Heating of the transistor surface occurred primarily around the emitter edges (see images). At higher drive conditions the entire transistor surface became hot, but the emitter edges were reaching the failure point. Failure occurred with melting of the aluminum conductor material. This is a typical condition that had already been characterized for operating conditions, which led to the eventual solution to the problem. The exact details of the problem will not be revealed here, but it was quite unusual and unexpected.
In an important large-scale project at another company, banks of high-power transistor drivers were failing, literally by the baskets full. No one on the project could come up with satisfactory answers, and Dr. Vlasak was tasked to solve the problem. He obtained as much data as possible and noted that the snubbers appeared to be well-designed, and that the transistors were apparently within their ratings. SPICE simulations also did not reveal any special anomalies.
The very large transistors were embedded in a very hard, strong plastic material that would not submit to removal without damaging the devices, so he sent the transistors to a lab where they were submitted to a high-energy surface removal process. The potting material would slowly flake off, and after several days of this treatment, the transistor chips appeared clean with no damage caused by the cleaning process. The failed transistors were also de-potted, and inspection showed that they were totally destroyed by heat. No other conclusions were reached by inspection of the failures, so not much was learned from the failed transistor. The unused opened transistors appeared bright and clean, and they did indeed work properly.
An unused transistor array was mounted on a small block under an infrared microscope(see picture at right). The knobs for adjusting the cross-feed stage are at the bottom. The IR detector, located at the focal plane, was cooled with liquid nitrogen to the operating temperature. The drive and operating voltages were slowly increased, and the temperature profile across the transistor array could be seen through the microscope. Photos of the transistor surfaces under various operting conditions were taken and later analyzed. The results indicated that the SPICE simulations were based on incomplete parametric data. See the results to the right.
See the IR Images of the operating transistors
See the Laundry Control System project
|[Previous]||Company Projects - page 7|
|[Next]||Company Projects - page 9|
|[Mail]||Send email to Adaptive Enterprises Web Site|