Basic Active Array for MEOLUT application Saša Dragaš1, Zoran Golubičić(1), Cristina Lavin (1) , Victorina Fernandez (1) , Zoran Cvetković (1) , Yolanda Fernandez (1) , Miloje Zečević (1) , Jagan Shanbhag (1), Josep Perlas Reyes (2) , Enric Obiols Bernaus (2) , Luca Salghetti Drioli (3) (1) TTI, Parque Científico y Tecnológico de Cantabria C/Albert Einstein no 14, 39011 Santander, Spain, Email:
[email protected] (2) (3)
INDRA ESPACIO, C/ Roc de Boronat, 133, 08018, Barcelona, SPAIN, Email:
[email protected]
European Space Agency - ESTEC, TEC-EEA, Keplerlaan 1, NL-2201 AZ Noordwijk, The Netherlands, Email:
[email protected]
ABSTRACT This paper describes activities performed during the first design and development phase within the frame of ESA GSTP project “Basic Active Array Module for MEOLUT Application”. The main goal of the project is to present a fully integrated and tested basic module of the future antenna array system for Search and Rescue (SAR) application with MEO satellites. Required multi satellite tracking operation and almost hemispherical coverage imply that the most suitable and the simplest antenna sub-system solution can be found in employment of a spherical array configuration implementation jointly with digital beam forming techniques.
2. SPHERICAL ANTENNA ARRAY DESIGN A large spherical antenna array (in terms of number of employed radiating elements) theoretically conserves all radiation properties independently of pointing direction. In reality MEOLUT antenna sub-system is approximation made of planar sub-arrays and some variation can be expected. In order to minimize this possible variation a strategy of forming the beam by employing about 100 elements laying over the surface of dodecahedron has been taken. These elements are determined by intersecting the dodecahedron with imaginary cone. This concept is presented on the next figure.
1. INTRODUCTION Among other approaches, digital beamforming antenna array for MEOLUT (MEO Local User Terminal) application is a fully justified solution from economical point of view, as a replacement of multiple motorised parabolic antennas field as well from operational reasons, offering competitive radiation performances and flexible architecture for possible future mission expansions. Real implementation of spherical array is based on well known geodesic dome approach. The array is shaped as a dodecahedron consisting of 55 identical triangles, that permits application of any known planar antenna and RF technology. The analysis showed that target gain of 25dB was possible to achieve employing ten, more directive radiating elements, per triangle. Also, good theoretical suppression of side lobes and high purity of circular polarisation was achieved. The beamformer board architecture is based on use of SPARTAN 3ADSP FPGA. The implemented open architecture permits functionality of the future system as a pure phased array or implementation of adaptive algorithms. The processing operations of digital multiplication, filtration and signal delay adjustment (phase shifting ) are incorporated into FPGA and fully tested.
Figure 1. Different active apertures for the synthesis of the main beam for the angles 0, 45 and 85 deg respectively
From the previous figures it can be denoted that for the each pointing angle the estimated aperture contain almost the same number of the radiating elements, placed inside the intersection of the sphere and the imaginary cone that starts from the centre of the dodecahedron (e.g. sphere). The distribution of the radiating element within aperture differs a bit from one pointing angle to another, but their density over the spherical surface is very similar. It means that radiation performances will be very stable in the whole coverage area. However, almost the same radiation aperture is conserved towards the new pointing direction. The distribution of the radiating elements over the new active aperture is very similar but with different sequential rotation. Due to the fact that this different sequential rotation has to be compensated by adequate phase shift component, the new rotated radiation aperture will preserve mostly the radiation properties. It must be pointed out that for a selected active aperture the 2-D electronic main beam scan should be produced within a narrow cone of 10degrees. For the further scan a new “shifted” aperture should be selected. In order to perform simulation of a larger antenna configuration and to speed up its execution it was a reasonably to simplify radiating element, conserving the important parameters as much as possible close to the ones obtained using full 3-D models. In order to make as much as possible, including large number of radiating elements, realistic prediction of the overall antenna array performances it was necessary even more to simplify the radiating element model. Therefore, it has been tested hexagonal approximation of the open circular waveguide with square patch exciter. The layout of the hexagonal waveguide approximation element with square patch fed with two delta generators is given on the following figure.
Polarisation Gain of isolated element Port Isolation Cross-polar discrimination Radiation pattern shape
Elevation gain pattern roll-of
Cross-polar spatial limit
discrimination
Wide angle matching
Dual Circular (LH and RH) 8dB >20dB >18dB High level of co-polar and cross-polar pattern axial symmetry. Target radiation element pattern should be similar to (cosi( ))^2.5 4-5dB within a cone of 45degree with respect to the vertical Better than 13dB within a cone of 45degree with respect to the vertical (e.g. axial ratio better than 4dB) Radiating element should provide optimum noise figure match up to the angle of 45degree of incident wave.
Table 1. Set of established requirements for the single radiating elements On the next figure is depicted a part of the envisaged antenna aperture with 70 hexagonal radiating elements.
Figure 3. Original radiating element model and its simplified model with conserved radiation properties The computed 3-D co-polar and principal cuts patterns are presented in continuation.
Figure 2. Original radiating element model and its simplified model with conserved radiation properties Selection of the employed radiating element has been done on base on the following assumption listed in the table below. It has been done on the base on requirements for two-dimensional scan operation where each element contributes up to certain scan angle, while after the new 100 elements are selected in forming of main beam. Figure 4. Computed 3/D radiation pattern
same radiating elements. In order to evaluate this possible unbalance on the element ports on the overall noise performances the design of LNA block was carried out using as loads predicted impedance variation of the radiating element. 2.2. Radiation element implementation The following figure presents the final concept of the radiating element that includes all parts necessary for its assembly.
Figure 5. Computed co-polar and cross-polar pattern principal cuts 2.1. Mutual coupling study Mutual coupling is treated between the elements that form a basic triangle as well as between adjacent triangles. This study has been carried out on the radiating elements showed in previous section.
Figure 8. Overview of the parts for radiating element assembly
Figure 6. Model of mashed triangle used in coupling study
Figure 7. Model of mashed triangle with a part of adjacent triangle used in coupling study Regarding the computed results it can be denoted that matching is kept while beam steering is performed on the most of the radiating elements within the triangular arrangement. Also, it was seen that matching can be rather different between the ports that belong to the
The dual circular polarisation operation is obtained using a 90 degree hybrid coupler. In continuation are given the simulated results for s matrix, at the ports before the coupler, as well as radiation properties in the principal cuts, showing expected symmetry and high purity of the circular polarisation.
Figure 9. Computed Matching on the ports before 90degree coupler and isolation between them
Figure 10. Computed radiation patterns in the principal cuts
Figure 13. Measured co-polar and cross-polar radiation pattern at the RH port in both principal cuts
Figure 11. Radiating element ready for integration into triangular array The next set of diagrams shows the measured results including the influence of radom cover. In general, can be concluded that measured results are in accordance with the required properties from Tab.1.
Figure 14. Measured matching at the both ports and isolation 3. BASIC MODULE ARCHITECTURE Antenna architecture is based on 55 basic modules. Modules contain 10 dual polarized radiating elements. MEOLUT signals received from all elements are down converted and transmitted to common signal processor. Common signal processor and common controller serve to emulate part of CBP (Core Beamforming Processor) function up to the moment when CBP will be design. Signals from all satellites, collected and processed by common processor are transmitted to CBP. General block schematic of the antenna hardware is presented at the next figure.
Figure 12. Measured co-polar and cross-polar radiation pattern at the LH port in both principal cuts
multiplexes are transmitted over 12 LVDS lines. Beam former controller communicates with CBFP and common processor by Ethernet connection.
Figure 17. Basic module blocks distribution Figure 15. Basic module block schematic Core beam forming processor transmits commands to the common signal processor and receives the control data. Core beam former processor transmits reference and control signal to frequency synthesizer. Frequency synthesizer feed all basic modules with microwave LO and VHF clocks. Basic module is organized in two levels. RF blocks are positioned at the bottom level. RF blocks are directly (without connectors) connected to radiating elements. RF blocks are connected to LO and clock signals by SMA connectors and cables. All other connections are realized by mezzanine high speed digital connectors (including power supply). Part of the second level presents level of power dividers intended for LO and clock signal distribution. Power dividers are equipped by SMA connectors. The main part of the second level presents Beam former Processing unit with beam former controller. 20 digital signals (16 bit each) are fed by mezzanine high speed connectors.
4. RF FRONT-END In order to minimize receiver noise factor first stages in both RF block channels are low noise amplifiers. LNA gain of 20 dB and noise figure of 0.45dB are assumed. Two LNA outputs are combined in quadrature hybrid combiner. Outputs of the combiner present LHCP and RHCP signals. Losses introduced by coupler are below 1dB. The second LNA is applied after hybrid in order to amplify signal before RF filters. Additional gain of 20dB is provided. The first IF filter is 20MHz wide and flat inside the 2 MHz central frequency and it is positioned at the second LNA output. Insertion loss of this filter is below 5dB. Last stage at RF frequency is third amplifier stage with additional 20dB gain. RF chain gain is 54dB. In reality amplifiers have more than 20dB gain but some losses of incorporated attenuators are encountered. Attenuators are incorporated for better isolation and better matching. Active mixer is applied with 7 dB gain. First IF filter with 2 dB losses serves to reject inter modulation products. Overall gain after this filter is 59dB. First IF amplifier with 26 dB gain increases overall gain to 85 dB. Second IF filter is SAW filter. Last two amplifiers with 54 dB gains are increasing overall gain to 130dB.
Figure 18. RF front-end block diagram
Figure 16. Basic module beamformer processor unit The same connectors are applied for control signal connections (between Beam former Controller and RF Blocks). LVDS connection will be provided for partially processed multiplex signal and 12 non processed signal
70MHz output signal is filtered before input to AD converter. Dual channel 14 bit AD converter enables 84 dB dynamic range. Two 14 bits signal and some auxiliary signals present RF block outputs. AD converter parameters are controlled by RF controller. RF controller controls AD converter, operation and can adjust the gain of the analogue chain in order to avoid saturation. RF control measures the voltages and current
of different analogue stages in order to protect antenna from serious failure.
Figure 21. A/D module block schematics Figure 19. Integrated RF front-end In the following figure, noise figure and gain results for RF module are shown. The gain is 63.4dB and noise figure is 0.66dB for the frequency band.
Five AD9600, with appropriate input circuits, construct AD conversion block. AD9600 is dual 10 bits, 105 MSPS converter produced by Analog Devices. This configuration permits ten independents analogue signals conversion at the same time. The final version of the Beam former digital processing board will have ten dual, 14 bits AD converts AD9640. 60 MHz 10 bits-RHCP(0:9)
Digital Down Converter
RAM
100MbpsEthernet MAC control
Beamformer
Figure 22. DBF block architecture Figure 20. Measured gain and NF of the RF front-end block 5. DBF REALIZATION, MEASUREMENTS AND VERIFICATION The next figure presents block diagram of A/D module that was developed and produced in order to have a preliminary results that will help to close the final design of the final Digital Beam Former module. The board consists of the following blocks: AD conversion block, Referent clock driver block, Line buffers, Power supply block.
DBF module captures the signal samples from A/D module and performs necessary signal processing.. Module receives the 10 bits RHCP samples from 10 antennas A/D module with 60 MHz clock. The digital down converter (DDC) allows obtaining the I-Q signals for one type of satellite. The beamformer contains the multipliers and adders to execute the operations in signed numbers representation required by the algorithm, multiplexers to select the correct operands and a synchronism control to perform the sequence of the arithmetic operations. In addition, there is a block for the 100 Mbps Ethernet MAC control whose task is sending the results buffered in FIFO to the computer for the representation. For the numeric multiplication, the direct digital synthesizer (DDS) generates two’s complement sinusoid values for the frequency of fH=10.4 MHz. The cascaded integrator-comb (CIC) filter generates a lowpass response output and implements the down
sampling. Three important parameters are defined: the number of CIC stage (N), the rate change for decimation (R) and the differential delay in the comb section stages (M). In addition, multi channel FIR filters are added in the last stage: a compensation finite impulse response (CFIR) filter and a programmable finite impulse response (PFIR) filter. These FIR filters also include a sample rate conversion (R). The DDC block introduces therefore a total rate change Rt = 50· 3·2 = 300, which means that, with a 60 MHz frequency clock, samples on DDC outputs are obtained each 5μs. With Xilinx CORE generator software is possible to configure and generate the cores. The algorithm is realised with VHDL on a Xilinx Virtex-5 XC4VFX100-2FF1136 FPGA and the design and implementation stages are carried out entirely with the Xilinx Integrated Software Environment. In order to fully test the A/D and DBF modules functionality, simulator test system was designed The main purpose of the system is to provide 10 signals with the possibility to control the phase shift between each signal. For this purpose, simulator board was developed together with the corresponding control software for Windows platform. DDS1
DDS2
DDS3
DDS4
DDS5
DDS6
DDS7
DDS8
Tx antenna 1544.91 MHz
A/D signal generator DBF
Figure 24. Integrated system test block schematic
DDS9 DDS10
μC
RS-485
clock
to PC
Figure 25. Integrated system during laboratory tests
Figure 23. Simulator board block schematic Control software for the Windows platform is written in order to have full control over individual phases of each of the 10 output signals. The software allows setting the output frequency for all DDS chips, and separate setting and control of each phase. Simulator test system was connected to the A/D module with the DBF module. The complete system was tested by changing the output phases and the functions of all modules were validated. After each of the modules was tested and verified, the next step was to perform fully integration of RF front-end subsystem, A/D and DBF modules. The following figures present the block diagram of system integration and testing in laboratory environment.
Figure 26. A/D converter input (upper figure) and
DBF output
6. SYSTEM LEVEL ANALYSIS AND TEST PROCEDURE DEFINITION The presented design and development work was preceded with an extensive research of COSPASSARSAT regulations and normative in order to establish a coherent set of requirements for each aspect of design which can be listed as follow: Functional requirements (Antenna array functions, reception of SAR downlink signals, satellite tracking, …). Performance requirements (G/T, polarization …). External interface requirements. Environmental requirements (temperature range, humidity range, …) Implementation requirements (time and frequency references, calibration, …) Operational requirements (autonomy, …) Verification requirements (SAR streams compatibility with MEOLUT, ..) Product assurance requirements (quality standards). Special requirements for the Core Beamforming Processor. In order to verify requirements and functional readiness of the basic module set of test procedures has been defined for the next stage of the project. Each test case consists of a first definition section where are mentioned the related requirements, verification objectives, test description overview, configuration, test conditions, tools needed and test organization, such as, estimated duration and scheduling constraints. 7. ADAPTIVE ALGORITHM IMPLEMENTATION ROADMAP Beam former architecture has been defined in such way that permits future implementation of an adaptive algorithm. Otherwise, the future spherical system is able to work as a pure phased array antenna providing radiation properties quite competitive with reflector counterparts. From that point of view it has been made an effort to find a suitable manner in radiation properties prediction using 3-D electromagnetic simulators while antenna element excitation are calculated using some adaptive algorithm. The main intention was to provide as much as possible closer picture using as much as possible realistic antenna array configuration, including the coupling effect. A short analysis presented in continuation is a result of initial investigations. The Fig. 27 shows the flowchart on implementing the system. In the first iteration, LMS algorithm is implemented on a theoretical Array (array with identical element patterns). The computed weights were then entered into practical Array designed in any simulation Software. If the desired Radiation pattern
doesn’t match with the theoretical pattern, then the radiation Pattern in Exported and LMS Algorithm is implemented till we have the desired output as the theoretical output.
Figure 27. Implementing algorithm flowchart To prove this, a LMS algorithm is implemented on a simplified and smaller current MEOLUT configuration as show on figure 28. The proposed method was applied on a part of simplified MEOLUT array configuration having one cross-dipole, as radiating element, per triangle.
Figure 28. Simplified conformal MEOLUT array The Signal of Interest is at (θ, φ) (100,600) and the interference is at (100,600) . Before the Implementation, The Difference between Signal and Interference is 12 dB and after LMS implementation, the difference is 38 dB.
Figure 29. Pattern before (on the left) and after implementing LMS algorithm 8. REFERENCES 1. TTI, INDRA ESPACIO: Test Readiness Report, Basic Active Array Module for MEOLUT Application: 21976/08/09-TTI, June 22nd 2010. 2. Kolundzija, B at al: WIPL-D Pro CAD 2009, Wipl-d d.o.o.
