| Abstract: |
High stability and accuracy of microwave and millimeter-wave radiometer measurements are required to obtain reliable radiometric data products for improving weather forecasting, hydrology, agriculture, climatology and cryospheric sciences. Frequent and accurate radiometric calibration is critical to achieve the required stability and accuracy for radiometric measurements. Two-point, end-to-end radiometer calibration is generally performed using external calibration targets (one of which may be the cosmic microwave background) at two widely separated temperatures, assuming that the radiometer has a linear response. However, radiometers exhibit non-linear responses due to inherent, undesirable nonlinearities in amplifiers and detector diodes. Furthermore, fluctuations in gain of amplifiers and responsivity of detector diodes in radiometric receivers increase the uncertainty in radiometric measurements. Internal calibration techniques, including noise diodes and reference loads, are used to improve the stability of microwave and millimeter-wave radiometers. However, the effectiveness of internal radiometer calibration techniques is limited since these techniques do not perform end-to-end calibration. Internal calibration techniques omit gain variations and instabilities in the antenna as well as the portion of the front end before the noise diode or reference load. Furthermore, these internal calibration references suffer from inherent limitations, e.g. instabilities in noise diodes that limit the accuracy of the calibration.
CubeSats and other SmallSats have emerged in recent years in increasingly significant roles in Earth observation and remote sensing. They offer a path to reduce cost, shorten development times, accelerate the introduction of new technology in space and increase overall access to space. CubeSats with microwave and millimeter-wave radiometer instruments are planned for in-space technology demonstration and science missions, including MIRATA, TEMPEST-D and TROPICS. The stringent requirements on mass, volume and power for CubeSats have introduced new challenges in the calibration of microwave and millimeter-wave radiometers for CubeSat deployment. External calibration targets tend to be large in size, making it difficult to meet dimensional requirements for a CubeSat instrument. In addition, external calibration targets tend to limit the available swath and duty cycle for Earth observation.
This study proposes a deep learning neural network model based technique for calibration of microwave and millimeter-wave radiometers, with particular application to those designed for CubeSat and SmallSat platforms. The noise-wave representation of a Dicke radiometer has been used to generate synthetic radiometric data to demonstrate the proposed deep learning calibration method. Calibration demonstration for synthetic radiometers with a variety of noise properties have shown that the new technique performs radiometric calibration without adding any significant noise to the system. The new model can be applied to microwave and millimeter-wave radiometers as the sole calibration technique or as an augmentation of existing methods. The proposed model is expected to improve the accuracy and stability of microwave and millimeter-wave radiometers by providing an improved method for radiometric calibration for satellite missions, especially CubeSats and other SmallSats.
|
