Charging precision analysis of a 40kW, 3kV soft-switching boost converter for ultra precise capacitor charging

Summary form only given. A compact and cost effective X-ray free electron laser facility is built at the PSI in Switzerland [1]. This laser system requires modulators with a pulse voltage of 370kV, a pulse power of 127MW, a pulse length of 3μs with a rise time of less than 1μs and a repetition rate of 100Hz. Furthermore, an ultra-high output voltage repetition accuracy of 10^-5 is required. The considered modulator including the capacitor charging unit is connected to the grid via an isolated PFC AC/DC inverter. To achieve a repetition accuracy of 10^-5, the capacitor bank has to be charged even more precise. In [2], the capacitor charger and the chargers control strategy have been presented. The converter is a MOSFET-based 40kW boost converter with a nominal input voltage of 1.3kV and output voltage of 3kV. The converter is operated in boundary conduction mode. Each switching cycle is independent of the preceding cycles in boundary conduction mode. The usage of MOSFETs instead of IGBTs enables a switching frequency of 70kHz up to 250kHz at the nominal operating point. The charge transmitted to the capacitive load during one cycle is lower with a high switching frequency and the system is less sensitive to random errors. In this paper, the precision of the converter presented in [2] is investigated. Since absolute accuracy is not required, only random errors like noise, jitter, quantization related errors or rounding errors during computation have to be considered. In order to investigate the repetition accuracy, the complete chain from the measurements, the quantization, the digital controller, the D/A conversion, to jitter in the switching signal is analyzed in this paper. First, all controller related measurements are investigated. This investigation includes an analysis of the sensor noise, the pre-amplifier noise and the ADC including quantization related errors. Second, the controller is studied. Since the controller is a d- gital controller, it has finite resolution. This in turn influences the repetition accuracy. Third, the D/A conversion including the related measurement and the comparator are examined. The D/A conversion also includes a quantization error and noise. Last, the jitter of the switching signal is investigated. The jitter is determined by measurements on the converter presented in [2]. The combined errors mentioned before result in a total rms jitter in the switching signal. By using an appropriate model of the converter, the rms output voltage jitter can be calculated.