Signal processing aspects of fusion plasma broadband reflectometry

Microwave reflectometry is a radar technique utilized by nuclear fusion diagnostics to evaluate the plasma electron density distribution (density profile) and its local fluctuations in experimental devices, e,g., tokamaks. It exploits the fact that an electromagnetic wave launched into the plasma is reflected at the layer where the refractive index vanishes. By mixing the incident and reflected waves, a phase-modulated reflectometric signal is produced. In O-mode broadband reflectometry, the density profile is determined by sweeping the frequency of the incident wave, estimating the phase-rate of the reflectometric signal and computing its Abel inversion integral. In this paper, a stochastic nonlinear filtering approach is adopted for the estimation problem. The joint phase and phase-rate dynamics is modeled as a vector Gauss-Markov process from which only the first component is observed. A suboptimal nonlinear filter tailored to the features of the problem under study is developed and tested by simulation and applied to real data. This estimator exhibits significant advantages over the extended Kalman-Bucy filter, which is used in this work as a benchmark