Data driven modeling for linearization of particle accelerator RF power source

Abstract Radiofrequency (RF) power source during high power operation of a particle accelerator often exhibit saturation nonlinearity in amplitude and phase for sinusoidal input signal. This leads to an undesirable distorted RF signal as input to the RF cavity, which is the basic building block of an accelerator. Linearization of RF power source is an important necessity for an operational accelerator to improve and simplify RF cavity control performance, in addition to improving efficiency. The RF power source being a final control element of the RF cavity electric field control loop, its linear operation over a wider RF input amplitude range is often sought. Digital predistortion has been one of the preferred approaches for linearizing RF power source. This approach needs the determination of the input-output characteristics of the predistorter from those of an RF power source. Among many approaches, least squares estimation techniques are suitable for the design of digital predistorter. Among them, a recursive least squares estimation is most suitable and is elaborated in this paper. Amplifier nonlinearity and its handling using adaptive least squares estimation is explained, followed by modeling and simulation results. Input-output data acquired from an experimental 200 W, 30 MHz RF amplifier is used as input for simulations to demonstrate linearization capability by predistortion. After building up the basics, a technique for amplitude and phase linearization for data generated from the Saleh model of an RF power amplifier is presented. For an assessment of linearization capability, one needs to measure it. The effect of nonlinearity extent on linearity as a function of the number of model coefficients is analyzed. Experimental data is used to analyze the effect of nonlinearity extent and measurement noise. This paper is a comprehensive coverage of the design and analysis of a linearization scheme and is applicable for any general engineering application. A digital predistortion algorithm needs to be implemented in a control system for the RF cavity. This problem forms an important inverse problem of general nature under the purview of the modern control system community.