Interpretation of measured data and the resolution analysis of the RTP 4-channel pulsed radar

Abstract

In the present work, the resolution of a 4-channel pulsed radar being built at Rijnhuisen for RTP tokamak is analyzed. The achievable resolution depends mainly on the accuracy of the time-of-flight measurements and the number of sampling frequencies; since the technological solution and the configuration have already been set, the emphasis is laid on the interpretation of the measured data (inversion problem) and minimization of the overall error. For this purpose, a specific neural network — Multi Layer Perceptron (MLP) — has been successfully applied.

The central density in the range 0.2–0.6×10²⁰ m⁻³ was considered, i.e., above the critical density for all four frequencies but not too high to restrict the measurements just to the edge of plasma. For a wide class of density profiles, by balancing the inversion error and the time measurement error, the overall error in estimating the reflection point position between 0.72cm (for the lowest frequency) and 0.52cm (for the highest frequency) root mean square was obtained, assuming an RMS error of 70 ps in the time of flight measurements. It is probably much better than what could have been obtained by Abel transform. Moreover, the mapping by MLP is incomparably faster and should be considered for routine multichannel pulsed radar data processing.