System identification of the linac RF system using a wavelet method and its applications in the SNS LLRF control system

For a pulsed LINAC such as the SNS, an adaptive feed-forward algorithm plays an important role in reducing the repetitive disturbance caused by the pulsed operation conditions. In most modern feed-forward control algorithms, accurate real time system identification is required to make the algorithm more effective. In this paper, an efficient wavelet method is applied to the system identification in which the Haar function is used as the base wavelet. The advantage of this method is that the Fourier transform of the Haar function in the time domain is a sinc function in the frequency domain. Thus we can directly obtain the system transfer function in the frequency domain from the coefficients of the time domain system response.

1 Introduction

The Spallation Neutron Source (SNS) Linac to be built at Oak Ridge National Laboratory (ORNL) consists of a combination of low energy normal conducting (NC) accelerating structures as well as higher energy superconducting RF (SRF) structures. The purpose of RF system modeling is to investigate the various cavity configurations in order to provide the correct requirements for the RF control system hardware. We use modeling as a way to specify RF components; verify system design and performance objectives; optimize control parameters; and to provide further insight into the RF control system operation.