Frazer BEAMFORMER TEMPORAL STABILITY Figure 10. Time evolving mode rejectability for the 1F2l mode for the case of a new rejection beamformer for each coherent processing interval (solid line) and the case of a fixed mode rejection beamformer where the beamformer solution is determined using only the first coherent processing interval (32 s) of data. The temporal stability of the mode rejection MVDR beamformer solution has been investigated. Figure 10 contrasts the mode rejectability of the 1F2l mode for two cases recorded over the Mt. Everard path. In the first (solid line) the 1F2l mode is rejected continually through the 110 s period shown. Mode rejectability of greater than 55 dB is achieved for all coherent processing intervals (32 s coherent interval and 10 s processing stride). By contrast the dashed line shows rejectability for the case where the rejection beamformer solution is determined during the first coherent interval (32 s) then held fixed and applied to every subsequent coherent processing interval (32 s coherent interval and 10 s processing stride). Rejectability drops quickly and is less than 40 dB after 30 s and less than 30 dB after 110 s. This result is typical and demonstrates that the ionosphere is sufficiently dynamic that the mode rejection beamformer becomes "stale" and rejection performance deteriorates after short intervals. ELEVATION ANGLE FROM DIRECTION-OF-DEPARTURE Consider an example where there are five modes propagated over the Coondambo to Kings Canyon path. The single waveform range-Doppler map is shown in Figure 9a. We have selected the 1E (first mode out in range) to preserve and our goal is to reject all other propagation modes. This was achieved at Kings Canyon using the MVDR mode-selective beamformer solution as shown in Figure 9b. All unwanted modes have been rejected to the level of the noise floor. The same beamformer solution applied to data received at Hermannsburg (range 844.4 km azimuth -21.3°) is ineffective and five propagation modes can clearly be seen in the range-Doppler map of Figure 9c. This poor rejection performance is caused by both a slightly different ionospheric structure on the Coondambo to Hermannsburg path compared with the Coondambo to Kings Canyon path and also the sidelobe properties of the transmit array. These factors are important in the radar backscatter case for spatially distributed clutter from the earth surface. The noise levels were higher at Kings Canyon due to local effects nearby the receiver. Mode elevation take-off angle estimates determined using the MISO array exploit direction-of-departure estimates measured using the array and the known bearing from the transmitter array to the receiver locations. Elevation angle estimates determined from the OIS data are computed using the known ground range between OIS transmitter and receiver and the measured OIS range between transmitter and receiver and assume a spherical ionosphere. The elevation take-off angle estimates determined using the MISO transmit array have been cross-checked with the same mode elevation take-off angles determined from the OIS data. Two example results are presented for the Mt. Everard path. These results are typical of many such elevation angle measurement comparisons analyzed. The first example was recorded during MSE-I and is listed in Table 2 while the second was recorded during MSE-II and is shown in Table 3. There is good agreement between the two different methods for determining elevation angle with root mean square difference between the two methods of less than 0.65° and 0.95°, respectively. Note that the MISO architecture enables these estimates to be derived from data recorded by a single receiver. Table 2. Table 3. Comparison of Take-Off Elevation Angle Estimates Determined Using the MISO Transmit Array and OIS Data Mode ElevDOD(°) ElevOIS(°) Comparison of Take-Off Elevation Angle Estimates Determined Using the MISO Transmit Array and OIS Data Mode ElevDOD(°) ElevOIS(°) 1E 12.4 13.0 1E 10.3 11.0 1F2l 31.4 31.0 1F2l 27.5 27.5 1F2h-o 34.3 34.3 1F2h-o 34.3 33.5 1F2h-x 38.4 39.4 1F2h-x 45.6 44.3 DECEMBER 2017 IEEE A&E SYSTEMS MAGAZINE 61