A deep learning-based pixelated reflectarray design method and reflectarray antenna

Abstract

The present application belongs to the field of reflectarray antenna design, and specifically provides a pixelated reflectarray design method based on deep learning and a reflectarray antenna, to solve the problems of long design cycle, high calculation cost and performance decline of the reflectarray antenna in the prior art. In the design process of the pixelated reflectarray, a deep learning agent model is used to replace full-wave simulation, which greatly shortens the optimization time, realizes fast reverse design, and significantly improves the performance of the pixelated reflectarray antenna obtained by design. In addition, topological smoothing constraints and physical connection constraints are introduced in the design process, which avoids the unprocessable microstructure, improves the manufacturing yield and robustness of the antenna.

Description

Technical Field

[0001] This invention belongs to the field of reflective array antenna design, specifically providing a pixelated reflective array design method and reflective array antenna based on deep learning. Background Technology

[0002] Among reflective array antennas, planar reflective array antennas have the advantages of being lightweight, high-gain, and low-cost, and are widely used in modern communication systems. Planar reflective array antennas consist of a feed and a reflective phase-shifting surface. To effectively illuminate the phase-shifting surface, a certain distance needs to be maintained between the feed and the reflector, resulting in a relatively high overall antenna profile, especially in large-scale, high-gain applications. Furthermore, the front-mounted feed introduces aperture obstruction, leading to a decrease in aperture efficiency. Compared to planar reflective array antennas, folded reflective array antennas introduce polarization-selective surfaces for primary reflection and secondary transmission of incident waves, reducing the profile height and minimizing feed obstruction. However, the design of the reflective array is crucial for both planar and folded reflective array antennas. Currently, existing technologies utilize pixelated units to increase design freedom, but these rely on multi-objective genetic algorithms, often requiring tens of thousands of full-wave simulation iterations, resulting in extremely long design cycles and failing to meet the demands of rapid design. Moreover, existing design methods often ignore physical constraints, leading to current singularities in the generated pixelated units, making simulation convergence difficult and hindering fabrication, thus degrading the performance of the reflective array antenna. Based on this, the present invention proposes a pixelated reflective array antenna based on deep learning and its design method, in order to shorten the design cycle and improve the performance of the pixelated reflective array antenna. Summary of the Invention

[0003] The purpose of this invention is to provide a pixelated reflective array design method and reflective array antenna based on deep learning, so as to solve the problems of long design cycle, high computational cost and performance degradation of reflective array antennas in the prior art.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] A deep learning-based pixelated reflective array design method, characterized by the following steps:

[0006] The pixelated metal pattern of the pixelated reflection array unit is converted into a binary pixel matrix, where 0 elements represent that the current pixel block has no metal filling and 1 elements represent that the current pixel block has metal filling.

[0007] A training dataset is randomly generated, and the binary pixel matrix of each sample simultaneously satisfies topological smoothness constraints, physical connectivity constraints, and symmetry constraints. Based on the symmetry constraints, a unique matrix block is taken and converted into a one-dimensional vector to form the input feature vector. At the same time, label values ​​are obtained using the full-wave simulation software (HFSS).

[0008] A multi-layer deep neural network is constructed and trained to obtain a surrogate model. The multi-layer deep neural network includes an input layer, a hidden layer, and an output layer. The hidden layer consists of multiple hidden blocks, and each hidden block is composed of an FCL (Full Connect Layer), a BNL (Batch normalization layers), an L_ReLU (Leaky ReLU) layer, and a DropL (Dropoout layer) connected together.

[0009] A genetic algorithm is used, with the initial population randomly generated. A surrogate model is used to quickly screen out candidate reflection units. The candidate reflection units are sent to the full-wave simulation software (HFSS) for accurate verification. For any target pixelated reflection array unit with a preset reflection phase, the candidate unit with the smallest reflection phase error is selected from all verified candidate reflection units and archived in the phase box until full phase coverage is completed.

[0010] The target pixelated reflection array elements with the preset reflection phase are filled into the array positions to obtain the pixelated reflection array surface.

[0011] Furthermore, the topological smoothness constraint includes: no isolated 1-element and the 1-element filling rate is between 20% and 70%; the physical connectivity constraint includes: no 1-element in any 2×2 element subset is distributed only on the diagonal; the symmetry constraint includes: axisymmetric distribution based on the diagonal or rotationally symmetric distribution at 90°.

[0012] Furthermore, when the pixelated reflective array is applied to a folded reflective array antenna, the output of the multilayer deep neural network is , and , For the reflected phase, This refers to the polarization rotation efficiency.

[0013] Furthermore, the selection criteria for the candidate reflective units are: polarization rotation efficiency greater than 0.9 and reflection phase error within ±5°.

[0014] Furthermore, when the pixelated reflective array is applied to a dual-circularly polarized planar reflective array antenna, the output of the multilayer deep neural network is: , , , The reflection phase under normal incidence conditions. The reflection phase under a 30° oblique incidence condition. This represents the phase deviation value.

[0015] Furthermore, the selection criteria for the candidate reflective units are as follows: Less than 10° and The error is within ±5°.

[0016] Meanwhile, the present invention provides a pixelated folding reflective array antenna, comprising: a feed horn, a polarization selective surface, and a pixelated reflective array; characterized in that the pixelated reflective array is designed by the above method; the feed horn is disposed in the central region of the pixelated reflective array, the polarization selective surface is disposed directly above the reflective array, and the distance between the polarization selective surface and the pixelated reflective array is half the focal length.

[0017] Furthermore, the pixelated reflective array is a circular array composed of several pixelated reflective array units. Each pixelated reflective array unit consists of a dielectric substrate, a pixelated metal pattern on the upper surface of the dielectric substrate, and a metal ground plane on the lower surface of the dielectric substrate. The pixelated metal pattern is composed of several pixel blocks arranged in a square array.

[0018] Furthermore, the polarization selection surface adopts a polarization gate structure, which is composed of uniformly arranged metal strips, and the metal strips are overlapped and printed on the upper and lower surfaces of the dielectric substrate.

[0019] The present invention also provides a dual-circularly polarized planar reflective array antenna, comprising: a feed horn and a pixelated reflective array, characterized in that the pixelated reflective array is designed by the above method; the horn feed is a dual-circularly polarized circular horn antenna, which is disposed at the focal point of the pixelated reflective array.

[0020] Furthermore, the pixelated reflective array is a circular array composed of several pixelated reflective array units. Each pixelated reflective array unit consists of a dielectric substrate, a pixelated metal pattern on the upper surface of the dielectric substrate, and a metal ground plane on the lower surface of the dielectric substrate. The pixelated metal pattern is composed of several pixel blocks arranged in a square array.

[0021] Based on the above technical solution, the beneficial effects of the present invention are as follows:

[0022] This invention provides a deep learning-based pixelated reflector array design method. In the pixelated reflector array design process, a deep learning surrogate model is used to replace full-wave simulation, which greatly shortens the optimization time, enables rapid reverse design, and significantly improves the performance of the designed pixelated reflector array antenna. In addition, topological smoothing constraints and physical connection constraints are introduced in the design process to avoid unmanufacturable microstructures, thereby improving the antenna manufacturing yield and robustness.

[0023] Specifically, for the folded reflector antenna, the high-degree-of-freedom topology of the pixelated structure achieves full 360° phase coverage while maintaining a polarization conversion efficiency of over 0.9 for all reflector elements, significantly outperforming traditional geometric scaling methods. Simultaneously, the folded architecture significantly reduces antenna height and eliminates feed smothering. Finally, experimental results show that the antenna achieves a peak gain of 22.38 dBi and an aperture efficiency of 40.51% at 10 GHz. For the planar reflector antenna, incorporating oblique incidence phase deviation into the optimization objective, the designed reflector elements exhibit a phase shift of less than 10° within the 0°–30° incidence angle range, effectively reducing phase errors caused by feed smothering on the reflector surface. The 90° rotationally symmetric structure design supports dual circular polarization operation. Finally, experimental results show that the antenna achieves left-handed and right-handed circular polarization gains of 26.46 dBi and 26.54 dBi respectively at 10 GHz, with an aperture efficiency as high as 45%, significantly outperforming similar designs. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the pixelated reflective array unit of the pixelated folded reflective array antenna in Example 1.

[0025] Figure 2 This is a schematic diagram of the structure of the multilayer deep neural network in Example 1.

[0026] Figure 3 This is a flowchart of the global search based on the genetic algorithm in Example 1.

[0027] Figure 4 The figure shows the simulation results of the electromagnetic characteristics of the pixelated reflective array unit in Example 1 across the entire phase range.

[0028] Figure 5 This is a schematic diagram of the polarization selection surface of the pixelated folded reflective array antenna in Example 1.

[0029] Figure 6 The figure shows the simulation results of the transmission coefficient and reflection coefficient of the polarization selective surface in Example 1.

[0030] Figure 7 This is a schematic diagram of the overall structure of the pixelated folded reflective array antenna in Example 1.

[0031] Figure 8 This is the radiation pattern of the pixelated folded reflective array antenna in Example 1.

[0032] Figure 9 The image shows the amplitude-frequency response curve of the pixelated folded reflective array antenna in Example 1.

[0033] Figure 10 This is a schematic diagram of the pixelated reflective array unit of the dual circularly polarized pixelated reflective array antenna in Example 2.

[0034] Figure 11 This is a schematic diagram of the structure of the multilayer deep neural network in Example 2.

[0035] Figure 12 The image shows the simulation results of the electromagnetic response of the pixelated reflective array unit in Example 2 across the entire phase range.

[0036] Figure 13 This is a schematic diagram of the feed horn structure of the dual circularly polarized pixelated reflective array antenna in Example 2.

[0037] Figure 14 This is the radiation pattern of the feed horn in Example 2.

[0038] Figure 15 This is a schematic diagram of the overall structure of the dual-circularly polarized pixelated reflective array antenna in Example 2.

[0039] Figure 16 The image shows the LHCP radiation pattern of the dual circularly polarized pixelated reflective array antenna in the yoz plane in Example 2.

[0040] Figure 17 The image shows the LHCP radiation pattern of the dual circularly polarized pixelated reflective array antenna in the xoz plane in Example 2.

[0041] Figure 18 The image shows the RHCP radiation pattern of the xoz plane of the dual circularly polarized pixelated reflective array antenna in Example 2.

[0042] Figure 19 The image shows the RHCP radiation pattern of the yoz plane of the dual circularly polarized pixelated reflective array antenna in Example 2.

[0043] Figure 20 The graph shows the result of the maximum radiation gain of the dual circularly polarized pixelated reflective array antenna in Example 2 changing with frequency.

[0044] Figure 21 The graph shows the result of the axial ratio of the dual circularly polarized pixelated reflective array antenna in Example 2 changing with frequency. Detailed Implementation

[0045] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0046] Example 1

[0047] This embodiment provides a pixelated folded reflective array antenna, employing a folded optical path, including: a feed horn, a polarization selective surface, and a pixelated reflective array. The feed horn is positioned in the central region of the pixelated reflective array, and the polarization selective surface is positioned directly above the reflective array, with the distance between the polarization selective surface and the pixelated reflective array being half the focal length. The polarization selective surface transmits y-polarized electromagnetic waves and reflects x-polarized electromagnetic waves. Therefore, the x-polarized spherical wave emitted by the feed horn is reflected back to the pixelated reflective array by the polarization selective surface. The pixelated reflective array performs polarization torsion (torsion into y-polarized electromagnetic waves) and phase adjustment on the incident x-polarized electromagnetic wave before re-emitting it, which then passes through the polarization selective surface and enters space.

[0048] The polarization selection surface adopts a polarization gate structure, which is composed of uniformly arranged thin metal strips; the metal strips have a width of 0.4 mm and a period of 0.3 mm, and are overlapped and printed on the upper and lower surfaces of the substrate.

[0049] The pixelated reflective array is circular and consists of several pixelated reflective array units. The reflection phase of each pixelated reflective array unit is calculated based on the array beamforming principle. In each pixelated reflective array unit, the dielectric substrate material is RF-35 with a thickness of 1.524 mm and a dielectric constant of 3.5. The physical size of each unit is 10 × 10 mm. 2 (Approximately 1 / 3 wavelength) A pixelated metal pattern is formed on the upper surface of the dielectric substrate, consisting of 9×9 pixel blocks, with each pixel measuring 1×1 mm. 2 A metal floor is provided on the lower surface of the dielectric substrate.

[0050] Furthermore, in order to achieve the design of the pixelated reflective array, this embodiment needs to obtain 36 pixelated reflective array elements with phase states ranging from -180° to 170° and spaced at 10° intervals, which is designed by the following steps:

[0051] Step 1: Convert the pixelated metallic pattern of the pixelated reflective array unit into a binary pixel matrix, such as... Figure 1 As shown, 0 elements represent that the current pixel block has no metal fill, and 1 elements represent that the current pixel block has metal fill; and the following physical constraints are also satisfied:

[0052] First constraint: No isolated 1 element, that is, there is no single 1 element that is surrounded by 0 elements;

[0053] Second constraint: There is no 2×2 element subset in which the element 1 is distributed only on the diagonal;

[0054] Adjacent elements are connected only by a shared edge;

[0055] Third constraint: The fill rate of element 1 is between 20% and 70%;

[0056] Fourth constraint: The binarized pixel matrix is ​​axially symmetric about any diagonal.

[0057] A training dataset of 100,000 samples was randomly generated. Each sample was converted into a binary pixel matrix, and then, based on the symmetry constraint (fourth constraint), the upper triangular matrix was converted into a one-dimensional vector to form the input feature vector (45-dimensional). Simultaneously, the reflection phase was obtained using full-wave simulation software (HFSS). With polarization rotation efficiency ,Will , and As a label;

[0058] Step 2: Construct a multi-layer deep neural network and complete its training;

[0059] The multi-layer deep neural network, such as Figure 2 As shown, it includes: an input layer, a hidden layer, and an output layer;

[0060] The input feature vector is input through the input layer;

[0061] The hidden layer consists of four hidden blocks with the number of neurons being 512, 512, 256, and 128 respectively. Each hidden block is composed of an FCL (Full Connect Layer), a BNL (Batch normalization layers), an L_ReLU (LeakyReLU) layer, and a DropL (Dropoout layer).

[0062] The output layer predicts the polarization rotation efficiency and the sin / cos values ​​of the reflection phase;

[0063] Based on the training dataset from step 1, the root mean square error of MSE is set as the loss function to complete the model training and obtain the surrogate model.

[0064] Step 3: Use a genetic algorithm, such as... Figure 3 As shown, the initial population is generated from random pixels. A trained surrogate model is used to quickly select candidate units with a polarization rotation efficiency greater than 0.9 and a reflection phase error within ±5°. These candidate units are then sent to full-wave simulation software (HFSS) for precise verification. For any target phase unit, the candidate unit with the smallest reflection phase error is selected from all verified candidate units and archived into the phase box. This process continues until 36 target pixelated reflector array units are collected, as shown below. Figure 4 As shown, the final optimized 36 pixelated reflective array units maintained high polarization conversion efficiency across the entire phase range, with a phase error within 5°.

[0065] The beneficial effects of the present invention will be explained in detail below with reference to simulation tests.

[0066] like Figure 5 As shown, the polarization rotating surface employs a double-layer metal grating structure, such as... Figure 6 As shown, in the 10 GHz band, when the incident angle is scanned within the range of 0° to 35°, the transmission coefficient |T| of the polarized rotating surface to the y-polarized wave is... yy | Remaining above -1dB indicates transmission; the reflection coefficient of the polarized rotating surface to x-polarized waves is|R xx | It remains within -0.1dB, in a reflective state, to meet the requirements of folded optical paths.

[0067] like Figure 7 As shown, in this embodiment, the pixelated folded reflective array antenna consists of a feed horn, a polarization-selective surface, and a pixelated reflective array. The feed horn uses a pyramidal horn with a -10dB beamwidth of 70°. The pixelated reflective array places the designed target pixelated reflective array elements at the corresponding array positions according to the required pencil beam phase distribution. The pixelated folded reflective array antenna is tested in a microwave anechoic chamber, and the results are as follows... Figure 8 and Figure 9 As shown, the antenna has a peak gain of 22.38 dBi at 10 GHz, corresponding to an aperture efficiency of 40.51%; the sidelobe level is below -10 dB, the cross-polarization level is below -15 dB, and the -3 dB gain bandwidth is approximately 8.85% (9.50–10.38 GHz); the measured results agree well with the simulation, thus demonstrating the effectiveness and superiority of the present invention.

[0068] Example 2

[0069] This embodiment provides a dual-circularly polarized pixelated reflective array antenna, including: a feed horn and a pixelated reflective array. The horn feed is a dual-circularly polarized circular horn antenna, which is located at the focal point of the pixelated reflective array and has a focal diameter ratio of 0.8.

[0070] The pixelated reflective array is circular and consists of several pixelated reflective array units. The reflection phase of each pixelated reflective array unit is calculated based on the array beamforming principle. In each pixelated reflective array unit, the dielectric substrate material is RF-35 with a thickness of 1.524 mm and a dielectric constant of 3.5. The physical size of each unit is 10 × 10 mm. 2 (Approximately 1 / 3 wavelength) A pixelated metal pattern is formed on the upper surface of the dielectric substrate, consisting of 9×9 pixel blocks, with each pixel measuring 1×1 mm. 2 A metal floor is provided on the lower surface of the dielectric substrate.

[0071] Furthermore, in order to achieve the design of the pixelated reflective array, this embodiment also needs to obtain 36 pixelated reflective array elements with phase states ranging from -180° to 170° and spaced at 10° intervals, which is designed by the following steps:

[0072] Step 1: Convert the pixelated metallic pattern of the pixelated reflective array unit into a binary pixel matrix, such as... Figure 10 As shown, 0 elements represent that the current pixel block has no metal fill, and 1 elements represent that the current pixel block has metal fill; and the following physical constraints are also satisfied:

[0073] First constraint: No isolated 1 element, that is, there is no single 1 element that is surrounded by 0 elements;

[0074] Second constraint: There is no 2×2 element subset in which the element 1 is distributed only on the diagonal;

[0075] Adjacent elements are connected only by a shared edge;

[0076] Third constraint: The fill rate of element 1 is between 20% and 70%;

[0077] Fourth constraint: The binarized pixel matrix is ​​90° rotationally symmetric.

[0078] A training dataset of 100,000 samples was randomly generated. Each sample was converted into a binary pixel matrix, and the matrix blocks were converted into one-dimensional vectors according to the symmetry constraint (fourth constraint) to form the input feature vector (21-dimensional). At the same time, the reflection phase under normal incidence conditions was obtained using full-wave simulation software (HFSS). Reflection phase under 30° oblique incidence condition The phase deviation value , , As a label;

[0079] Step 2: Construct a multi-layer deep neural network and complete its training;

[0080] The multi-layer deep neural network, such as Figure 11 As shown, it includes: an input layer, a hidden layer, and an output layer;

[0081] The input feature vector is input through the input layer;

[0082] The hidden layer consists of five hidden blocks with the number of neurons being 256, 512, 512, 256, and 128 respectively. Each hidden block is composed of an FCL (Full Connect Layer), a BNL (Batch normalization layers), an L_ReLU (Leaky ReLU) layer, and a DropL (Dropoout layer).

[0083] Output layer prediction phase bias value , , ;

[0084] Based on the training dataset from step 1, the root mean square error of MSE is set as the loss function to complete the model training and obtain the surrogate model.

[0085] Step 3: Using a genetic algorithm, the initial population is generated from random pixels, and the trained surrogate model is used to quickly select those that meet the criteria of step 1). Less than 10°; 2) Candidate elements are selected for -180±5°, -170±5°, ..., 170±5°. These candidate elements are then fed into the full-wave simulation software (HFSS) for precise verification. For any target phase element, the candidate element with the smallest reflection phase error is selected from all verified candidate elements and archived into the phase box. This process continues until 36 target pixelated reflective array elements are collected. Figure 12 As shown, the reflection amplitude of the generated unit is above 0.97, and the phase deviation value is... The fact that the angle is no more than 10° and the error of the reflection phase is within ±5° indicates that these reflection array elements have excellent working performance.

[0086] The beneficial effects of the present invention will be explained in detail below with reference to simulation tests.

[0087] like Figure 13 As shown, in this embodiment, the feed source adopts a dual-circular polarized feed horn, and its radiation pattern is as follows. Figure 14 As shown in the figure, the main polarization radiation patterns of the horn antenna coincide on the xoz and yoz planes, indicating that it is suitable for use as a feed excitation for a reflector array.

[0088] like Figure 15 As shown, in this embodiment, the dual-circularly polarized pixelated reflective array antenna consists of a feed horn and a pixelated reflective array. The pixelated reflective array places the designed target pixelated reflective array elements at the corresponding array positions according to the required pencil beam phase distribution. The dual-circularly polarized pixelated reflective array antenna is tested in a microwave anechoic chamber, and the results are as follows... Figures 16-21As shown in the figure, the peak gains of left-hand circular polarization and right-hand circular polarization are 26.46 dBi and 26.54 dBi, respectively, with corresponding aperture efficiencies of 44.27% and 45.09%. Within a bandwidth of 10.1%, the axial ratio is less than 2 dB, and the gain drop is less than 2 dB.

[0089] The above description is merely a specific embodiment of the present invention. Any feature disclosed in this specification may be replaced by other equivalent or similar features unless otherwise specified. All disclosed features, or steps in all methods or processes, may be combined in any way except for mutually exclusive features and / or steps.

Claims

1. A pixelated reflective array design method based on deep learning, characterized in that, Includes the following steps: The pixelated metal pattern of the pixelated reflection array unit is converted into a binary pixel matrix, where 0 elements represent that the current pixel block has no metal filling and 1 elements represent that the current pixel block has metal filling. A training dataset is randomly generated, and the binary pixel matrix of each sample simultaneously satisfies topological smoothness constraints, physical connectivity constraints, and symmetry constraints. Based on the symmetry constraints, a unique matrix block is taken and converted into a one-dimensional vector to form the input feature vector. At the same time, label values ​​are obtained using the full-wave simulation software (HFSS). A multi-layer deep neural network is constructed and trained to obtain a surrogate model. The multi-layer deep neural network includes an input layer, a hidden layer, and an output layer. The hidden layer consists of multiple hidden blocks, and each hidden block is composed of an FCL (Full Connect Layer), a BNL (Batch normalization layers), an L_ReLU (Leaky ReLU) layer, and a DropL (Dropooutlayer) layer. A genetic algorithm is used, with the initial population randomly generated. A surrogate model is used to quickly screen out candidate reflection units. The candidate reflection units are then sent to the full-wave simulation software (HFSS) for accurate verification. For any target pixelated reflection array unit with a preset reflection phase, the candidate unit with the smallest reflection phase error is selected from all verified candidate reflection units and archived in the phase box until full phase coverage is achieved. The target pixelated reflection array elements with the preset reflection phase are filled into the array positions to obtain the pixelated reflection array surface.

2. The pixelated reflector array design method based on deep learning according to claim 1, characterized in that, The topological smoothness constraint includes: no isolated 1-element and a 1-element filling rate between 20% and 70%; the physical connectivity constraint includes: no 1-element subset exists where 1-element is distributed only on the diagonal; the symmetry constraint includes: axisymmetric distribution based on the diagonal or rotationally symmetric distribution at 90°.

3. The pixelated reflector array design method based on deep learning according to claim 1, characterized in that, When pixelated reflective surfaces are applied to a folded reflective array antenna, the output of the multilayer deep neural network is: , and , For the reflection phase, This refers to the polarization rotation efficiency.

4. The pixelated reflector array design method based on deep learning according to claim 3, characterized in that, The selection criteria for the candidate reflective units are: polarization rotation efficiency greater than 0.9 and reflection phase error within ±5°.

5. The pixelated reflector array design method based on deep learning according to claim 1, characterized in that, When the pixelated reflective array is applied to a dual-circularly polarized planar reflective array antenna, the output of the multilayer deep neural network is: , , , The reflection phase under normal incidence conditions. The reflection phase under a 30° oblique incidence condition. This represents the phase deviation value.

6. The pixelated reflector array design method based on deep learning according to claim 5, characterized in that, The selection criteria for the candidate reflection units are as follows: Less than 10° and The error is within ±5°.

7. A pixelated folded reflective array antenna, comprising: The invention comprises a feed horn, a polarization selective surface, and a pixelated reflective array; characterized in that the pixelated reflective array is formed by the pixelated reflective array design method of claim 1; the feed horn is disposed in the central region of the pixelated reflective array, the polarization selective surface is disposed directly above the reflective array, and the distance between the polarization selective surface and the pixelated reflective array is half the focal length.

8. The pixelated folded reflective array antenna according to claim 1, characterized in that, The pixelated reflective array is a circular array composed of several pixelated reflective array units. Each pixelated reflective array unit consists of a dielectric substrate, a pixelated metal pattern on the upper surface of the dielectric substrate, and a metal ground plane on the lower surface of the dielectric substrate. The pixelated metal pattern is composed of several pixel blocks arranged in a square array.

9. A dual-circularly polarized planar reflector antenna, comprising: The feed horn and pixelated reflector array are characterized in that the pixelated reflector array is formed by the pixelated reflector array design method of claim 1; the horn feed is a dual-circularly polarized circular horn antenna, which is located at the focal point of the pixelated reflector array.

10. The dual-circularly polarized planar reflective array antenna according to claim 9, characterized in that, The pixelated reflective array is a circular array composed of several pixelated reflective array units. Each pixelated reflective array unit consists of a dielectric substrate, a pixelated metal pattern on the upper surface of the dielectric substrate, and a metal ground plane on the lower surface of the dielectric substrate. The pixelated metal pattern is composed of several pixel blocks arranged in a square array.

Images