| Abstract: |
Artificial Radio Frequency Interference (RFI) is an ever-increasing problem in remote sensing microwave radiometry. Satellite data have shown that a significant amount of RFI is present up to Ku-bands at many locations around the world. In addition, the RFI problem is getting worse. Therefore, effective methods and systems to detect and remove RFI are needed for the next generation satellite radiometers. Ideally, such systems should combine sub-Kelvin RFI detection performance, low down-link data rate, and intelligent RFI filtering that preserves the uncontaminated radiometer data.
In this work, we have developed and tested a real-time RFI processor breadboard to detect and filter out RFI. The processor was designed to be compatible with the European MetOp Second Generation mission’s Microwave Imager (MWI) instrument, 18.7 GHz channels (MWI-1). The developed RFI processor applies the following detection algorithms for RFI: (1) kurtosis, (2) cross-frequency, and (3) anomalous amplitude detection. To improve performance, data is divided into sub-samples in time and frequency with fine resolution. The RFI processor can detect and filter out RFI with this fine resolution in real time and then integrate the clean sub-samples over time and frequency. Thus, a cleaned version of the radiometer data can be downlinked at traditional, low data rate. The processing functions are implemented in a reprogrammable Xilinx Virtex-5 FPGA with high processing capacity, providing high flexibility.
The electronics of the RFI processor includes the analog-to-digital conversion of two RF input signals (vertical and horizontal), data processing in FPGA, voltage conditioning, generation of clock signals, and telemetry circuitry. All power, data, and command interfaces have been duplicated for redundancy. The RFI processor was designed and implemented to provide high degree of equivalence with a potential Flight Model, both at equipment and component level. As baseline, the applied components are form-fit-function and manufacturer equivalent with Flight Model components.
The measured performance of the RFI processor corresponds the simulations and good overall detection capability has been achieved, especially for narrow-band RFI. Detection threshold values are dependent on the selected parameter values (e.g., integration time, number of frequency sub-bands, false alarm rate, detection probability, etc.). With the parameter values applied here, detection level of 0.4 K has been demonstrated for 0.1 % duty cycle RFI. Detection level of 1.3 K has been demonstrated for any duty cycle above 1 %. In addition, detection level varying from 1.3 K to 2.3 K has been demonstrated for narrowband (1 MHz) and wider band (25 MHz) continuous RFI.
The budget values are reasonable and within target values: power consumption is <12 W (at room temperature) and mass is <1.1 kg. The test results obtained indicate that the developed RFI processor works well and as expected. The bandwidth requirement of MWI-1 (200 MHz + 25 MHz guard bands at both sides) can be fulfilled and different RFI detection algorithms could be implemented. However, the capabilities of the RFI processor hardware have not been fully exploited; additional algorithms and higher number of sub-bands could be incorporated in the FPGA. Especially, additional algorithms would be beneficial to improve the detection threshold of wider band RFI (e.g., 100 MHz). |
