Active phased array antennas do not only have the advantage of fast and flexible control of the patterns but also the ability to be mathematically modelled. Such an antenna model provides a software tool to accurately determine the antenna beam patterns based on detailed characterisation of the antenna hardware and knowledge of the antenna control parameters. To achieve the required radiometric quality, this concept requires highly accurate pre-launch characterisation data. After pre-launch validation against near field range pattern measurements and in-flight verification the antenna model will be used to generate the in-orbit calibrated beam patterns required for radiometric corrections in the SAR processing throughout the satellite lifetime. The model has to be capable of accurately determining not only the individual beam patterns but also the relative gain variations from beam to beam. Determination of the absolute gain from measurements over external calibration targets can then be reduced to few beams.
As part of the external calibration activities during commissioning, the antenna model has to be verified using rain forest data (Figure 1) and azimuth patterns recorded by ground receivers. A further important task is the determination of beam pointing offsets using monopulse (notch) patterns over rain forest and receivers. Beam to beam relative calibration can be verified in ScanSAR mode over homogeneous targets.
In case of TerraSAR-X, the antenna model verification as well as the absolute calibration will be performed for three (low, medium, and high incidence angle) beams. Compared to traditional in-flight calibration, that would require deployment, maintenance and data collection of transponders, and the need of long-lasting antenna beam characterisation using repetitive passes over the rainforest for at least all performance beams, the required effort and the duration of the commissioning phase can be significantly reduced.
This antenna model is the key element of the antenna pattern section shown in Figure 2, which has been implemented into the TerraSAR-X IOCS (Instrument Operations and Calibration Segment) (Web page ‘TerraSAR-X Cal). Beyond the above capabilities the antenna pattern module also features a tool for generating optimised beam coefficients under given constraints (e.g. side lobe suppression, main lobe width and gain, gain ripple). This so-called Antenna Excitation Generator is used to calculate the launch set of TerraSAR-X beam coefficients and will be used for beam optimisation in the event of major degradations of the antenna during the mission.
One example of calculated patterns is shown in Figure 3. The mask is derived by different performance parameters like the minimum gain in the main lobe for sufficient power within the coverage region (the blue line) and the maximum gain in the side lobes for sufficient suppression in the ambiguity regions (the green line). The black curve is the pattern for the best performance, calculated for an ideal antenna array before launch. Then, during the lifetime of the SAR instrument some transmit/receiver modules (TRM) may drift or fail and consequently the antenna pattern violates the mask, as indicated by the brown line. An optimisation of the remaining TRM yields the red curve which doesn’t violate the ambiguity mask, while the achieved gain in the coverage region reaches the minimum required.
This Antenna Pattern Optimisation Software was first developed for the ASAR front-end under ESA/ESTEC contract. The ASAR Antenna consists of 320 sub-arrays and TRMs, respectively. For TerraSAR-X, two optimisation approaches were realised and compared, an analytical and a randomising method. The first approach uses a genetic algorithm on top of a deterministic inversion. The solution is achieved by calculating several “generations” with certain “individuals” of patterns and recombining the associated excitation coefficients through mutation. The second method calculates thousands of different pattern realisations by randomly varying the excitation coefficients to find the optimum solution (500 realisations per second on a Pentium IV, 3.4 GHz). In case of TerraSAR-X the second method showed better performance and results.