Data transfer methods in real time controller of Ion Cyclotron High Voltage Power Supply

摘要

Ion Cyclotron High Voltage Power Supply (IC-HVPS) with dual output (27kV & 15kV, 3MW) is operational at ITER-India lab with a 1.5 MW Diacrode based RF source to be used for ICRF (Ion Cyclotron RF) system. The controller for ICHVPS is designed with LabVIEW Real-time (RT) PXI controller to support all control and monitoring operations of the Pulse Step Modulation (PSM) based power supply. Besides regulation of output voltage, the controller supports all essential features like, fast dynamics, low ripple and protection for source and loads. This is managed by close integration between RT system and FGPA (Field Programmable Gate Array) inside the controller. Two types of data transfers inside the RT controller are implemented, 1) Between FPGAs and RT controller, 2) Between internal algorithms. RT controller and FPGAs are interconnected through PXI bus. Critical functions, signal generations, monitoring & data acquisition and interfaces are assigned to FPGAs. All these processes are managed by Real Time Operating System (RTOS) installed in the controller. Most of the processes initiated by RT controller ends up performing certain task by invoking the algorithm inside FPGA directing some action to outer interfaces. Some of the process routines are implemented for managing the internal algorithms. This paper discusses different techniques used mostly for transfer of information between RT controller and FPGAs. Information includes critical signals, data, measurements, commands and configuration required for internal algorithms. Algorithms use various methods like DMA-FIFOs, local variables, arrays, Trigger lines to transfer information. RT-FIFOs are used to handle bulk data in runtime to improve performance. Methods have been chosen carefully to fulfill critical requirements, balancing the available resources while managing limitations. Discussion also includes some specific choices made for synchronized and asynchronous transmission of parallel and sequential data.