Compact and Planar End-Fire Antenna for PicoSat and CubeSat Platforms to Support Deployable Systems

A miniaturized planar Yagi-Uda antenna for integration with PicoSats or other SmallSat missions is proposed. Miniaturization techniques, such as meandering and 1-D artificial dielectric concepts to reduce the guided wavelength, are employed to overcome space constraints imposed by the SmallSat footprint while still maintaining good performance for the FR-4 antenna. Simulations and measurements have been carried out on the Unicorn-2 PicoSat chassis from Alba Orbital and are in good agreement. Also, antenna dimensions have been reduced between 15% and 66% when compared to a more conventional planar Yagi-Uda antenna working at the same frequency. This compactness allows for simple integration with the deployable solar panel array of the Unicorn-2 PicoSat spacecraft. Full end-fire radiation is achieved and peak gain values are about 5 dBi for the antenna when fully integrated on the satellite chassis, offering an attractive solution for downlink connectivity. This compact antenna design can also be used within an array for beam steering or integrated within the solar cell modules of other PicoSats, CubeSats and SmallSats. Applications include Earth observation, remote sensing, as well as SmallSat to ground station communications. The planar Yagi-Uda antenna may also be useful wherever end-fire radiation is required from a compact antenna structure.


FIGURE 1. Diagram of the Unicorn-2 PicoSat from Alba Orbital [7] with the proposed end-fire antenna integrated with the deployable FR-4 solar cell array module.
One key element of any payload offering different communication services is the antenna. Suitable beam pattern characteristics are also vital to maintain communications. In EO for instance, the power radiated from the antenna is required to be directed towards the Earth for data exchange, as depicted in Fig. 1. Depending on the required location of the antenna and how the satellite has been designed or is to be positioned in space, antenna selection may be limited when iso-flux, broadside or end-fire radiation patterns are required. Keeping these antennas compact, low-cost and lightweight is critical specially in missions where the wavelength can be large in comparison with the available footprint. As a result, designs based on turnstile monopoles, helix antennas [3], [4], or metallic and PCB-based structures [5], [6] may be considered.
Depending on the gain requirements, deployable antennas are versatile structures with a wide variety of applications not only limited to low frequencies. Antenna systems, as in [8], [9], based on reflectors for high frequency bands, were achieved with meshed or inflatable antennas. Although these deployable systems can make optimum use of the footprint in SmallSats for large antennas, the required stowage volume, and at times, the complicated deployment mechanisms can limit commercial and industrial adoption. On the other hand, low-cost planar solutions, despite typically providing moderate gain values, can be a reliable and more low-profile alternative. Furthermore, they have more degrees of freedom in terms of design flexibility. In this respect, integrating the antenna on the solar modules [10], [11], [12], [13], [14] is an attractive option to optimize the available space on the SmallSat whilst not requiring any specific deployment mechanism for the antenna itself.
In the Solant project [10], crossed slots were etched on amorphous silicon solar cells to achieve circularly polarized (CP) broadside radiation. This antenna design set the precedent for future configurations taking advantage of the available surface on the solar modules [11]. Conventional and modern techniques, such as grid antennas [12], [13], [14] or transparent conductors [15], [16], are enabling new planar structures which use the solar cell layer as the antenna substrate. This can avoid the perforation or relocation of the solar modules [17] whilst maintaining solar power harvesting capabilities. However, these antennas may provide lower efficiencies, which can be critical in the case of transparent conductors, since the conductivity can be of order 10 6 or 10 5 S/m depending on the employed conductive oxide [16]. Moreover, when full end-fire radiation is needed, antenna integration on top or near the solar cells might not be easily feasible. This is mainly due to the continuous ground plane needed under the thin solar cells (required for proper DC biasing and solar power collection circuitry) which could short-circuit important antenna elements. In this case, antenna radiation can be limited to quasi-end-fire or near broadside [18]. One solution is the removal of one of the solar cells to use the substrate underneath and pattern it as desired using conventional PCB design approaches. In this way, full end-fire radiation can more easily be achieved by using a variety of printed elements.
When requiring CP radiation at end-fire, antipodal configurations as in [19], [20], [21] could be considered. In these structures, complementary magnetic dipoles and mirrored open loops etched on top and bottom of the substrate are required. However, these designs provide optimum performance involving low permittivity substrates (ε r = 1.1). If considered for PicoSat or CubeSat integration, higher permittivities might be needed depending on the antenna design frequency. This size constraint can result in a lower radiation efficiency or reduced CP performance due to volume restrictions. In particular, antenna thickness requirements need to be on the order of 0.05λ 0 [19], [20], [21]. Moreover, such a physical requirement might make SmallSat integration, when operating in the L-or S-bands, not easily feasible, and, without the adoption of some vertical substrate thickness reduction technique or advanced antenna miniaturization.
Linear polarized (LP) antennas instead could be investigated when the possible antenna size and thickness is limited as is the case for the deployable solar cell array on the Unicorn-2 Picosat from Alba Orbital (see Fig. 1). When considering planar design solutions, there are a number of miniaturization techniques. Meandering of microstrip, the use of artificial magnetic materials [22], [23] or artificial dielectrics [24], can be considered for such SmallSat antenna integration. For end-fire radiation, conventional Yagi-Uda antennas or arrays, based on microstrip dipoles [25], can also be implemented and miniaturized to satisfy the size and operating frequency requirements [18]. For example, the preliminary investigations in [18] showed promising simulations where a full metallic chassis was assumed. In this case, the beam was diverted away from end-fire but gain values were kept around 3 dBi.
Some other works have also investigated planar Yagi-Uda antennas for SmallSats. For instance, [26] reported a 6 dBi printed Yagi-Uda array for 2.45 GHz and was simulated on the chassis of a 3U CubeSat. This antenna design with no miniaturization, needed a substrate of 90 × 150 mm 2 which could be too large for any PicoSat with additional payloads. This led to a modified version in [27] where copper rods embedded in a FR-4 substrate, formed a Yagi-Uda array that was mechanically steered to control the beam in the elevation plane. This might increase the complexity and cost, whilst needing the satellite to have the solar panels facing the Earth to achieve full end-fire radiation. However, this would prevent efficient power harvesting from the Sun.
Some techniques have also been reported for planar Yagi-Uda miniaturization include meandering, C-shaped elements, and folded dipoles [28], [29], [30], [31]. Despite providing some design compactness, the size reduction was not significant, and in most cases, the number of directors were sacrificed due to space constraints resulting in low gain performances. On the other hand, a compact structure which maintained directivity was achieved using an intricate technique; i.e., parasitic interdigitated strips as in [32], which required some complex and design-specific loading considerations.
Advancing on these developments whilst providing an alternative design approach, we propose a miniaturized planar Yagi-Uda antenna structure based on low-cost PCB technology. The design achieves full end-fire radiation in the S-band and is fully integrated within the solar panel array, and this array extends from the PicoSat Chassis (see Fig. 1). In particular, structure meandering and artificial dielectric concepts were employed by loading the three directors of the planar antenna (see Fig. 3) in a 1-D sequence to achieve overall structure compactness whilst maintaining directivity. This advances on our preliminary simulation work in [18], which mainly reported the design of a Yagi-Uda-like structure using an array of microstrip patches for operation at 8 GHz for placement on the glass layer of a solar cell.
To the best of the Authors' knowledge, 1-D artificial dielectric concepts and meandering have not been effectively applied on fully planar Yagi-Uda antennas previously. These miniaturization techniques are newly adopted in this paper, mainly, in an effort to fit the proposed S-band end-fire antenna on the deployable solar panel array of the Unicorn-2 Picosat (see Figs. 1 to 3). These techniques are further described in Section II together with a comparison with a conventional Yagi-Uda design, where some important dimensions have been reduced by more than 65% whilst still maintaining an operating bandwidth of more than 6%. It will also be shown that our compact design achieves a 40% and 15% reduction of the total required antenna length and width, respectively, when compared to a more conventional and non-compact version. Simulated results will also be discussed in Section II, including the scenario were the antenna is in frees-space and integrated onto the Unicorn-2 PicoSat. Measured results are reported in Section III with a summary in Section IV.

II. ANTENNA DESIGN AND SIMULATIONS
Typical antenna performance specifications for EO downlink transmission in the S-band for the Unicorn-2 PicoSat commercial mission [7] are an optimum LP gain of 5 dBi and a minimum −10 dB impedance reflection coefficient bandwidth of 10 MHz (0.42%) [18].
For the Unicorn-2 PicoSat, the chassis has already most of its space occupied with other payloads [7]. Thus, the integration of a new antenna was limited to the solar module nearest the main body of the satellite of the deployable wings (see Figs. 1 and 4, inset). These solar modules were constructed using epoxy resin material, defining the antenna substrate. Also, due to the need to maintain a weight balanced PicoSat structure, removal of one of the solar cells on the other side of the chassis was required and a second end-fire antenna can be added (see Fig. 1). Also, when the solar cell is removed and the epoxy substrate is used for antenna integration, the available footprint [7] is 43.80×90 mm 2 with a substrate thickness of 0.4 mm. This defines the available footprint for our miniaturized end-fire antenna which must operate at 2.4 GHz for ground station connectivity. Using  FR-4 material enables low-cost experimental demonstration and proof-of-concept for the space-ready antenna.

A. DESIGN CONSIDERATIONS
This antenna is composed of several parasitic microstrip dipoles [25] acting as directors, which are typically placed at a distance of 0.15 − 0.3λ 0 from the driven element and other directors [33]. Classic and well known formulas for Yagi-Uda antennas [33] were initially used to obtain the parameters of the non-miniaturized version (see Fig. 2 (b)) which is the starting point in our design process. Relevant dimensions are outlined in Figs. 2 (a) and (b) and Table 1. The antenna is fed by a 50-microstrip line and has been modelled and simulated in CST [34].
In order to achieve a higher gain, more directors are exploited. However, the length of the available footprint is very limited (90 mm or 0.72λ 0 where λ 0 is the free-space wavelength at the design frequency of 2.4 GHz) and cannot accommodate many directors to comply with the required 5 dBi realized gain requirement. Given this space available, the substrate losses whilst considering conventional design approaches, the maximum gain achievable is 3 dBi. Also, size constraints do not permit a fourth director to increase gain. Nevertheless, this would still not comply with gain requirements for the Unicorn-2 PicoSat mission. In addition, to improve the matching, the feeding microstrip line L m of the non-miniaturized version should be larger than 0.1λ 0 (optimally around 0.25λ 0 , as in [18]). Moreover, the length of the driven element L dr needed to operate at 2.4 GHz, and it could not fit within the substrate width (W sub = 43.80 mm). For these two reasons and to comply with the antenna requirements, several miniaturization techniques were adopted.
Adapting the driven element into the constrained width W sub was achieved using meandering and its length L dr was physically miniaturized to maintain operation at 2.4 GHz. This meandering reduced the width of the proposed Yagi-Uda by about 15%. On the other hand, the distance between directors D can also be reduced by applying artificial dielectric concepts [24], realized by local printed inclusions. This achieves a representative 1-D configuration for guided wavelength reduction along the y-direction and near the driven and director elements. Accordingly, the surface reactance near the printed metallic segments is locally tailored to achieve a higher effective relative permittivity, as in [24], [35], [36], allowing for the reduction of the distance D. For example, the minimum (D = 0.15λ 0 ) required for a good performance [25], [33] can be shortened by 60% to 0.06λ 0 , and this allowed for more space for the required directors.
This technique also allowed for an improvement in the front-to-back ratio, increasing the directivity with just three directors [24]. Basically, the surface reactance is modified as desired by the addition of a finite periodic grid of metallic strips in the y-direction (see Fig. 2 (a)). The length of these metallic strips L AD was kept constant for all directors. This is because the width W ADn is the main factor influencing the capacitive coupling and thus increasing the effective relative permittivity near the driven and parasitic elements. It should also be mentioned, that during our many optimisations and simulation studies using CST microwave studio [34], it was found that the best antenna performance (in terms of matching and directivity) was achieved when W AD1 , W AD2 , and W AD3 were positioned near the directors.
The final parameters for the miniaturized antenna can be found in Table 1, where they are also compared to the original, non-miniaturized structure. It can also be observed that the parameter L yagi was miniaturized, requiring a length of 0.35λ 0 which is about a 60% reduction when compared to the conventional version (0.58λ 0 ). Also, due to the improvement of the front-to-back ratio and the meandering of the driven element, the reflector implemented on the ground plane can be further reduced to a third in width W ref and a half in length L ref when compared to the non-miniaturized version (see Table 1). This supported further compactness.  The two designs are also placed side-by-side in Fig. 3 for size comparison.

B. SIMULATION RESULTS
A comparison of the reflection coefficients for the nonminiaturized and the miniaturized antennas is reported in Fig. 4. The miniaturized design offers a better impedance matching (|S 11 | ≤ −15 dB) although the −10 dB bandwidth over the required S-band operating frequency range is very similar for both cases and compliant with specifications required for the Unicorn-2 PicoSat.
In Fig. 5, the simulated realized gain values at full end-fire (θ = 90 • , φ = 90 • ) are reported for the two configurations. The standard Yagi-Uda antenna provides a peak gain of around 3 dBi while, thanks to the improvement of the front-to-back ratio due to the controlled surface reactance, the miniaturized antenna has a simulated realized gain of 5.1 dBi, also complying with the requirements. Additionally, the distance D between the directors in the non-miniaturized case has been reduced to the dimensions of the miniaturized antenna (from 0.15λ 0 to 0.06λ 0 ) to show the difference in performance and the need to include the reactive loading of the 1-D artificial dielectric structures for efficient radiation in a constrained footprint. It is shown in Fig. 5 that the realized gain is even lower than in the standard case as it does not comply with the required theoretical values for D, which is usually 0.15 to 0.3λ 0 due to the effective coupling between directors. The performance comparison between the miniaturized and non-miniaturized version is reported in Table 2.
Furthermore, the miniaturized antenna has been simulated on the lossy polyamide PicoSat chassis (see Fig. 1 and Fig. 4, inset), where a UFL connector is used for surface mount feeding. Initial analysis were carried out in [18], where the antenna was simulated next to a PEC satellite body. This deviated the beam towards θ = 100 • and reduced the gain to 3 dBi at full end-fire. The material of the body has been updated according to Alba Orbital instructions and lossy polyamide carbon fiber has been used. As shown in Fig. 4, the antenna offers reflection coefficient values up to −25 dB with a slightly diminished bandwidth of 5.13%. In this case, the beam is still pointing towards full end-fire (θ = 90 • , φ = 90 • ) with gain values of 4.3 dBi as depicted in Fig. 5. There is a reduction of around 0.7 dBi with respect to the non-integrated scenario. This degradation in gain is caused by the addition of the connector and the influence of the body of the satellite and the metallic surroundings. This loss could be compensated by the addition of another director in the space available.

III. EXPERIMENTAL RESULTS
The miniaturized Yagi-Uda S-band planar antenna was manufactured using an FR-4 substrate ( r = 4.4) having a thickness of 0.4 mm, and dimensions of 43.80 × 50 mm 2 for low-cost experimental testing and simple integration with the existing FR-4 solar panel array. It should be mentioned that FR-4 dielectric properties may not be stable in harsh environments such as space and could degrade antenna radiation characteristics. Basically, non-space qualified FR-4  PCB laminates may not maintain stable physical properties at extremely low or high temperatures. Further design work using space qualified TMM4 laminates from Rogers, which have low thermal variations and similar electric properties to FR-4, together with a metal coating additive (e.g., Alodine 1200 to help with corrosion resistance), could be used in future to maintain performance making the design more suitable and robust for space and other harsh operating environments. This PCB antenna is depicted in Fig. 6 (a) and the measured reflection coefficient results are reported in Fig. 7. There is a shift in frequency towards 2.45 GHz due to the addition of an SMA connector. Simulations have been repeated including the connector and, as it is shown in Fig. 7, the minimum of the reflection coefficient for the antenna is shifted to 2.45 GHz. Regardless, simulations and measurements are in good agreement (when including the connector in the simulation model) with |S 11 | ≤ 20 dB with a measured −10 dB impedance bandwidth of 6.25%.
Far-field measurements have been carried out in an anechoic chamber. It should be noted that the far-field positioner had an impact on the radiation pattern measurements shifting the resonance frequency for the compact antenna (similar to [38]). This is specific to the measurement setup available, and could be avoided in other measurement facilities. Polar plots, comparing the simulated and measured results for the miniaturized antenna from 2.45 GHz to 2.55 GHz, are reported in Fig. 8 where the simulation model is shown for reference. The expected performance is observed for the realized gain (Figs. 5 and 8) with a loss of 0.3 dB (4.76 dBi) at the maximum peak realized gain frequency of 2.5 GHz (when compared to the simulated case). Also, the radiation pattern displays a broad-beam shape. However, the main lobe is maximum along the axis of the antenna (θ = 90 • ) (see axis definition in Fig. 8) while gain values of −3 dBi are achieved at broadside, indicating end-fire performance. The front-to-back ratio is about 3.5 dB. Results for the crosspolarization levels are also shown and are below −25 dB at the end-fire and back-fire directions. A summary of the measured values with a comparison with simulation results has been reported in Table 3.
A new prototype has been also manufactured to be fully integrated on the deployable wing of the PicoSat Unicorn-2. This prototype is depicted in Fig. 6 (b). As it can be observed, a defected ground plane has been added at the back of the design to emulate the effects of a solar cell. The simulated and measured results for the co-and crosspolar radiation patterns of the antenna integrated on the chassis at 2.4 GHz for the φ = 90 • plane are depicted in Figs. 9 and 10. Both polarizations agree well, showing maximum gain values of 4.9 dBi for the co-polar component and a improved front-to-back ratio of 10 dB. Cross-polar levels are higher with respect to the non-integrated version (Fig. 8) due to the additional metallic bodies surrounding the compact antenna.
A comparison has been included in Table 4, where the advantages of our proposed antenna are highlighted when compared to other state-of-the-art compact designs. Most of  these works are based on WLAN and similar applications, however existing Yagi-Uda antennas for CubeSats [26], [27] have also been included in Table 4 to highlight the novelty in our integration approach for small satellites (in that the existing solar cell array of the Unicorn-2 PicoSat spacecraft is exploited) without requiring independent deployable systems for just the antenna. It should also be noted that the proposed three-director design achieves a competitive size when compared with systems of only one or no directors [29], [30], [31], [32], without sacrificing performance, bandwidth, or involving complex parasitic loading configurations such as parasitic interdigitated strips. Moreover, these previous works achieve the miniaturisation in one dimension only, while our proposed design is reduced in length and width by combining 1-D artificial dielectric loading and structure meandering.

IV. CONCLUSION
A planar miniaturized end-fire Yagi-Uda for PicoSat or other SmallSat integration was proposed which can support placement within deployable solar panel arrays. Miniaturization techniques based on artificial dielectric concepts and meandering dipoles have also been employed to reduce key antenna parameters, such as the distance required between directors or the length required for the driven element to operate at 2.4 GHz, from 0.15λ 0 to 0.06λ 0 and from 0.21λ 0 to 0.17λ 0 respectively, when compared to a more conventional planar Yagi-Uda antenna. Regardless of this reduction, RF performance of the antenna is still maintained and with an improvement in the front-to-back ratio that allows for higher gain. Also, the minimum length of substrate to accommodate the antenna is reduced from 0.58λ 0 to 0.35λ 0 .
As reported in the paper, efficient antenna performance was achieved by complying with the restricted dimensions of the FR-4 substrate under the solar cells. Measured realized gain values are about 5 dBi which are in agreement with the simulations for the antenna fully integrated on the satellite chassis. Performance of the compact antenna could be further improved by the addition of a fourth director and artificial dielectric element for compactness. This low-cost planar antenna can also be useful for CubeSats and other SmallSats, and generally where compact and low-profile antenna elements are needed to generate end-fire radiation.
TOM WALKINSHAW is the Founder of Alba Orbital, a company focused on building prototype PocketQubes, started in 2012. He has more cumulative years developing PocketQubes than anyone else in the world and has led Alba to its position as the leading company in its field. He has won many awards and accolades, including the The Glasgow Caledonian University Alumni of the Year in 2019, 39th Coolest person in U.K. and Forbes 30 under 30, to name a few.
CONSTANTIN CONSTANTINIDES received the M.Sc. degree in signal and image processing in 2008, and the Ph.D. degree in image processing in 2012. He is an RF and Signal and Image Processing Engineer. After having studied in France, he went to Heriot-Watt University, Edinburgh, U.K., to work as a Researcher on leaky-wave and compact antennas starting in 2015. He worked in space-related projects, and then moved to Alba Orbital, Glasgow, a company that designs picosatellites, to work as an Engineer in 2017. He currently works on designing and optimizing antennas for picosatellites as well as working on satellite night imaging techniques.