A new antenna layout for a Super Dual Auroral Radar Network (SuperDARN) HF radar has been developed. The new layout utilizes two auxiliary arrays; one behind and one in front of the main array, rather than the single auxiliary array that existing radars use. The rear auxiliary array consists of three antennas providing beam-steering capability while the front auxiliary array consists of a single antenna. This layout is expected to greatly improve the calculation of elevation angle of arrival. Simulations presented show the advantages and disadvantages of using twin-terminated folded dipole (TTFD) antennas and log-periodic dipole arrays in standard and modified SuperDARN array configurations. TTFD antennas are shown to have superior front-to-back ratio and beam-steering capability but suffer from shadowing effects due to the presence of corner reflectors. Impedance-matching techniques used in SuperDARN radars are discussed, and the results of a new matching method, exhibiting a superior voltage standing-wave ratio over the SuperDARN frequency band, are presented. Shadowing of the main array by the front auxiliary array is investigated, and it is shown that the impact of the front array on the main array gain pattern is significantly less for the case of a single front antenna than for a four-antenna front array. Radar phase calibration techniques are discussed, and it is proposed that the additional single-antenna front array be used for system-wide radar phase calibration. An algorithm for the determination of elevation angle of arrival using the new layout is also given.
 The Super Dual Auroral Radar Network (SuperDARN) is a global network of similar HF (8–18 MHz) radars employing linear arrays of horizontally polarized antennas [Greenwald et al., 1995; Chisham et al., 2007]. There are currently two distinct types of antennas used in SuperDARN radars; log-periodic dipole array (LPDA) antennas and twin-terminated folded dipole (TTFD) antennas. A list of current SuperDARN radars is provided in Table 1, which includes the type of antenna deployed. In addition to the main 16 antenna array, SuperDARN HF radars employ a single auxiliary linear array with fewer antennas as an interferometer for the detection of backscatter elevation angles. Nominally, the displacement between the main and auxiliary arrays is 100 m, yet over the years there have been considerable variations in array layouts (see Table 1). Most variations are due to local constraints imposed by terrain variation over the array area and interfering objects, while some variations have been specifically chosen. Of all the existing SuperDARN radars, nine have significantly different displacement distances along the boresight (Y), five have offsets along the line of antennas (X), and six have the main and auxiliary arrays at different altitudes (Z).
Table 1. Interferometer Position Relative to Main Array for Existing SuperDARN Radarsa
X is along the line of antennas with +Xtoward higher antenna numbers, Y is in the direction of the boresight, and Z is the altitude difference.
denotes TTFD antenna installations, while the rest are LPDA installations.
 Since the main interferometer displacement (Y) is greater than the maximum radar wavelength, an elevation angle of arrival aliasing effect is observed where only elevation angles below a certain angle, Δmax, are correctly interpreted (see A. J. McDonald et al., Elevation angle-of-arrival determination for a standard and a modified SuperDARN HF radar layout, submitted to Radio Science, 2013). For the Hankasalmi and Pykkvibaer SuperDARN radars for example, Δmax is equal to 35° and 45°, respectively [Milan et al., 1997]. From Table 2 it can be seen that a reduction in the Y component of the main auxiliary array displacement increases Δmax, allowing a larger range of elevation angles to be detected. However, reducing the Y displacement also increases Radio Frequency (RF) interaction and shadowing between the main and auxiliary arrays.
Table 2. Variable Interferometer Y Displacements and Δmaxfor SuperDARN Operational Frequencies (8–18 MHz)a
Y displacement (m)
Δmax 8 MHz
Δmax 18 MHz
The calculation of these values is based on expressions found in Milan et al. .
 The quality of elevation angle data is influenced by many factors including HF propagation characteristics, backscatter signal spreading, array interferometry, and radar hardware phase variation. While the elevation angle errors associated with signal spreading and HF propagation are unavoidable, they are small in comparison to those associated with elevation angle aliasing and radar hardware phase variation [Ponomarenko et al., 2011].
 Analog electronic paths cause phase delays inside the transmitters and receivers which vary with frequency and temperature and also drift with time as components degrade. In existing SuperDARN radars the phase variability is left unchecked which can cause inaccuracies during beam forming and also during backscatter elevation angle determination [Milan et al., 1997; Ponomarenko et al., 2011]. These inaccuracies become even more evident when a multiple-transceiver radar is employed. Having a multiple-transceiver system means that all analog paths must have matched phase delays for correct radar operation. There have been various attempts to calibrate SuperDARN radars by modifying hardware. However, since phase calibration needs to be performed on a routine basis and due to the remote locations of many of the radars, regular phase calibration has often proved inconvenient or impractical.
 The antenna type and its characteristics can impose restrictions on the ideal array layout and as a result can also influence elevation angle measurements. SuperDARN radars almost exclusively use LPDA and TTFD antennas, with the latter being used in most SuperDARN installations since 2005 (see Table 1). There are many performance and mechanical design differences between the antennas, as shown by Custovic et al. [2011a] and Sterne et al. .
 In any antenna array, RF coupling and shadowing between nearby antennas is an important consideration. The TTFD has recently been favored over the LPDA in SuperDARN installations for its superior front-to-back ratio, superior beam-steering capability, and overall cost reduction [Custovic et al., 2011a]. However, the use of corner reflectors in TTFD antennas to improve the front-to-back ratio can introduce significant shadowing and signal attenuation between the main and auxiliary arrays.
 In this paper we present simulation results for the gain patterns of the TTFD antennas including corner reflectors to understand the potential for RF interaction between nearby antennas and shadowing between the main and auxiliary arrays. We describe in detail a new SuperDARN HF radar antenna layout which is optimized for TTFD antennas and the calculation of elevation angle of arrival. We analyze the gain patterns of one-, two-, and four-antenna linear arrays and discuss issues related to the signal-to-noise ratio of the front auxiliary array. We also discuss improvements to radar phase calibration which will be made possible by the new antenna layout.
2 SuperDARN Antennas: LPDA Versus TTFD
 SuperDARN radar installations deployed between 1985 and 2005 exclusively use LPDA antennas (see Table 1). The SuperDARN LPDA is made up of 10 tapered dipole elements; the longest of which is 14.94 m. The last two elements are inductively loaded to extend the operating frequency without physically extending the length of the longest element. It is well known that in phased linear arrays the optimum antenna separation for beam forming is λ/2 [Ma, 1974]. For the SuperDARN frequency range of 8–18 MHz, this results in an optimum spacing between 8.3 m (at 18 MHz) and 18.5 m (at 8 MHz). The LPDA arrays in existing radars have an antenna separation of 15.24 m which cannot be reduced due to the length of the longest dipole element. For an operating frequency of around 10 MHz this is close to the optimum antenna separation and provides good beam steering. For operation at frequencies far away from 10 MHz, however, beam-steering capability can degrade significantly. Due to RF coupling between adjacent antennas, as the azimuthal direction of the desired beam forming moves away from the boresight, side lobes in the gain pattern become increasingly large. Azimuthal simulations of a 16 antenna LPDA array at 14 MHz, 0°, 22°, and 40° from the boresight are shown in Figure 1. For beam directions greater than about 30° azimuth from the boresight, the power of side lobes for a LPDA antenna is comparable to the forward power, such that echoes detected from side lobes will be mixed in with the backscatter from the main beam direction. The front-to-back ratio of LPDA antennas is also poor, with large back lobes observed at all azimuthal angles. The presence of large back lobes has been shown by Milan et al.  to greatly complicate data interpretation. In addition, the tapered LPDA dipole elements tend to sag over time and have the potential to result in additional coupling between antennas as the dipole elements deviate from the horizontal plane.
 Most SuperDARN installations since 2005 utilize TTFD antennas (see Table 1). While simulations have shown that the TTFD has a marginally lower average gain across the 8–18 MHz range compared to the LPDA [Custovic et al., 2011a], the beam can be easily steered to large azimuthal angles with greatly reduced side and back lobes. The compact size of the TTFD allows adjacent antennas to be placed closer together, enabling effective beam forming at angles greater than ±30° from the boresight, while maintaining a main lobe to side lobe ratio of at least 10 dB. The TTFD antenna installations are also supported by a corner reflector behind the antenna which almost completely eliminates the back lobes. Figure 2 shows simulations of a 16-antenna TTFD array with beam forming at 0°, 22°, and 45° from the boresight.
 With each SuperDARN installation, the TTFD has undergone optimization processes which has resulted in several variations of the TTFD antenna design. For example, the Wallops Island radar utilizes 11 corner reflector wires and 100 Ωterminating resistors, while the Blackstone radar utilizes 21 corner reflector wires and 75 Ωterminating resistors. The increase in corner reflector wires increases the power in the forward direction, while the decrease in terminating resistance minimizes the absorbed power across the frequency band. Most importantly, the antenna spacing was reduced from 15.24 m to 12.9 m to provide better beam-steering capability. Based on the improvements, the Blackstone TTFD model has been deployed on several other sites (see Table 1).
 As part of the deployment of the third Tasman International Geospace Environmental Radar (TIGER) at Buckland Park, South Australia (described in Custovic et al. [2011b]), the TTFD has been further optimized. The length of the individual antenna element has been increased from 10.8 m to 12.8 m. The additional length minimizes the peak real and imaginary impedance below 10 MHz which result from the mismatch between antenna length and the half wave length at those frequencies. Also, the structure of the TTFD has been modified by replacing the common central wire with two individual wires, eliminating the use of a splice.
 An important consideration when designing an interferometer array is shadowing between the main and auxiliary arrays. Shadowing refers to the blocking and attenuation of RF signals of one linear array by a nearby linear array. In the case of the LPDA, the presence of the auxiliary array in front has a negligible effect on the main array gain pattern because it is confined to a horizontal plane. In contrast, TTFD antennas have large corner reflectors to direct the antenna gain in the forward direction. The corner reflectors rise approximately 17 m above ground level while the most elevated TTFD wire of the radiating element is no higher than 11 m above ground [Custovic et al., 2011a]. This means that the auxiliary array, if it is placed behind the main array, is partially occluded by the main array corner reflector. If, on the other hand, the auxiliary array is in front of the main array, the corner reflector of the auxiliary array spreads and deflects the main array beam. In particular this can impact signals with elevation angles less than about 25°. Figure 3 shows simulations of the difference in gain along the boresight at the main array, with and without a four-antenna auxiliary array located 100 m (violet line) and 67 m (green line) in front of the main array. It is observed that main array signals are deflected from lower to higher angles by the presence of the four auxiliary array corner reflectors with a loss of gain of up to 1 dB at a displacement of 67 m and up to 0.5 dB at a displacement of 100 m. Note that at very low elevation angles the difference is smaller since the absolute gain at these angles is much smaller.
2.1 Antenna Impedance Matching
 A primary difference between LPDA and TTFD antennas are their frequency dependence. The LPDA is a multiple-element antenna, while the TTFD is a simple long wire. Consequently, the same impedance-matching technique may not be adequate for both types of antenna arrays.
2.1.1 LPDA Matching
 The LPDA utilized in older SuperDARN radars has a nominal impedance of 200 Ω. Each of the 10 dipole elements are individually fed via a transmission line, and as a result there is minimal impedance variation across the frequency band. The impedance is then matched using a simple 2:1 toroidal-based balun.
2.1.2 TTFD Matching
 The TTFD antenna is a long wire antenna and is therefore highly frequency dependent, with large impedance variation over the most commonly used band of SuperDARN frequencies, 9–16 MHz (see Figure 4). The measured impedance graphs presented here are for a 12.8 m dual wire TTFD.
 SuperDARN radars typically use toroidal-based transformers for antenna matching which can incur a significant power transfer loss due to ferrite saturation at higher powers. This is a particular challenge for the Buckland Park radar which is equipped with a 2.4 kW solid state amplifier [Custovic et al., 2011b]. Additionally, toroidal transformers are not very effective in matching large variations in inductive and capacitive antenna impedances. Figure 5 shows voltage standing-wave ratio (VSWR) results for the Buckland Park TTFD antenna using three types of baluns used by other TTFD-based SuperDARN radars: the Sil 5 core, Sil 4 core, and Array Solutions (AS) 1 core baluns. It can be seen that none of the toroidal-based baluns provide an ideal response for the SuperDARN range of frequencies. It is highly desirable to have a VSWR under two for most of the frequency band. More importantly, none of the baluns can provide a flat response and this is inherent to all toroidal-based baluns. For this reason a custom-made inductor-capacitor (LC) matching network, made from aircore wound inductors and microstrip transmission lines, has been developed by the authors to adequately match the nonreal impedance components. The network is optimized using gradient and pattern search techniques based on the Genesys RF and Microwave Design package from Agilent Technologies (see www.home.agilent.com). This ensures an optimum solution for the designated band. Figure 6 shows VSWR readings taken at the Buckland Park radar site using the new matching network. The custom matching network effectively matches the impedance over the frequency range 9–17 MHz, to provide a relatively flat VSWR response. In addition, the antenna system real impedance does not need to be a squared multiple of 50 as is the case with a matching transformer. Hence, the antennas have been optimized to a nominal impedance that minimizes the VSWR variation across the band.
3 New SuperDARN Antenna Layout
 The new antenna layout has been developed in conjunction with the Buckland Park radar and is shown in Figure 7. The 16-antenna main array is consistent with existing SuperDARN designs. The interferometer layout consists of a rear auxiliary array with three TTFD antennas plus a front auxiliary single-antenna subarray. This configuration gives the advantage of multiple main auxiliary array displacements normal to the boresight while maintaining some beam-steering capability for the rear auxiliary array.
 The relative values of d1and d2(see Figure 7) have been chosen to permit the unambiguous calculation of elevation angles. An algorithm for the determination of elevation angle using the new layout is given in section 4. By making the main interferometer array spacing different for the front and back interferometer arrays it is possible to remove the aliasing effect in the elevation angle calculation observed in existing SuperDARN radars. For full details, see A. J. McDonald et al. (submitted manuscript, 2013).
 The distance between adjacent antennas governs the overall radar beam-steering capability, with a narrower spacing allowing beam-steering capability to greater azimuthal angles away from the boresight. The antenna separation for the Buckland Park radar was chosen as 14 m which is larger than the 12.9 m separation used for the recent Blackstone TTFD installation. This increase is to accommodate the longer TTFD wires used as described in section 2 while still providing good beam-steering capability. While the antenna spacing could be reduced to as little as 13 m, this would increase coupling between antennas as well as the VSWR.
3.1 Single-Antenna Subarray Considerations
 While there are many combinations of d1and d2which will enable unambiguous calculation of elevation angles (see A. J. McDonald et al., submitted manuscript, 2013), the front auxiliary subarray is limited to 67 m in front of the main array due to restrictions imposed by the radar site. Figure 3 (blue line) shows a simulation at 14 MHz of the difference in gain along the boresight at the main array, with and without a single-antenna subarray located 67 m in front of the main array. This shows that a single TTFD with a corner reflector in front of the main array has a minimal impact on the main array transmit beam, even at the reduced displacement of 67 m. The same simulation was conducted for the highest and lowest SuperDARN operational frequencies of 8 MHz (red line) and 18 MHz (violet line) with the results shown in Figure 8. Simulation results are also provided for a four-TTFD front array at 67 m for 8 MHz (green line) and 18 MHz (blue line). The loss of gain at low elevation angles is observed to increase with increasing frequency. This is due to the decrease in takeoff angle with increasing frequency which results in more RF energy being deflected off the front array reflector at higher frequencies.
 Figure 9 shows the azimuthal gain pattern of a single TTFD antenna with corner reflector for 10 MHz (violet line) and 14 MHz (green line). For reference, the gain pattern for two- and four-antenna arrays at 14 MHz (black and blue lines, respectively), with beam forming directed along the boresight, is also shown. The gain along the desired azimuth, in this case the boresight, of the single antenna is observed to be around 5–7 dB less than that of a four-antenna array with beam forming.
 The random noise contribution of a multiple-antenna system scales with the square root of the number of antennas, while the coherent receive power scales linearly. Hence, we expect the SNR of a multiple-antenna array to scale with the square root of the number of antennas. Therefore, going from a four- to one-antenna interferometer array means SNR is halved; that is, we see a reduction of 10 log10(1/2)=3 dB.
 To offset these reductions, the new output HF amplifier has four times higher power output [Custovic et al., 2011b], increasing from 600 W to 2400 W, which represents a 6 dB increase in received power. In addition, the new antenna array utilizes improved output filtering, antenna impedance matching, and lower loss cabling [Custovic et al., 2011b], which accounts for at least a 2 dB increase in system SNR.
 Overall, 6 dB is lost from the lack of beam forming and 3 dB due to the reduction in SNR. From the increased transmit power 6 dB is gained and 2 dB from low loss cabling and improved output filtering/impedance matching. Therefore, the single-antenna subarray of the new system is expected to have only a marginally reduced SNR performance compared to the original four-antenna interferometer array.
 Ideally, a 4-16-4 layout would be preferable over the present 1-16-3 in terms of SNR. However, the number of antennas in the front auxiliary array is a compromise between increasing SNR and minimizing the impact to the transmit and receive radiation pattern of the main array. The degree of shadowing of the main array expected for a two- and four-antenna front auxiliary array at an offset of 67 m at 14 MHz is shown in Figure 3 (red and green lines). It has also been observed in simulations (results not shown) that the front-to-back ratio of the main array decreases by approximately 3 dB when a four-antenna TTFD array is in front of the main array at a displacement of 67 m.
 The 1-16-3 layout has been developed to provide maximum performance while maintaining the traditional number of antennas for a SuperDARN radar. It is possible, however, that a 2-16-4 layout would be preferable with an additional antenna added to the front array to improve the signal-to-noise ratio.
3.2 Radar Phase Calibration
 There are three sources of phase variability within a radar system: the transceiver, the transmission cables, and the baluns/antennas (see Figure 10). Currently there is no comprehensive calibration method available for SuperDARN radars [Ponomarenko et al., 2011]. In this subsection we briefly describe two calibration methods developed for the Buckland Park SuperDARN radar.
3.2.1 Radar Transceiver Phase Calibration
 Multiple-transceiver radar systems are prone to variable phase lags within hardware systems such as transmitters, receivers, cables, baluns, and antennas. This phase variability between transceivers leads to phase ambiguities and problems with data interpretation. An in-transceiver phase calibration method using a direct digital synthesizer (DDS) signal as a reference has been developed for Buckland Park. In order to resolve the ambiguities posed by multiple-transceiver phase variability, each transceiver will utilize several analog to digital converters (ADCs). The ADCs sample and measure transmitting and receiving signal phase differentials which can be used to adjust each DDS so that all transceivers are phase aligned. This calibration method has been described previously in Custovic et al. [2011b].
 The in-transceiver calibration method does not take into consideration the coaxial feeder cable between transceiver and antenna, antenna balun or the phase variability induced by the antenna impedance. However, for SuperDARN radars using LPDA antennas, phase calibration within the transceiver may be sufficient. The antennas are industry manufactured and have a nominal impedance of 200 Ω, with minimal variation across the frequency band [Custovic et al., 2011b]. In addition, the use of a single-core toroid balun to match the antenna impedance, with careful design, can result in a relatively small phase differential between the 16 antennas.
3.2.2 System-Wide Phase Calibration
 The LC matching network described in this paper is a multiple-order filter consisting of hand wound inductors and microstrip transmission lines. Regardless of careful design and construction, the total combined component tolerances will result in much larger differential phase shifts at the antenna input Custovic et al. [2011b]. An additional calibration method is proposed using the single-antenna front array as a reference. System-wide independent radar calibration will be performed on site at regular intervals. This technique is currently under development and will be tested upon completion of the radar.
4 Angle-of-Arrival Calculation With the New Antenna Layout
 For the new radar layout described in section 3, the elevation angle from the horizontal Δ can be expressed in terms of the measured main auxiliary array phase lags ψ:
where ψ1and ψ2 are the measured main auxiliary phase lags for the front and rear auxiliary arrays, respectively, and φ0 is the azimuth from the boresight at zero elevation angle. a=d1/λand b=d2/λ with d1 and d2 the main auxiliary array displacements (see Figure 7) and λthe signal wavelength. m and n are unknown integers resulting from the aliasing effect or 2πambiguity in the total phase lag (see, for example, Milan et al. ).
 For the choice of d1=67 m and d2=−80 m used in the construction of the Buckland Park SuperDARN radar, for ψ1,ψ2∈[−π,π), it can be shown that if ψ2≥ψ1then n=mand if ψ2<ψ1then n=m+1 (see A. J. McDonald et al., submitted manuscript, 2013). So in practice, we can measure ψ1 and ψ2, express n in terms of m, and then solve (1) and (2) uniquely for m and Δ. Note that for the measured phase lag ψ2associated with the rear interferometer array, the substitution should be made before applying the algorithm.
 The new Buckland Park SuperDARN HF radar will utilize an auxiliary single-antenna subarray in front of the main array, in addition to the rear auxiliary array. System-wide phase calibration of the radar will be possible using the single-antenna subarray as a reference against which to calibrate the radar system. Dual interferometer arrays will also facilitate calculation of elevation angles without the aliasing effect, which prevents measurement of elevation angles below a certain angle.
 It is found that for frequencies significantly above 10 MHz, the beam-steering capability of LPDA antenna arrays gets progressively worse, with large side lobes appearing in the gain pattern for large azimuthal angles. For TTFD antenna arrays, a closer separation between adjacent antennas of the array is possible, allowing much better phased beam forming away from the boresight. The corner reflectors used with TTFD installations are also shown to greatly improve the front-to-back ratio.
 Simulations of a TTFD array in the standard SuperDARN layout show that the corner reflectors of a four-antenna auxiliary array in front of the main array cause deflection of the transmit beam and partially occlude the received signal at the main array. However, simulations of the new antenna layout indicate minimal impact on the gain at the main array due to the presence of a single front auxiliary antenna, even at a reduced displacement of 67 m.
 An LC matching network developed for matching the large impedance variation of the TTFD antennas is shown to be more effective than existing toroidal-based matching techniques. As a result the TTFD antennas can be expected to operate more efficiently over a wider frequency band.
 While a single-antenna auxiliary array will have a reduced signal-to-noise ratio compared with a four-antenna auxiliary array, the additional sensitivity and power provided by the new digital radar system is expected to compensate for this, allowing the single-antenna subarray to still perform its role as interferometer. The new layout has been developed to provide maximum performance while maintaining the traditional number of antennas for a SuperDARN radar. However, the authors acknowledge that additional antennas may need to be added to the front array, particularly for existing SuperDARN radar systems, to achieve the required signal-to-noise ratio. This would be a compromise between increasing SNR and minimizing the impact of shadowing on the transmit and receive radiation pattern of the main array.
 Andrew McDonald is supported by a postdoctoral fellowship through the Department of Defence, Defence Science Technology Organisation, and by an ARC LIEF grant in partnership with IPS Radio and Space Services.