Upgrade of DIII-D toroidal magnetic field power supply controls

The toroidal magnetic field power supply for the DIII-D tokamak is of the 12 pulse line commutated variety. It consists of four individual modules and a main system control cabinet which are combined to deliver 127,000 A and 1000 V to the toroidal field (TF) coil. The modules are connected in a series-parallel configuration but can be run alone or two at a time as well. Normally on DIII-D experiments, the series-parallel configuration is required. The original design provided each individual module with its own voltage (E) and current (I) control loop and a main control loop. A problem with this design was that the individual control loops would cause a current sharing imbalance in the parallel modules if the calibrated loops drifted by the slightest amount. It was determined that individual control loops were not needed and a single phase lock firing circuit was employed in the system cabinet with fiber optic links to the modules for gate drive signals. Since all four modules have to be on line for DIII-D to operate, a problem in any of the five E&I control loops resulted in the supply, and, therefore, the tokamak, being idled. By reducing the number of control loops to one, the sharing problem was eliminated, as well as 4 out of 5 potential control failures. The original supply employed relay logic for sequence control and fault monitoring. There were over 130 relays in each module plus an additional 100 in the system cabinet. The combination of the number of relays with the required interconnecting wiring, the age of the supply, the vibrations of the cabinets and the harsh environment, resulted in a continuously escalating number of "phantom," and often intermittent, faults. In some cases, a day of testing and searching was required to isolate the problem. The fault and sequence logic relays were replaced by a new programmable logic controller (PLC). All existing interconnect wire was removed and replaced with multiconductor cables that connect directly from fault sensors and input devices to the PLC. It is anticipated that an average of 12 h per year in control loop downtime will be eliminated. In addition, 100 man-hours of annual calibration time will he eliminated. New capabilities introduced by the PLC should provide for at least one additional plasma shot per day.