Shared-Aperture Stacked-Patch Electronically Steered Antenna (ESA) for Ku-Band Low Earth Orbit (LEO) Satellites

Table of Contents

1. Introduction 2. Problem Definition 2.1 Hardware Inefficiency in Current ESA Architectures 2.2 Protocol Fragmentation and Vendor Lock-in 2.3 Scalability and Future-Proofing 2.4 Problem Statement 3. Value Proposition 3.1 Value Proposition Canvas 3.2 Value Proposition Statement 4. Design Statement/Methodology/Strategy 4.1 Design Statement 4.2 Design Methodology 4.2.1 Literature Review 4.2.2 Competitor Analysis 4.3 Design Strategy 5. Concept Design 5.1 Concept Selection 5.1.1 Morphological Chart 5.1.2 Selection Criteria 5.2 Design Solution 5.2.1 Stack Architecture 5.2.2 Material Selection 5.2.3 Dimensioning of Key Parameters 6. Prototype & Testing 6.1 Prototype 6.1.1 Unit-cell Design and Simulation 6.1.2 Finite Array Design and Simulation 6.2 Testing 6.2.1 Fabrication of Array 6.2.2 Testing of Array 6.2.3 Test Results 7. Analysis 7.1 Simulaton vs Fabricated 2x2 Subarray 7.2 Scaling Projections for 32x32 Array Architecture 8. Shortcomings and Future Work 9. Conclusion Appendices Appendix A Appendix B Appendix C Appendix D Appendix E Appendix F References

1. Introduction

Geostationary Earth Orbit (GEO) satellites operate at ~35,786km above Earth’s equator, appearing fixed relative to the ground. This allows the use of fixed, non-tracking ground antennas[1], simplifying infrastructure[2].

Low Earth Orbit (LEO) satellites operate much closer to Earth (~160-2,000km) and travel rapidly (~90-120minutes/orbit)[1], enabling high-throughput and low-latency services[2]. As a result, ground terminals face challenges like frequent hand-overs, rapidly varying geometry, Doppler effects, and tracking needs[5]. The growth of LEO constellations is projected to reach tens of thousands in the coming decade to deliver global coverage[6], thus pushing demand for low-complexity, mass-manufacturable, and high performance user terminals[7][3][8].

Phased-array electronically-steered antennas (ESAs) provide a compelling solution[3]. Unlike mechanical dishes, ESAs steer beams via phase-shifting/beamforming across an array without moving, enabling rapid beam switching, a low profile, multi-satellite/multi-beam capability, reduced wear and maintenance, and improved links latencies/dynamics[3][8][9][10]. Figure 1 substantiates their increasing commercial demand.

In Singapore, ST Engineering Advanced Networks and Sensors (STE ANS) are establishing leadership in 5G and satellite communications. Their work in radar, satellite, and ground antenna manufacturing underlines Singapore’s ambitions in this sector. This report documents our collaboration with STE ANS to develop a pioneering ESA terminal for LEO constellations.

2. Problem Definition

2.1 Hardware Inefficiency in Current ESA Architectures

Traditional ESA terminals use separate apertures for transmit (Tx) and receive (Rx), resulting in larger physical footprints, higher size, weight, and power (SWaP), and cost[5][15]. Moreover, achieving necessary performance like wide scan angles, high gain, and circular polarization to mitigate fading and polarization mismatch, is highly challenging[16][17]. The demand is further accentuated in mobile or rooftop terminals where size and manufacturability matter[3][8].

2.2 Protocol Fragmentation and Vendor Lock-in

LEO operators often employ non-standardized and proprietary protocols[5][18]. This forces terminal manufacturers to design provider-specific hardware and modem integrations, restricting interoperability[3][18]. As a result, customers are locked into single-vendor ecosystems, reducing flexibility and driving up costs[3].

2.3 Scalability and Future-Proofing

End-users and integrators increasingly require scalable, interoperable terminals that operate seamlessly across networks. Current solutions lack this versatility, creating a gap in the market for a high-performance, multi-orbit ground terminal, which adapts to evolving standards and supports multiple service providers, enabling long-term viability and return on investment[3][8].

2.4 Problem Statement

This leads us to our problem statement:

3. Value Proposition

3.1 Value Proposition Canvas

Our value proposition canvas demonstrates how our shared-aperture, stacked patch ESA design will align with real customer needs, from terminal developers/integrators to commercial users.

3.2 Value Proposition Statement

Our solution directly addresses industry challenges by integrating both Tx and Rx functions within a single, co-located aperture, significantly reducing SWaP while maintaining high link reliability and polarization purity.

This inherently interoperable design is software reconfigurable for different operators, frequency plans, and multi-orbit constellations, minimizing hardware redesign[3].

Furthermore, it integrates seamlessly with commercial modem/beamforming IC, eliminating vendor lock-in[3]. Operating across a relatively wide Ku-band also supports multi-operator operation[3]. Moreover, our scalable ESA array allows flexible adaptation to varying performance and cost requirements.

This leads us to our value proposition statement:

4. Design Statement/Methodology/Strategy

4.1 Design Statement

To identify industrial LEO ESA terminal requirements, interviews were conducted with ground terminal developers (STE ANS and Astrolynk Technologies Inc.). Using a “Demand & Wish” table, the requirements are mapped to our antenna architecture.

The table below details the technical requirements based on the system and user needs.

4.2 Design Methodology

4.2.1 Literature Review

The following subsections review the technical foundations that justify and realize the project’s choice of a shared-aperture, stacked-patch ESA.

4.2.1.1 Shared-Aperture

Shared-aperture antennas co-locate multiple functions within a single panel, eliminating separate arrays. This is highly attractive for platforms with constrained SWaP[20][21]. Successful implementations rely on isolation techniques to minimise cross-band interference[21].

4.2.1.2 Stacked-Patch

Stacked-patch antennas use multiple radiating layers to increase impedance bandwidth, improve axial-ratio (AR) bandwidth for circular polarization, and elevate realized gain without significantly raising the profile[24][25][26][27]. To further improve the performance, several structural enhancements such as air gap layer, parasitic patches, and vias could be introduced into the stack (refer to Appendix B). Our novel design stacks the Tx patches above the Rx’s, further decreasing the footprint[3][8]. The stacked patches form a dual-resonant structure which could radiate at both Tx and Rx frequencies, achieving a wide impedance bandwidth and a smoother reflection coefficient (S11)[48][49].

4.2.1.3 Circular Polarization (CP)

CP is widely used in satellite links because it mitigates polarization mismatch between moving platforms/satellites[33]. By putting Tx above Rx patches, opposite-sense CP (Left-Hand and Right-Hand CP) is required instead of linear polarization to ensure isolation and prevent cross-interference between the two paths[24][27].

4.2.1.3.1 Single-Feed

From Figure 10(a), a square patch with truncated corners introduces perturbations that generate two closely spaced, equal-amplitude orthogonal modes[34]. A single offset feed achieves the necessary 90° phase difference for CP generation, though at the expense of narrower bandwidth[34].

4.2.1.3.2 Dual-Feed

This method excites two orthogonal linear modes in the patch using dual feeds. It provides excellent CP purity and wider CP bandwidth, but increases complexity and feed loss[34].

4.2.1.4 Feed

Patch antennas require a feeding mechanism to deliver RF power effectively. The excitation method directly affects impedance matching, bandwidth, polarization, and fabrication complexity[35][36]. The feeding can be achieved through:

4.2.1.4.1 Contacting Methods

A coaxial probe/inset feed that connects directly to the patch. This is simple and offers decent impedance matching, but may cause spurious radiation, mechanical stress, and a narrow bandwidth[35].

4.2.1.4.2 Non-Contacting Methods

This method uses aperture coupling, where a microstrip feedline beneath the ground plane couples energy to the patch via a slot. This isolates the feed from the radiating element, improving bandwidth, CP performance, and easier impedance/polarization control and integration with multilayer designs[24][35], though more complex fabrication[36].

4.2.2 Competitor Analysis

From Table 4, our proposed design targets are aligned with real-world system requirements, and thus are ambitious relative to commercial offerings, with our topology leading to low SWaP being the key differentiation[3][8].

4.3 Design Strategy

The table below summarises our design strategy.

5. Concept Design

5.1 Concept Selection

5.1.1 Morphological Chart

To systematically identify and compare possible designs, a morphological chart was constructed.

5.1.2 Selection Criteria

5.1.2.1 Route A

Although simple and compact, this configuration suffers from limited impedance bandwidth (<10%) due to probe inductance and increased dielectric loading at Ku-band frequencies[35]. The vertical probe also introduces unwanted asymmetry, leading to cross-polarization degradation[40]. Moreover, mechanical drilling of coaxial probes at mm-level accuracy complicates array fabrication and integration into planar PCBs[3].

5.1.2.2 Route B

The truncated corner patch (TCP) with single offset feed offers a simpler feeding structure and lower insertion losses[34]. However, this structure provides limited control over orthogonal modes, making it more challenging to achieve robust CP and maintain a stable AR across wider bandwidths, especially under array scanning conditions[64].

5.1.2.3 Route C (Chosen)

The dual-feed scheme is selected due to its superior control of orthogonal modes, enabling more stable CP and improved AR performance across the operating band[41][42]. The benefits outweigh the added feed complexity and potential insertion losses[42]. Together with the stacked-patch structure, aperture coupling enables better isolation between the feed and radiating layers, allowing robust multilayer integration[3][43]. Moreover, aperture coupling offers wide impedance bandwidth, reduced spurious radiation, higher CP performance, and preserves the mechanical integrity of radiating patches[3][24]. Thus, route C presents the best balance between polarization performance, impedance and AR bandwidths, manufacturability, and scalability for our ESA.

5.2 Design Solution

5.2.1 Stack Architecture

The proposed stacked geometry consists of multiple layers arranged vertically. Table 7 shows the order of top to bottom layer of our ESA. Additional layers (Appendix B) will be added to further optimize our design.

5.2.2 Material Selection

Commercially, the material used for patches, feeds, and ground planes is copper, while the dielectric substrate material must exhibit low dielectric loss, stable εr, and thermal consistency[3]. Commercial laminates need to be evaluated in terms of thickness, manufacturability, and cost to enable optimization and mechanical rigidity[3].

Rogers RT/duroid 5880 is selected as the final dielectric substrate due to its low dielectric constant (εr=2.2), which supports wideband operation and improves impedance matching[5][8]. Its low loss tangent also enhances radiation efficiency at high frequencies[3]. Figure 16 shows its properties and its range of commercially available thicknesses.

5.2.3 Dimensioning of Key Parameters

5.2.3.1 Patch

A square patch offers easy polarization control and stable radiation pattern symmetry[54]. It simplifies dual-band scaling[54], and is standard PCB-compatible for easy fabrication and arraying[3]. Its well-studied mode structure (TM₁₀ and TM₀₁) also ensures highly predictable performance and modeling accuracy[3][54].

"

For a square patch, the resonant length, L, is λg/2[55]. Using the above 3 equations[56], we can find the length/width of patches, given the choice of substrate and its thickness. Given Tx patch is smaller than Rx (due to smaller λ0 as Tx resonates at a higher frequency), it would be placed above Rx in our stacked geometry[57].

5.2.3.2 Slot

Rectangular slots (Figures 18(a), (b), and (c)) provide simple coupling, while 'H' and bowtie slots (Figure 18(d) and (e)) offer stronger coupling due to their more uniform field magnitude[59]. Combining these 2 into the hourglass-shaped slot (Figure 18(f)) achieves the most uniform field distribution and strongest coupling[59], which is a design we plan to test.

5.2.3.3 Feed

A straight microstrip feedline (refer to Figure 13) is chosen as the excitation mechanism for the patch, because it provides clean coupling, broad bandwidth, and structural simplicity[59]. Thinner feed lines couple more strongly to slots[24].

5.2.3.3.1 Feed offset

Feed offset refers to the lateral displacement of the microstrip feed from the patch’s centerline, controlling coupling strength and mode balance to generate CP. For square patches, the offset feed point is on the diagonal[60].

5.2.3.3.2 Stub

The stub length is the distance between the slot’s center and the feed line’s open end. This section behaves as a reactive tuning stub[61]. Tuning the stub’s length and width allows impedance fine-tuning for resonance matching[61].

6. Prototype & Testing

This project uses CST Studio Suite 2025 to design, optimize, and validate a stacked-patch antenna element, then scale it into a finite 2×2 array before fabrication. The workflow is split into three stages:

1. Single element simulation.

2. Finite 2x2 array simulation.

3. Finite 2x2 array fabrication and testing.

6.1 Prototype

6.1.1 Unit-element Design and Simulation

6.1.1.1 Boundary Conditions and Excitation Setup

Periodic Boundary Conditions (PBCs) are applied on the X- and Y-axes to mimic an infinite array, while the Z-axis uses Open boundaries to simulate free-space radiation.

The feed excitation is introduced through a waveguide port at the feedline input, matched to 50Ω.

6.1.1.2 Building of Unit-element

The prototyping process was carried out in three progressive stages to ensure systematic validation and controlled integration of complexity. To achieve the aforementioned array footprint in Table 3, each unit-element is designed with a target dimension of 10mm by 10mm. The design evolved from a single-patch configuration to a stacked dual-patch structure, and finally to a dual-feed circularly polarized antenna element.

Stage 1: Single Patch with Offset Feed and Slot

The first stage focused on developing a single-patch aperture-coupled antenna to establish baseline performance and fundamental design understanding. Each layer and feature from Table 7 was constructed systematically and parameterized using variables to enable efficient tuning and optimization in future iterations.

At this stage, only the Tx patch was implemented, targeting a single resonant frequency at 14.25GHz. An hour-glass slot was implemented.

The feedline and slot was offset from the patch center to achieve proper impedance matching and effective excitation of the patch’s dominant mode.

The microstrip feedline width was also tuned to ensure a 50Ω characteristic impedance.

Copper components (ground planes, feed, patch) were set to a thickness of 0.035mm, according to standard manufacturing practices.

The dielectric substrate selected at this stage was RO4450B (εr=3.54) with a thickness of 0.5mm. Standard manufacturing practices were incorporated by setting the substrate thickness between the lower ground plane and the microstrip feedline, as well as between the microstrip and the slotted ground plane, to 0.203mm. The last layer of substrate was set to 0.138mm, insulating the lower ground plane.

Using the Parameter Sweep Tool, parametric sweeps were conducted to optimize key design parameters, including patch dimensions, stub length and width, feed offset, and slot geometry, while maintaining sufficient geometric constraints to accommodate a future addition of orthogonal feed for CP.

The resulting S₁₁ response demonstrated satisfactory impedance matching (<-10dB) at 14.25GHz, validating the initial design approach.

Stage 2: Dual-Patch Stacked Configuration

In the second stage, the design was extended to a stacked configuration by introducing a Rx patch beneath the Tx patch, enabling dual-band operation. The vertical separation between the two patches was set to 0.5mm. This configuration allowed the two patches to resonate at different frequencies. However, the impedance bandwidth was not wide enough to cover both Tx and Rx bands (10.9GHz-14.5GHz).

To improve bandwidth, the substrate material was changed to RT/duroid 5880, which has a lower dielectric constant (εr=2.2), and the thickness was increased from 0.5mm to 0.79mm.

Rigorous parametric optimization was then performed on key parameters using both the Parameter Sweep Tool first, then Optimizer Tool, and the most optimized S₁₁ was achieved as shown below.

Given that the target resonant frequencies are 14.25GHz for Tx and 11.7GHz for Rx, from Figure 32, the design is unable to meet the target Tx as the S₁₁ went above −10 dB beyond 14.1GHz.

Moreover, prior to obtaining results as shown in Figure 32, it was observed that the hour-glass slot occupies a relatively large area near to the patch centre, which may pose challenges when incorporating an additional orthogonal feed and slot subsequently. Hence, alternative slot geometries, including bowtie and H-shaped slots, were explored.

Among these, the H-slot demonstrated better performance in terms of impedance matching, bandwidth, and greater positional flexibility, as it can be more significantly offset from patch centre, which would minimise cross-polarization between the two feeds subsequently.

Parametric optimization was then performed to further fine-tune the design parameters, resulting in improved S₁₁ characteristics.

Although dual resonant frequencies were not observed, the S₁₁ remained below −10 dB across both the operating bands, which is considered acceptable given that only a single feed is employed.

Stage 3: Dual-Feed Circularly Polarized Configuration

In the final stage, the antenna was upgraded to a dual-feed configuration to enable CP. An additional orthogonal feedline and corresponding slot were introduced to excite two orthogonal modes of the patches.

Careful consideration was given to the spacing between slots to reduce cross-polarization degradation.

Vias were also incorporated between the two ground planes to improve electromagnetic confinement and suppress surface waves[65][66]. A diagonal configuration was employed to introduce a controlled perturbation along the patch’s diagonal, supporting CP while facilitating effective mode control and frequency tuning[67].

Extensive parametric tuning was carried out on the patch dimensions, slot geometries, feed offsets, stub lengths, and via configurations. The Optimizer Tool was then used to achieve optimal performance in terms of S₁₁, S₂₁, AR, and realized gain.

The S₁₁ demonstrates dual-resonance behavior with a primary resonance near 11.2GHz and a secondary around 14.1GHz. The -10dB impedance bandwidth covers both Tx and Rx bands (10.9–14.5GHz), confirming the antenna is well-matched.

For S₂₁, the design achieves excellent port-to-port isolation, remaining below -15dB across the whole band. This minimal mutual coupling between the dual feeds ensures that phase quadrature is preserved, directly supporting the realization of high CP purity and cross-polarization discrimination (XPD)[3][8].

To validate the dual polarization capabilities, we isolated the excitation of each port while terminating the other. Results were taken at Rx resonant frequency (11.7GHz) and Tx’s (14.26GHz).

When Port 1 is active, the unit-element produces a stable, directive radiation pattern with a peak realized gain of 4.21dBi. In the ϕ=0° plane, the 3dB angular width is 97.9° (from -51.8° to 46.1°), while in the ϕ=90° plane, it remains broad at 91.6° (-43.0° to 48.6°). This wide beamwidth is essential for achieving the target ±53° scan volume in the full-scale array. Furthermore, the cross-polarization level (GainTheta in ϕ=0° and GainPhi in ϕ=90°) is suppressed to -12.0dBi, representing an XPD of 16.2dB (4.21-(-12.0)dBi) at boresight.

Port 2 performance closely mirrors Port 1, confirming the structural symmetry of the dual-feed architecture with a consistent peak gain of 4.22 dBi. The 3 dB beamwidth measures 91.5° (ϕ=0°) and 98.5° (ϕ=90°). The inversion in angular width between the two ports is expected, as the E-plane and H-plane orientations rotate by 90° between the vertical and horizontal feed ports. The isolation remains robust, with cross-polarization suppression levels identical to Port 1, ensuring minimal interference between the two signal paths.

At 14.26GHz, Port 1 excitation yields a peak realized gain of 4.30dBi. The radiation patterns maintain a broad profile, with 3dB beamwidth of 88.9°(ϕ=0°) and 77.38° (ϕ=90°). While the cross-polarization isolation remains robust, the XPD decreases to ~10dB at this higher frequency.

Port 2 exhibits high symmetry to Port 1, achieving a peak realized gain of 4.60dBi. The 3dB beamwidth measures 77.13° (ϕ=0°) and 89.69° (ϕ=90°). This expected rotation in principal planes between ports confirms that the feed structure is delivering orthogonal linear polarizations with nearly identical power envelopes. The consistency in gain and beamwidth across both ports is essential for generating high-quality CP when combined with a 90° phase shift[8].

Overall, a comparison between the 11.7GHz and 14.26GHz results shows that the unit-element maintains a stable gain profile across both Rx and Tx bands, with less than 0.4dBi deviation between the two resonant frequencies. The persistent symmetry between the vertical and horizontal feed ports across the entire operating spectrum validates the unit-element as reliable and predictable.

Both ports were then excited simultaneously with a 90° phase shift to produce CP. As shown in Figure 47, the unit-element maintains the standard high-quality threshold of AR<3dB across 10.9GHz to 14.26GHz. This confirms the aperture-coupled feed is highly symmetrical and well-matched for the Rx band (11.7 GHz). However, the AR ascends sharply beyond 13.6GHz, reaching the 3dB limit at the 14.26GHz Tx resonance, which directly correlates with the aforementioned lowered 10dB XPD.

As a final refinement, patch, stub, and slot geometries were filleted with a 0.4mm radius to improve manufacturability and reduce fabrication sensitivities at sharp discontinuities[3][8]. The design was subsequently re-optimized to ensure that performance was preserved following these modifications. Below shows the final dimensions of the unit-element before scaling to an array.

6.1.2 Finite Array Design and Simulation

6.1.2.1 Boundary Conditions Setup

Open Boundary Conditions are applied on all the X-, Y-, and Z-axes to simulate free-space radiation.

6.1.2.2 Building of 2 x 2 array

The transition from a single unit-element to a 2x2 planar array was implemented by arranging the four elements in a symmetrical grid as shown below.

6.1.2.3 Performance of 2x2 array

The peak realized gain at boresight increased to 8.79dBi, a ~4.6dB improvement over the unit-element (~4.2dBi). This aligns with theoretical array factor summation, accounting for minor mutual coupling and feed network losses. Consequently, the beamwidth has sharpened to 61.9° in the ϕ=0° plane and 62.8° in the ϕ=90° plane. This enhanced directivity is a critical baseline for achieving the high EIRP (Equivalent Isotropically Radiated Power) required for the scaled 32x32 array.

The array also exhibits good polarization health. The cross-polarized component (LHCP), represented by the lower curves in both figures, is suppressed to -10.36dBi at boresight. This results in an XPD of 19.15dB, indicating that the feeding network and the 90° phase-shift mechanism maintain high precision across all four elements. Furthermore, the symmetry between the ϕ=0° and ϕ=90° planes ensures a stable phase center and minimal squint, which is essential for maintaining a low AR across the main lobe.

At 14.26GHz, the array achieves a peak realized gain of 10.19dBi at boresight, representing a notable increase. This gain enhancement is accompanied by further beam narrowing (52.49° in the ϕ=0° plane, 55.02° in the ϕ=90° plane), validating the array's ability to focus energy effectively.

The array maintains a high degree of LHCP polarization purity. The cross-polarized RHCP component is suppressed to -12.79dBi, yielding a XPD of 22.98dB. However, the divergence between the RHCP and LHCP envelopes at wider angles suggests increasing ellipticity toward the beam edges, where circularity is slightly degraded. This behavior aligns with the previously established unit-element AR.

From Figure 58, the array maintains excellent CP purity, with the boresight AR remaining below the critical 3dB threshold across the operating band.

Overall, the simulated performance of the 2x2 array confirms a successful transition from a unit-element to a subarray, providing a reliable foundation for hardware fabrication. At both the 11.7GHz(Rx) and 14.26GHz(Tx) frequencies, the array demonstrates stable radiation characteristics, significant gain enhancement, and well-defined beam narrowing.

6.2 Testing

6.2.1 Fabrication of Array

Following the simulation-based optimization, the 2x2 patch array was fabricated to validate its performance under real-world conditions.

Additionally, a corporate feeding structure, specifically the Wilkinson Power Divider (Appendix C), was integrated into the back of the PCB. The network is designed with two primary ports: Port 1 (S₁₁), which feeds the four vertical polarization components of the patches, and Port 2 (S₂₂), which feeds the four horizontal polarization components. To manage the phase and amplitude distribution required for beamsteering and CP, the array was integrated with a Beamforming Integrated Circuit (BFIC) too.

6.2.2 Testing of Array

The fabricated array was mounted in a far-field anechoic chamber to eliminate external interference and multi-path reflections. The testing setup involved a standard gain horn antenna as the transmitter and the 2x2 array prototype as the receiver, mounted on a precision multi-axis positioner. This configuration allowed for the high-resolution mapping of the S-parameters, realized gain, and CP across the targeted frequency spectrum. This comprehensive testing environment confirms the efficiency of the integrated feeding network and the functional readiness of the BFIC-controlled array for high-speed communication applications.

6.2.3 Test Results

The measured S-parameters confirm a robust dual-band resonance characteristic across the frequency spectrum. Both S₁₁ and S₂₂ were below -10dB across 10.9GHz-14.5GHz. Furthermore, the S₂₁ curve, representing the isolation between the vertical and horizontal feed structures, remains below -20dB across the entire operating range. This high level of isolation is critical for maintaining polarization purity and preventing signal leakage between the two orthogonal feeding paths.

At the lower resonance of 11.7GHz, the array exhibits highly symmetrical radiation patterns in both planes. The measured peak realized gain reaches ~8dBi, which aligns closely with the simulated results (8.79dBi). The slight drop is likely due to minor fabrication tolerances and dielectric losses in the physical substrate. The measured patterns show a slightly narrower beamwidth in the ϕ=0°, but a strong overlap between the simulated and measured curves, particularly at the boresight. The slight ripples observed in the measured traces at wider angles (θ=±60°) are typical of anechoic chamber measurements and reflect minor reflections or ground-plane edge effects.

Performance at the upper resonance (14.25GHz) remains stable, with measured gain peaking ~10dBi. While the patterns maintain a directive boresight beam, there is a slight observable deviation between the measured and simulated data at wider angles. Nevertheless, the array maintains a clean main lobe and well-suppressed side lobes. The consistency between the ϕ=0° and ϕ=90° planes further validates its ability to support wideband operation.

7. Analysis

7.1 Simulation vs Fabricated 2x2 Subarray

The measured S-parameters of the 2x2 subarray exhibit primary resonances at 11.5GHz and 13.85GHz (Figure 62), shifting slightly from the intended design targets. These deviations are primarily attributed to the cumulative effect of fabrication tolerances, specifically variations in the dielectric constant of the substrate and the precision of the aperture-coupled slot etching[3][8].

Despite these shifts, both S₁₁ and S₂₂ remain below -10dB across the operational bandwidth, ensuring efficient power transfer. Furthermore, the S₂₁ isolation remains <-20dB throughout the resonances, confirming that the structural shifts have not compromised the port-to-port decoupling required for stable CP. Moreover, at the 11.7GHz(Rx) and 14.25GHz(Tx) resonances, measured gains of 8dBi and 10dBi, respectively, show strong correlation with simulated models. These measurements validate that the design is resilient to minor manufacturing variances while maintaining the necessary impedance bandwidth and gain.

7.2 Scaling Projections for 32x32 Array Architecture

The transition from a 2x2 subarray to a full 32-element linear dimension implies a scaling factor of 16 in each dimension, totaling 1024 elements. A significant advantage of scaling to a 32x32 architecture is the transition from an edge-element dominant environment to an embedded element environment[3][8]. While the 2x2 subarray is limited by asymmetrical mutual coupling and fringe effects, the 1024-element array functions as a highly periodic structure. This uniformity allows the active reflection coefficient to converge to a steady state, resulting in improved and more predictable impedance matching. By mitigating edge-related discontinuities, the full-scale aperture will achieve enhanced S₁₁ and S₂₂ stability, provided active impedance is managed to prevent scan blindness during wide-angle steering.

7.2.1 Size, Weight, and Power Requirements

From Table 4, The proposed 32x32 phased array is required to meet a volume constraint of 320x320x30mm, a mass limit of 6.8kg, and a power envelope of 180W. At the unit-element level, the footprint is 10x10x2.3mm. The small vertical profile of the unit-element leaves ample headroom within the 30mm thickness for multi-layer PCB integration of the distribution network, cooling and control electronics. While the current research focused primarily on electromagnetic performance and beamforming validation, with less emphasis placed on the specific mass and power optimization of the control electronics and BFICs, the measured efficiency of the 2x2 prototype provides a positive outlook on meeting the weight and power requirements[3][8].

7.2.2 Tx and Rx Gains Requirements

Theoretically, increasing the array size by a factor of 16 provides a gain enhancement of approximately 10log10(256)=24dB. Based on the measured 8dBi from the 2x2 subarray, the 32x32 array is projected to achieve a boresight gain of 32dBi at 11.7GHz. Similarly, given the 10dBi measurement at 14.26GHz, the scaled Tx gain is projected to reach 34dBi. Referring back to Table 3 for the full array’s targeted Tx and Rx gains, these gain projections confirm that our design provides the necessary aperture efficiency to meet the high-gain requirements of LEO satellite links.

7.2.3 Tx EIRP Requirements

To accurately forecast the performance of the full array, the analysis must also account for the integration of active RF front-end components and layout constraints. For the Tx path, the EIRP is calculated by:

With a channel output power of ~9dBm, a projected Tx gain of 34dBi, and 1024 individual PAs (+30dB), the system achieves an overall EIRP of 73dBm (43dBW). This exceeds the target specification of 42.5dBW (Table 3), allowing for a healthy buffer against feed network losses.

7.2.4 Rx G/T Requirements

For the receive path, the G/T is calculated using the projected Rx gain (32dBi) and the noise performance of the selected LNA. With a Noise Figure (NF) of 1dB, the system is well-positioned to surpass the 6.8dB/K requirement (refer to Appendix E for calculations).

7.2.5 Scan Angle Requirements

The target scan volume of ±53° (Table 4) is a rigorous design constraint that requires the suppression of grating lobes and the maintenance of polarization purity at high offsets. Appendix F shows that the element spacing of 10mm is adequate to meet the maximum scan angle. Moreover, the high port isolation (S₂₁<-20dB) verified in the 2x2 prototype shows low mutual coupling which reduces the active impedance variations that typically cause scan blindness at wide angles. Furthermore, the broad-beam characteristics observed in the unit-element (91.5°-98.5°) ensure that the "element factor" does not prematurely attenuate the "array factor" during steering. At the maximum ±53° scan angle, the array will experience a projection-based scan loss of approximately cos(53°)=2.3dB. Given the projected boresight gain of 32-34 dBi, the system maintains sufficient headroom to ensure that the gain at the scan edge remains above the operational threshold for consistent satellite connectivity.

In conclusion, the 2x2 results provide a technically sound foundation for the fabrication of the full-scale 32x32 array. The stability of the measured radiation envelopes, combined with the successful integration of the 90° phase-shifting logic for RHCP and LHCP, validates the dual-band, shared-aperture architecture. By correlating the empirical subarray data with link budget requirements, it is demonstrated that the proposed ESA will meet EIRP and G/T targets while maintaining stable performance across a wide scanning volume, confirming its readiness for final assembly and experimental validation.

8. Shortcomings and Future Work

While the 2x2 subarray has provided a robust validation of our design, several technical limitations were identified.

A primary limitation identified in the 2x2 subarray is the degradation of polarization purity at higher frequencies. For future work, a re-optimization of design will be performed to balance XPD across the full bandwidth.

Furthermore, the transition to a full-scale 32x32 active ESA introduces severe thermal challenges. Dissipating <180W within a compact 32cmx32cm footprint necessitates a multi-layer PCB design with dense thermal vias and integrated heat sinks to prevent gain droop and phase instability.

Additionally, as the array scales, the cumulative phase and amplitude errors within the corporate feed network can lead to beam squint and increased sidelobe levels. To mitigate this, a digital beamforming backend will be integrated to allow for real-time calibration and adaptive null-steering. "Active element pattern" simulations are required to mitigate scan blindness caused by cumulative mutual coupling in the 1024-element environment.

To enhance scanning performance out to ±53°, future iterations will implement array tapering (amplitude weighting). While uniform excitation maximizes gain, it produces high sidelobes that can lead to interference. By applying a tapering profile, we can suppress sidelobes and improve the beam shape during wide-angle steering. This must be balanced against the resulting "taper loss" to ensure the 42.5dBW EIRP and 6.8dB/K G/T targets are maintained.

Finally, to manage the dynamic nature of LEO constellations, we are looking at integrating the digital beamforming backend with Machine Learning algorithms which reside within the BFIC logic to predict satellite trajectories and optimize beam-switching. This predictive approach is essential for handling the frequent handovers required as LEO satellites move rapidly across the sky, ensuring seamless connectivity and reduced latency for moving-platform applications. This intelligent backend will be critical for maintaining a resilient link in high-density LEO satellite environments.

9. Conclusion

In conclusion, this project presents a proof-of-concept of a high-performance 32x32 ESA that addresses the critical SWaP constraints of modern satellite ground segments by integrating Tx and Rx functions into a single, co-located aperture. By employing a software-reconfigurable, wideband architecture compatible with commercial BFICs, the design effectively eliminates vendor lock-in and protocol fragmentation. Furthermore, the future integration of array tapering and ML-driven handover logic ensures that this scalable solution remains interoperable and resilient across evolving LEO constellations, providing a long-term viable terminal for mobile/rooftop applications.

Appendices

Appendix A

Additional Literature Review

Ku-Band Operation

The Ku-band is strategically chosen for this project as it offers the optimal balance between antenna size, available bandwidth, and atmospheric performance[27]. Compared to lower bands like C-band, Ku-band allows for smaller antenna apertures and higher gain within the same physical footprint, reducing SWaP[30]. It also experiences less atmospheric and rain attenuation than higher bands like Ka-band, making it highly suitable for broadband LEO satellite communication systems where consistent link performance is critical[31].

The selected frequency allocations for our antenna are Rx: 10.7-12.7GHz and Tx: 14.0-14.5GHz. From Figures 42 and 43, these ranges are globally harmonised by the ITU for mobile satellite applications and are widely adopted by major LEO operators like Starlink and OneWeb[32]. This standardisation ensures that our design remains interoperable and compliant across multiple regions and service providers, addressing the lack of cross-network compatibility[3].

Dielectric Substrate

Dielectric substrate layers are used to house the patches and the feed. The substrate dielectric constant (εr) and loss tangent (tan δ) affect guided wavelength, patch size, bandwidth, and radiation efficiency[37]. Low-εr, low-loss laminates are commonly used in high-performance Ku antennas to maximise efficiency and reduce dielectric loading[3]. Other considerations include providing good manufacturability (blind via plating) at modestly higher loss[3], a trade-off between performance (loss/bandwidth) and cost/manufacturability.

Impedance Bandwidth

Impedance bandwidth (frequency range where |S₁₁| ≤ −10 dB) is a primary performance metric for antenna usability across required bands. Stacked-patch and aperture-coupled topologies have been shown to achieve wide impedance bandwidths, frequently ~20-40% in range, when parasitic resonances and aperture geometry are optimally tuned[24][26]. Practical Ku-band designs can therefore meet the Tx (14.0~14.5 GHz) and Rx (10.7~12.7 GHz) (combined 30.2% bandwidth) requirements[26][27], which require computational design and careful tuning of key geometries like air gaps, slot geometry, and parasitic layer dimensions.

AR Bandwidth

AR bandwidth (frequency range where AR ≤ 3 dB typically) defines the usable CP range critical for maintaining stable polarization alignment in dynamic LEO links[24]. Wide AR bandwidth ensures robust performance against polarization mismatch, Faraday rotation, and dynamic scan angles[27]. Stacked and aperture-coupled patches have been shown to significantly enhance AR bandwidth, often achieving 10-25% compared to <3% for single-layer designs[25][28].

Cross-Inteference

Cross interference arises from electromagnetic coupling between co-located Tx/Rx or dual-band sub-apertures, leading to signal degradation if not properly mitigated[20]. Techniques such as orthogonal polarization, frequency separation, via fencing, and feed isolation are proven to maintain a good (>30 dB) isolation in shared-aperture ESA designs[21][27].

Appendix B

1. Air Gap Layers

The air gap effectively lowers the effective dielectric constant (εeff) and reactive energy, and broadens Tx/Rx resonances, significantly improving impedance bandwidth (~25-35%)[46]. It also reduces mutual coupling for better isolation and minimizes surface-wave losses for higher efficiency and gain[47]. Its thickness must be tuned for optimization[47].

2. Parasitic Patches

The addition of coupled parasitic patch(es) forms a dual-resonant structure with closely spaced frequencies, which achieves a wide impedance bandwidth and a smoother reflection coefficient (S11)[48][49]. They redistribute current density, reduce surface-wave propagation and improve aperture efficiency[50]. Furthermore, it stabilizes CP performance[50].

3. Via Fences

Via fences act as electromagnetic shields to prevent surface-wave leakage, significantly reducing cross-polar radiation and enhancing polarization purity[23][51]. They also reduce back radiation and parasitic modes, thus stabilizing the impedance response across frequency[51].

Appendix C[68]

The primary distinguishing feature of this divider is the incorporation of an isolation resistor connected between the two output ports. This component is critical for dissipating reflected power and ensuring that the two output arms remain electrically isolated from one another, preventing signal leakage between channels. Because the potential difference across the isolation resistor is zero during balanced operation, no power is dissipated in the resistor under ideal conditions, maintaining high radiation efficiency for the overall system.

The performance of the Wilkinson divider is characterized by its port matching and high isolation levels. Technical evaluations indicate that while ideal schematic simulations suggest near-perfect splitting, physical fabrication introduces electromagnetic effects such as parasitic capacitance and edge effects that can slightly degrade the return loss and isolation. Despite these practical constraints, the Wilkinson power divider remains the preferred choice for antenna feed networks, such as those used in phased array radars, because it provides the stability and port-to-port independence required for precise beamforming and signal distribution across a large-scale aperture.

Appendix D[69]

Starlink Standard Kit Design which suggests the use of H-slots and array tapering design to improve scan performance.

Appendix E

1. Determining System Noise Temperature (Tsys)

The system noise temperature is primarily composed of the antenna noise temperature (Ta) and the receiver noise temperature (Tr).

A. Receiver Noise Temperature (Tr)

The noise performance of the LNA is given as a Noise Figure (NF) of 1 dB. To use this in a G/T calculation, we convert the NF to an equivalent noise temperature (Te) using the reference temperature T0=290K:

B. Total System Temperature (Tsys)

Assuming an antenna noise temperature (Ta) of approximately 100 K (typical for an Earth station antenna pointed at a satellite at Ku-band with some atmospheric and sidelobe noise), the total system temperature is:

2. Calculating G/T in Decibels

The G/T ratio is expressed as the antenna gain (G) minus the system noise temperature in dBK (10log10Tsys).

A. Convert Tsys to dBK

B. Final G/T Calculation

Using projected Rx gain of 32 dBi:

Projected G/T: 9.57dB/K

Required G/T: 6.8dB/K

Margin: +2.77dB

Even if we account for the thermal noise contribution of the distribution network (losses between the antenna elements and the LNA), which might add another 1-2 dB of noise figure depending on the layout, the system still comfortably exceeds the 6.8 dB/K target. With a margin of nearly 2.8 dB, the design is robust enough to handle the 2.3 dB scan loss expected at a 53 degree scan angle while remaining above the minimum required sensitivity.

Appendix F

Calculation of Element Spacing, d. θmax = 53° (refer to Table 4)

Calculating: d < 0.55λ0

Using formulas from Figure 17:

  • λ0,Tx = 21.05mm (f = 14.25GHz)
  • λ0,Rx = 25.64mm (f = 11.7GHz)

To avoid the grating lobes at maximum scan angle, the element-to-element spacing should be based on the shortest wavelength (the Tx frequency, 14.25 GHz). Hence, d < 10.5mm. In the 2x2 array, d = 10mm was used.

References