Thermal Response of a High-Power Switch to Short Pulses
- Created on Sunday, 01 June 2014
Simulations are used to calculate temperature changes that occur inside semiconductor switch modules, where measurement is not possible.
Army Research Laboratory, Aberdeen Proving Ground, Maryland
Semiconductor switch modules composed of Super Gate Turn-Off Thyristors (SGTOs) have been evaluated. The switches are intended to handle kiloamplevel currents and may dissipate peak powers measured in megawatts. Recent experiments measured the response of a switch module composed of eight SGTOs to single-short, high-current pulses. Simulations of those experiments were performed to calculate the temperature changes that occur inside the devices, where measurement is not possible. Worst-case operating conditions in which the switches handle several pulses within the space of 4 or 5 seconds (s) also were simulated. Modeling and simulation were performed with SolidWorks 3D modeling software and SolidWorks Simulation computational fluid dynamics software from Dassault Systèmes.
The original switch model includes a module composed of eight SGTO die, lugs and spacers, and base, as well as an aluminum cold plate to conduct heat from the module. In order to reduce the computation time spent in thermal simulations, the original model was simplified, reducing the curves and fillets to a geometry suitable to a rectangular mesh, while attempting to preserve the dimensions and volume of the parts (and materials) for time-dependent thermal calculations. The model was further simplified, and simulation run times reduced, by taking advantage of the symmetry of the model and using one-quarter of the simplified model for simulation. Spacers were eliminated from the model; they have no significant influence on the thermal response of the switch in the time-spans studied (see figure).
The initial temperature of the model and of the air surrounding it was defined to be 70 °C. The bottom surface of the cold plate on the switch module was also defined to be 70 °C; this served as a sink for heat generated in the SGTOs. The top surface of the circuit board was subject to natural convection, which was included in the simulation by defining those surfaces to have a convection coefficient of 10 W/m2-K.
The time-dependent thermal inputs for the simulations were derived from measurements of voltage drops and current in the SGTOs made during experimental characterizations performed with the solidstate switch modules. Curves of power versus time were culled from much larger files of data recorded at 0.1-microsecond (μs) intervals. It was assumed that the switch module distributes the power equally among the SGTOs. In the simulation, power was applied uniformly throughout the active part of the die, rather than applying it solely to a junction within the die.
While the active part of the die reached a temperature in excess of 130 °C, that part of the die is covered with a lid, and the observable part of the module has reached a maximum temperature of about 100 °C. Little of the heat in the first pulse has been conducted outside the active die in 35 μs. The “off” time is 1.66 s, at the end of which the die has almost returned to its initial temperature of 70 °C.
Next, how the temperature of the SGTOs rose and fell during a series of pulses was calculated. Two cases were used. The first was five 35-s pulses in a 4-s period, and the second was four 35-s pulses in a 5-s period to see whether or not the temperature of the devices stays within a safe operational range. SGTOs return almost to the original baseline temperature between pulses, and the average temperature rises with each pulse to just over 130 °C. Silicon devices may not operate as efficiently at such temperatures as they do at room temperature, but are not likely to be damaged by such short transits to high temperatures. There still is little sign of heat spreading through the module after 5 s, and visible temperatures are within one degree of those at the peak of the first pulse.
The second set of calculations used a 54-μs-wide pulse as a thermal input to the SGTO switch model. Peak power is 1.5 MW per SGTO, assuming equal distribution of power among the SGTOs in the module. As with the 35-ms pulse case, the thermal effects of a single pulse, a series of five pulses in 4 s, and four pulses in 5 s were calculated. The longer pulse raises the device temperature over 140 °C, about 10 °C higher than the 35-ms pulse.
The effects of 4- and 5-pulse trains with the 54-ms pulse were calculated. As each pulse was applied, the average temperature in the active part of the SGTOs rose to more than 140 °C, and returned almost to the baseline temperature between pulses. The SGTO temperature at the end of the last pulse in both cases was approximately 144 °C.
The model shows a quick rise in the temperature of the SGTOs during the pulse, but peak temperatures are within safe operating limits for Si devices. The heat in the devices dissipates quickly into the surrounding package and into the cold plate to which it is mounted. Heat conduction occurs rapidly enough that the devices’ temperature returns very close to their initial temperature within about 1 s after the end of the pulse. Calculations predict that the peak temperatures in the SGTOs will increase little over time when a series of short pulses over 4 or 5 s is applied. This indicates that the switches should be able to withstand repeated cycling at high power levels without sustaining heat-related damage or suffering serious degradation of performance.
This work was done by Gregory K. Ovrebo of the Army Research Laboratory. ARL-0164