Preliminary thermal overstress tests were conducted on IGBTs during system development. An International Rectifier IRG4BC30KD IGBT with a 600V/15A current rating in a TO220 package was attached to the transistor test board with no external heatsink. The collector-emitter junction was connected in series with a load power supply and a 0.2Ω load resistor. A 50Ω resistor was placed between the gate driver and the IGBT gate for current measurement. A thermocouple was attached to the IGBT case for temperature measurement. The gate signal was chosen to be a Pulse Width Modulated (PWM) signal with amplitude of 10V, a frequency of 10 KHz and a duty cycle of 40%, similar to a slow SMPS. A hysteresis temperature controller with set points of 329°C and 330°C was connected to the system, switching the gate voltage for its control mechanism. The load power supply voltage was increased from 0V to 4V over the course of several minutes until the heat sink temperature reached 330°C and the temperature controller began cycling. An additional temperature threshold controller, with a set point of 340°C, was programmed to turn off the load power supply and end the experiment in the event of thermal runaway and latching failures. IGBTs tested with this process were found to fail early in the test, within the first several minutes, or survived 1 to 4 hours before loss of gate control and thermal runaway was observed.
Figure 1: Latch up failure of an IGBT during thermal overstress.
Figure 1 shows the averaged collector-emitter current during a failure. The average ON-state current increases from approximately 4A to 10A indicating a transition from a 40% PWM duty cycle to a latched ON-state, resulting from a loss of gate control. This behavior may be attributed to the temperature-stimulated parasitic Silicon Control Rectifier (SCR) found within the IGBT. The experiment was ended quickly after the failure and the IGBT was found to be functional when returned to room temperature of 24°C. Characteristics of the degraded IGBT are currently being examined.
Both steady state and transient switching signals recorded during the test were analyzed. Steady-state voltages and currents showed minimal change throughout the test. Transient gate voltage and current also remained constant. Changes to collector-emitter voltage transient characteristics during turn-ON were also minimal. The collector-emitter current characteristics were not collected during this stage of development.
Figure 2: a) A voltage transient is generated across the collector-emitter junction. This peak decreases as the device degrades. b) A close-up of the transient peak.
A strong degradation indicator was observed when viewing the collector-emitter voltage turn-OFF transient. The peak voltage of this transient decreased significantly with both increases in temperature and thermal overstress degradation. Figure 2 displays the switching transient voltage, measured near 330°C, at different degradation stages.
Figure 3: A scatter plot of the package temperature vs. switching transient peak voltage for a single IGBT under degradation.
Figure 3 shows a scatter plot of transient peak voltage versus heatsink temperature with grayscale indicating aging state. A degradation trend can be clearly seen with transient peaks in similar temperature ranges decreasing over 10% during the course of the experiment. An indicator of semiconductor degradation under severe conditions is clearly observed. The root-cause is of course in question. This indicator could be intrinsic degradation; however, a more likely cause is thermal impedance degradation of the package causing increases in internal temperatures. Further investigation of this failure precursor’s cause is planned.