprison of all these various data sets.
Design and Optimization of Patch Antenna Using EM.Picasso
Figure 1 shows the geometry of a rectangular patch antenna of length L and width W, fed by a micostrip line of width wf and length Lf. For this design, the feed line is connected at the center of the edge of the patch, that is L1 = W/2. The patch is printed on a conductor-backed Rogers 4003C substrate with a thickness of 1.58mm and relative pemittivity εr = 3.38 as given by the manufacturer data sheet. On this substrate, a 50Ω microstrip transmission line has a width wf = 3.5mm. The length of the feed line is set to Lf = 30mm. The feed recess has a length of d and a gap width of g. In EM.Cube's Planar Module, the patch structure with the recessed feed is easily created using a large rectangle of dimensions L × W and subtracting from it two smaller rectangles of dimensions d × g. Note that this a constrained optimization problem. A square patch is assumed; hence, W = L. The three parameters L, d and g are used as optimization variables to tune and impedance-match the patch antenna. The coordinates of the recess rectangles and the feeding microstrip line are all tied to these design variables. Figure 2 shows the parameterization of these quantities based on three project variables: L, D and G. The initial values of these variables are set to 35mm, 10mm and 2mm, respectively.
EM.Picasso offers several ways to excite a patch antenna. The most accurate way for calculating the scattering parameters of the structure is to use a de-embedded port. In this approach, the feed line is extended to about two guide wavelengths, and an ideal gap souce is placed near the open end of the line. The surface current on the feed line forms a standing wave pattern. EM.Picasso uses an exponential interpolation scheme known as Prony's method to extract the scattering parameters of the structure from the standing wave data. Figure 3 shows the planar mesh of the patch antenna after its feed line has been extended for this purpose. After running a few parametric sweeps of the design variables and using one of EM.Cube's optimization algorithms to fine-tune those variables, their optimal values are found to be L = 34mm, D = 12mm and G = 2mm.
Fabrication and Characterization of S-Band Patch Antenna
Based on the above optimal values of the design parameters, a patch antenna was fabricated on a Rogers 4003 substrate and connectorized using an SMA connector as shown in FIgure 4. The frequency response of the fabricated antennas was measured using an Agilent 5378 performance network analyzer. The measured return loss (|S11|) is shown in Figure 5. It is seen from the figure that the antenna is tuned and matched at 2.349GHz rather than the original design frequency of 2.4GHz. It was subsequently found that Rogers Corporation recommends a value of εr = 3.55 for design purposes. Back in EM.Picasso, the relative permittivity of the dielectric substrate was change to the new larger value and an adaptive frequency sweep of the final optimized design was performed to examine the resonance behavior of the antenna. The simulated return loss (|S11|) is plotted in Figure 6. It shows a resonant frequency of 2.342GHz, which shows an error of less than 0.3%. Figures 7 and 8 show the surface current distribution and 3D far field radiation pattern of the simulated patch antenna, respectively, at f = 2.35GHz using EM.Picasso.
Figures 9 and 10 show the simulated 2D polar graphs of the radiation pattern of the patch antenna in the two principal planes YZ and ZX, respectively.
The radiation pattern of the fabricated antenna was also measured in EMAG Technologies' in-house anechoic chamber as depicted in Figure 11. Both co-pol and cross-pol patterns were emasured using a broadband log-periodic array that was oriented in vertical and horizontal positions. The patch antenna was also oriented on the turntable's pedestal in vertical and horizontal positions to measure the E- and H-plane pattern cuts. The measured results are shown in Figures 12 and 13 for YZ and ZX plane cuts, respectively.
Additional Design Verification Using EM.Tempo
The fabricated antenna obviously has a finite-sized substrate and a finite-sized ground plane. This affects the radiation pattern of the patch antenna and contributes to significant back-radiation as seen in Figures 12 and 13, which is absent from EM.Picasso's simulated patterns shown in Figures 9 and 10. To simulate a patch antenna over a finite-sized substrate with a finite-sized ground plane, we used EM.Cube's FDTD Module, also known as EM.Tempo. Figure 14 shows the project setup in EM.Tempo. A microstrip feed line of 30mm length is assumed. EM.Tempo offers different ways of exciting a microstrip line. For this project, we placed a +Z-directed distributed source at the open end of the microstrip line on a vertical plane right underneath the metal strip extending all the way down to the ground plane with a uniform impressed field distribution. This is illustrated in the inset of Figure 14. The time-domain source was excited with a modulated Gaussian waveform. An FDTD simulation of the structure was carried out in the time domain, and the scattering parameter S11 of the antenna was computed by EM.Tempo using a discrete Fourier transform. The return loss of the antenna is plotted in Figure 15. It is seen that the FDTD simulator predicts a resonant frequency of 2.278GHz, which has an error of about 3% with respect to the measured data. It must be noted that the feed mechanism of a radiating structure greatly affects its circuit characteristics such as S/Z/Y parameters. As a result, the resonant frequency might be shifted and the depth of the S11 notch may vary depending on the quality of impedance match. An accurate model must take into account the actual feed mechanism as well as the connector geometry. The reason why EM.Cube's Planar MoM solver (EM.Piasso) gave such good results was due to the fact that Prony's method indeed de-embeds the effects of feed mechanism from the simulation results.
Figure 16 shows the simulated 3D radiation pattern of the patch antenna with the finite-sized substrate and finite-sized ground plane. One can see that the radiation pattern of the finite-sized antenna is more directional with a higher directivity. Figures 17 and 18 show the simulated 2D polar graphs of the radiation pattern of the patch antenna in the two principal YZ and ZX planes, respectively. Considerable back-radiation is seen in both graphs. These graphs indeed agree quite well with the measured patterns of the fabricated in the anechoic chamber shown in Figures 12 and 13. Note that the measured pattern data are inverted with respect to the simulation data. In EM.Tempo we also set up a field sensor to compute the near-zone electric and magnetic field distributions. A horizontal field sensor plane was placed 1mm above the surface of the patch antenna. The plane extends across the computational domain parallel to the principal XY plane. Figures 19 and 20 show the computed magnitude and phase of the X-component of the electric field distribution on the plane, respectively, while Figures 21 and 22 show the computed magnitude and phase of the Y-component of the electric field distribution on the same plane.
Mapping Patch Antenna's Near Fields Using NeoScan
EMAG Technologies' NeoScan System is a turnkey field measurement system, which can be used for near-field scanning of antennas and other RF devices when combined with a translation stage. Figure 23 shows the fabricated patch antenna secured using a plastic holder fixture mounted on a percision XY table. An electro-optic (EO) field probe move above the surface of the patch antenna and samples the electric field on a horizontal plane above the device under test. For this measurement, a tangential EO probe was used as shown in Figure 24. The probe is first oriented along the X-axis to scan the X-component of the elctric field. It is then re-aligned along the Y-axis to measure the Y-component. During each scan, both the magnitude and phase of the particular electric field component are measured. Figure 25 shows the magnitude and phase of the X-component of the electric field at a plane about 1mm above the surface of the patch antenna. Note that the X-axis is oriented along the feeding microstrip line. You can clearly see the radiating edges of the patch antenna. This figure shows very good agreement with the simulated Ex magnitude and phase data computed by EM.Tempo and shown in Figures 19 and 20, respectively. Also note that you have to rotate the map by 90° to compare the two data sets. Moreover, the color scales of the graphs are opposite of each other. In other words, the red regions of Figures 19 and 20 correspond to the purple/blue regions of Figure 25. Figure 26 shows the magnitude and phase of the Y-component of the electric field at a plane about 1mm above the surface of the patch antenna. The contour of the feeding microstrip line is clearly visible in the figure, where the field exhibits singularity. This figure, too, shows very good agreement with the simulated Ey magnitude and phase data computed by EM.Tempo and shown in Figures 21 and 22, respectively.
Far Field Pattern Estimation Based on Near Field Maps
The near field maps of the patch antenna measured by the NeoScan system can effectively be used to estimate its radiation patterns using a near-to-far-field transformation. This transformation is indeed identical to the one used by EM.Cube's FDTD simulation engine for far field calculations at the completion of an FDTD time marching loop. The main difference is that the FDTD Module uses a closed surface, known as the Huygens box, which completely encircles the radiating structure. In this case, however, we use a 2D horizontal plane placed slightly above the surface of the antenna aperture. This plane, supposedly, must extend to infinity in the lateral directions. However, it is evident from Figures 25 and 26 that the fields decay very rapidly when one moves away from the boundary of the antenna. As a result, a knoweldge of tangential fields is required only at a finite portion of the horizontal plane immediately above the radiating aperture and its vicinity. Figure 27 shows the 3D far field radiation pattern of the fabricated antenna estimated from its near field maps. Note that estimated radiation pattern is more directional than the simulated pattern by EM.Picasso's Planar MoM solver. In fact, NeoScan's estimated directivity value of 5.389 is very close to 5.514 computed by EM.Tempo's FDTD simulation engine. Figures 28 and 29 show the estimated 2D polar graphs of the radiation pattern of the fabricated patch antenna using the NeoScan field maps in the two principal YZ and ZX planes, respectively. Note that the near-to-far-field transformation used for these results are valid only in the upper half-space due to the 2D surface scanning. That's why the polar graphs have been cut from the bottom and do not show any back-radiation of the finite-sized antenna structure.
Figures 30 and 31 put together all the radiation pattern data in the principal YZ and ZX planes, respectively, for comparison in a Cartesian graph. Specifically, the solid red lines represent the radiation patterns of the fabricated patch antenna estimated from NeoScan field maps, red symbols represent the measure radiation patterns of the fabricated antenna in the anechoic chamber, the blue solid lines represent the simulated radiation patterns of the patch antenna with a finite-sized substrate and finite-sized ground plane computed by EM.Tempo (FDTD) and the green solid lines represent the simulated radiation patterns of the patch antenna with an infinite substrate and infinite ground plane computed by EM.Picasso (Planar MoM). Generally speaking, a reasonable agreement is observed between the four different data sets. Note that in Figure 30 (YZ Plane), positive θ corresponds to the plane cut φ = 90° and negative θ corresponds to the plane cut φ = 270°. In Figure 31 (ZX Plane), positive θ corresponds to the plane cut φ = 0° and negative θ corresponds to the plane cut φ = 180°. Form these figures, it is seen that at grazing angles, the NeoScan pattern data approach those of the Planar MoM (PMOM) solver. This is all but expected due to the assumption of an infinite 2D Huygens planes covering the patch antenna and its substrate from the top.
