COMPRESSIVE SAMPLING AT FREQUENCY MODULATED CONTINUOUS WAVE POLARIMETRIC DOPPLER WEATHER RADAR (PDWR)

• Main Author: Purnamasari, Rita
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/63980
• Institution: Institut Teknologi Bandung

A weather radar is a type of radar used to detect rainfall, its movement direction, and speed to determine the level of rain intensity and cloud composition. IDRA (IRCTR Drizzle Radar) is a weather radar in the Netherlands used to measure sensitive atmospheric phenomena such as drizzle. IDRA is a Polarimetric Doppler weather radar developed at the raw dataBand frequency with the principle of wave transmission in the form of FMCW (Frequency Modulated Continuous Wave). Several of these techniques are sought so that IDRA produces high resolution in measuring rainfall. Although the IDRA PDWR (Polarimetric Doppler Weather Radar) has received good results, its measurement process produces a tremendous amount of polarimetric data. The receiving device on the IDRA can collect almost one Giga Byte of data every minute. All of this important data will later be processed in the signal processing hardware section to become a weather radar product that can be displayed and read by users. In the conventional signal processing method, signal sampling uses the provisions of the Nyquist theorem so that information regarding the distance to the observed object can be reconstructed (reconstructed) perfectly without any errors. This provision requires that the sampling frequency is at least twice the signal frequency; as a result, data acquisition in the signal processing system has a heavy burden because the required storage media is also getting more significant. In this decade, a new technology emerged called CS (compressive sampling) to overcome these problems. CS is a new paradigm that states that signal sampling can be carried out with frequencies far below the Nyquist rate to retain the original information of the signal as long as it meets the signal rarity requirements in a particular domain.In CS theory, signal sampling and reduction are carried out simultaneously (CS). With CS efforts, signal acquisition on the main signal processing device PDWR can save bandwidth. In this dissertation, we propose a CS technique for the beat signal by exploring the rarity of the beat PDWR signal in its transformation domain. The proposal to reduce the beat signal in this transformation domain is considered a scientific choice because of the rare signal that will be obtained when performing IFFT (Inverse Fast Fourier Transform) on the beat signal. Since the IDRA radar is an FMCW radar, the beat signal contains both phase and frequency components in it. Conventionally, by performing IFFT on the beat signal, the distance information between the radar and the observed object can be known, and the beat signal becomes a compressed form from before. For this reason, the spacing transfor- mation with IFFT is appropriate to be applied to the proposed CS method. Unlike other studies that use simulation data to prove the proposal, the data is processed from real weather data in this dissertation. The proposed CS-based PDWR scheme begins with sampling and reducing the beat signal using a random type of measurement matrix. Then after being successfully reduced and sent to the signal processing device, the compressed beat signal is reconstructed again using the OMP algorithm (Orthogonal Matching Pursuit). This proposed method is then evaluated by constructing a PPI (Plan Position Indicator) to measure reflectivity and average Doppler velocity. Several IDRA weather raw stress data at high and medium rainfall intensity levels were tested in the simulation stage. Compared to the conventional method, the CS method on the PDWR is able to reduce the data significantly without eliminating the main information from the actual weather. In this study, the proposed method works well for quantizing and classifying whether deposit conditions in the atmosphere, especially when the amount of sample used is above 25% of the original data amount.