An octave bandwidth dipole antenna Abstract: Achieving wideband performance from resonant structures is challenging because their radiation properties and impedance characteristics are usually sensitive to frequency. Therefore innovations in structural design are necessitated in order to transform any simple resonant structure into a wideband antenna. To be a practically useful antenna, the structural transformations have additional constraints in that (i) the antenna is required to be low loss, ii) the structural complexity is required to be compatible with manufacturing methods, and ii) the antenna needs to have close to uniform impedance over the wide frequency band. We have designed and developed an octave band short fat dipole with a sine square profile. It operates over the frequency range of 87.5 MHz to 175 MHz with frequency independent radiation patterns that very nearly cosine square and an input return loss of more than 8 db over the octave range. The structure has been optimized using EM modeling. The design has been validated by constructing a prototype and comparing measurements with expectations from the EM simulations. 1. Introduction: A dipole is a simple antenna that works on the principle of resonance. Inherently it has less bandwidth and as a result its impedance is a function of frequency. Thus achieving good impedance match over wide frequency range is difficult. Several structures exist in the literature which exhibit broadband performance. In general, any antenna structure that has its dimension varying smoothly would radiate energy efficiently over a wide frequency range. Increased efficiency of radiation in those structures is brought out by acceleration of charges along the surface. However, its impedance characteristics would be independent of frequency only if its surface current decays linearly in a frequency dependent fashion away from the feed point. Ideally this requires a structure of infinite extent. Since in practice structures are of finite dimension, they tend to have lesser bandwidth of maximum efficiency. A study was made of truncated structures with a variety of profiles like exponential, conical and sinesquare for the dipole arms. The electromagnetic behaviour of each was simulated using the WIPL-D CAD package to characterize their electrical properties. 2. The goal of the antenna design: The antenna being designed will be primary used for detecting the red shifted 21cm line generated by the neutral hydrogen in the intergalactic medium during the epoch of reionization (EoR) of the universe. This signal is expected to be several orders of magnitude weaker than the galactic foreground at low radio frequencies. Nevertheless, it is expected that its spectral signatures may be detectable because other celestial sources in the foreground of the cosmological gas, whose emission mechanisms are thermal brehmsstrahlung and synchrotron radiation, have relatively smooth continuum spectra. However, foreground sources have a distribution in intensity on the celestial sphere, which will result in frequency structure in measurements of the sky spectra that are made with antennas with frequency dependent beam patterns. Therefore, one of the primary requirements of the antenna is that it should be frequency independent.
The EoR measurement targets a redshift range corresponding to the frequency range of 87.5 MHz to 175 MHz, over this range the antenna should have identical characteristics. This ensures equal sensitivity across the frequency band for the sky signal in different directions. If this criterion is not satisfied, then the antenna spectral response will have undesirable features that may be confused with the sky spectrum. Beam symmetry is one more important parameter which characterizes the antenna performance. Since in the EoR experiment, sky data is averaged over the entire 24 Hrs. of LST, asymmetry in the beam leaves behind features in the spectrum due to incomplete cancellation of the imaginary component of the receiver response for the sky signal. A highly desirable feature in any antenna is good impedance match. Mismatch would result in reflections of the sky and receiver noise signals between the antenna and the first amplifier module to which it is connected giving rise to ripples in the pass-band. Generally ripples in the passband should be minimized since it may confuse and limit our ability to detect the signatures of EoR. Since the EoR signal is randomly polarized, the polarization of the antenna may be either linear or circular. We have designed the EoR antenna to have linearly polarization. Based on these requirements, the antenna is designed to a) be linearly polarized over the frequency range of 87.5 MHz to 175MHz b) have its 3dB beam width varying not more than a few percent across the entire band c) have symmetrical beam shape d) have its input return loss better than -10 db. 3. Investigation of structures for their broadband performance: A two wire uniform transmission line of infinite length is known to exhibit broad band performance by offering a constant characteristic impedance independent of frequency. When the transmission line is truncated, the bandwidth over which it has a constant characteristic impedance tends to decrease due to the production of reactive components to the reflection of signals at the end of the structure. It also suffers from poor radiation efficiency since charges travel on them with uniform velocity. It has been established that a flare at the end of a transmission line would improve both its impedance bandwidth as well as radiation bandwidth (Kraus). Careful studies were conducted through EM simulations to understand the effect of flare on the overall performance of the transmission line. Since a dipole may be viewed as a truncated transmission line, profiles like exponential, conical and many others were attempted for the dipole arms in the process of investigations. Among these the dipole with sine square profile outclassed the performance of all others in meeting the specifications required for the EoR experiment. 4. Simulation of a dipole having sine square profile A fat dipole was simulated in WIPL-D CAD package between 87.5 MHz to 175 MHz. It was initially modeled to be half wavelength long close to the highest frequency of operation, with a small finite gap inbetween the two arms. The surface was given a sine square profile and
was terminated at the end in a conical section. This would avoid abrupt termination of the structure and minimize consequent reflection of surface current. The shape of the surface was found to have a major impact on the production of reactive components of radiation impedance due to finiteness of the radiating structure. Hence the shape parameters were required to be tuned to achieve good impedance match over wide bandwidth. The feeding section was also given a conical profile while performing the simulations. The structural parameters like length of dipole, the amplitude of the sine square profile and its index, gap between the two halves of the dipole and thickness of the dipole at the feeding point were all optimized during the process. The dipole profile optimized for best performance is given in Eq (1) along with the optimized structural parameters. 15 where x All dimensions are in mm 435mm 200 x= r + A*sin α (kz) (1) r is half the length of the input square waveguide = 90.7 mm λ A is the amplitude of the profile = 1.2* 8 α is the power of the sinusoidal function = 2.022 λ is the wavelength corresponding to 184 MHz 41.4 x, z are the co-ordinates along x and z axes respectively 2 3.14159265 k = λ 660 Sine square z profile A=245.1 367.72 mm Fig. 1(a) shows the schematic of the dipole structure with dimensions tuned for optimal 25.86 performance in the frequency range 87.5 175 MHz and (b) shows the computer modelused for the EM simulation. 2r=181.47
(a) (b) Fig. 1. (a) Schematic diagram of the EoR dipole antenna with optimized parameters (b) computer model The simulation results revealed that the profile that has sine squared shape would make the dipole useable in the frequency range 87.5 175 MHz with a radiation pattern that is independent of frequency and an impedance match corresponding to a return loss of -10 db. Fig. 2 Photograph of the antenna built based on the EM optimization ( front view)
Fig. 3 Photograph of the antenna built based on the EM optimization In the process of ( Side fabrication, view) the surface of revolution of the dipole antenna was approximated by a finite number of flat aluminium sheets, each bent along the optimized profile. This was done to minimize the complexity in the fabrication and to make it possible for the antenna to be disassembled and transported in a compact package to remote observing sites. Fig. 2 and 3 show the photograph of the fabricated antenna. Several antenna measurements were made to characterise the radiation and impedance properties of the antenna. Sections below describe these measurements and results. 5. Measurement of the radiation Pattern of the antenna The antenna radiation pattern was measured at several frequencies in the frequency range 87.5 175 MHz. In general, measurement of the radiation pattern of non-directive antennas at low frequencies is challenging since multiple reflections from nearby objects and particularly the ground poses formidable challenges to accurately characterizing the antenna performance. Since antennas at low frequencies tend to be bigger in size and have larger regions of reactive fields, the zone of avoidance around the antenna under test tends to be larger at lower frequencies exacerbating the problem. The electrical size or cross-section of stray wires and other metallic objects in the environment is greater at lower frequencies, making the requirement of having a clean and large volume for measurements important. In the experimental set up for the measurement, the antenna under test was used as a receiver and a short dipole antenna was used as the transmitter. Both the antennas were kept off the ground with the fat dipole arms parallel to the ground. Ferrite tiles were spread across the ground over which the measurement was made. The ferrite tiles that were selected for this measurement setup are good absorbers of radio waves over a large range of low frequencies; their reflection in the frequency range 87.5 MHz to 175 MHz is more than db down. To this accuracy, these tiles are expected to significantly reduce ground reflection from affecting the pattern measurement. A stable signal generator (Agilent 8754 ) is used to feed the transmitter with CW at discrete frequencies across the band and at a constant power level. The receiver output is connected to a spectrum analyser. While conducting the pattern measurement, the transmitter is kept fixed and the receiver is rotated in azimuth over a range exceeding 180 degrees. The orientation of the receiver antenna (antenna under test) was measured using a magnetic compass fixed to the antenna. For each static position of the receiver antenna, the power level of the CW received is recorded. The general arrangement of the experimental setup is as shown in Figure 4. The measurement layout was evolved till reliable measurements were obtained. Issues that had to be addressed included stray pickups in cables connected to the antenna under test, metallic objects that disturbed the compass orientation, proximity to the absorber surface covering the ground. The pattern was measured at 90, 130 and 170 MHz. The measured patterns were symmetrical in gain and had no significant offsets in peak that might arise from errors in feeding. The
nulls on either side of the peak were as deep as 27 db down.the patterns over the octave band of frequencies were very similar indicating the broad band performance of the antenna. Figs. 5 and 6 show the plots of radiation pattern at various frequencies in both logarithmic and linear units: log scale plots are useful to show the stop band performance and depth of the nulls, linear scale plots are useful in examining the pass band performance and shape of the main lobe. The mean half power beam width (HPBW) across the entire frequency range is about 87.9 degrees. This matches very well with the WIPL-D simulation result of 89 degrees as shown in Figs. 7 and 8. The difference between the cosine square curve and the measurements is given in Fig. 8a. Short dipole: transmitter Antenna under test: receiver CW Signal Generator Spectrum Analyser Ferrite absorber tiles Fig. 4. Experimental setup for the measurement of Radiation Pattern Fig 5: Radiation pattern of the EoR broad band dipole. Vertical scale in db units. Continuous line shows theoretical cosine square pattern of an ideal dipole; symbols show measurements at 90, 130 and 170 MHz.
Fig 6: Radiation power pattern in linear scale. Fig 7: Simulated response of the 3D radiation power pattern in logarithmic scale around the antenna structure.
Fig 8: Simulated response of the radiation power pattern in logarithmic scale. Fig. 8a. Deviation of the measured radiation pattern from Cosine square 5. Measurement of Return loss of EoR antenna Return loss is an important specification for any antenna. It is an indicator of the loss in signal while getting transferred from free space via the antenna to the receiver terminals, and includes the impedance mismatch between the antenna and free space as well as between the antenna and transmission line; it also includes resistive loss in the transmission between the
antenna and receiver. It is usually preferred to have minimum mismatch and minimum resistive loss for maximizing the signal transmission and radiation efficiency. The EoR antenna was characterized for its input return loss in the frequency range 87.5 to 175 MHz. In our experimental setup (Fig. 9), an Anritsu vector network analyser was used both to excite the antenna and measure the reflected signal. The analyser is capable of generating signals over a wide range of frequencies with varying power levels. While measuring the return loss, the antenna under test is placed over ferrite tiles in order to ensure minimum reflection from the ground as well as to make the measurements in conditions similar to that of the EoR experiment. Transmitted signal Antenna Vector Network Analyser Balun / Transfer Switch Reflected signal Fig. 9 Experimental setup used for the measurement of the return loss of antenna Fig. 10 The photograph showing the methodology of connection at the feeding point of the antenna In the measurement of the EoR signal, two antennas are used in the zero-spacing interferometer mode. The two monopole outputs are connected to the center conductors of two semi-rigid cables whose outer shields are soldered all along their length. Subsequently, the pair of signals from the antenna arms goes through a balun for converting the pair of
balanced outputs to a single unbalanced form. In one antenna, a cross-over switch is introduced between the antenna arms and balun. The arrangements are shown in Fig. 10. In one of the antennas (Antenna 2) the transfer switch is introduced and used for phase modulating the antenna signal. The switch can either connect directly the monopole outputs of the antenna to the balun or swap them before connecting, as shown in Figs. 10 b and c. Swapping of the monopole outputs results in introducing 180 degrees phase to the antenna signal. Balun Tr. St. Sw. Tr. Sw. cross Balun Balun Unbalanced output of Ant 2 with Unbalanced straight connection output of Ant in the 2 with Transfer cross Switch connection in the Transfer Sw Unbalanced output of Ant - 1 a b c Fig. 10 E(a) Schematic showing the balun and Transfer switch connected to EoR antenna balun (b)(c) Three return loss measurements were made to characterize the EoR antennas. In the first measurement, return loss of Antenna 1 is measured by connecting it to the scalar network analyser. In the second and third measurements, return loss of Antenna 2 was measured in both (swapped and unswapped) configurations of the transfer switch. These two measurements indicate the behaviour of the switch in both its states. We expect the switch to behave identically in both the modes, with equal insertion loss and phase change for the signal. Fig. 11 shows the plot of return loss of Antenna 1 as a function of frequency. Its return loss is better than 8 db over the frequency range of 87.5 MHz to 175 MHz. Similarly the return loss of Antenna 2 taken in both the configurations of the transfer switch are shown in Fig. 12. Both the plots are nearly identical indicating identical characteristics of the switch in both the states. Due to the insertion loss of the transfer switch, the return loss of Antenna 2 appears to be better than Antenna 1. The measured return loss matches closely with the EM simulation results shown in Fig. 13. However, we observe that the maximum value in the return loss has shifted towards lower frequency. This might be because the antenna would have become electrically longer due to the copper foils introduced at the feeding section for establishing electrical contact with the balun.
Fig. 11. Plot of the return loss of Antenna 1
Fig.12. Plot of the return loss of Antenna 2 in both swapped and unswapped modes: (Phase switch control : 01 corresponds to unswapped mode and 10 to swapped mode) Fig. 13. Return loss of the Antenna 1 as expected from EM simulations Conclusions: We have successfully developed a design for a low loss octave bandwidth resonant antenna of the fat-dipole class. The antenna has close to identical radiation patterns and an input return loss of < -8 db over the entire frequency range of 87.5 MHz to 175 MHz. The radiation
pattern is close to a cosine square pattern. The measurement results match closely with the expectations of EM simulations. The fat-dipole was used as a first antenna to test the feasibility of the zero-spacing interferometer concept.