Drive Inverter Mechanical Design Intern

Semiconductor failures are notoriously difficult to characterize, especially since they can be impossible to detect through non-destructive testing. The failure rate of Tesla's semiconductor devices (TPAKs) is very low, but production volume is so high that failures still matter. One failed TPAK means replacing the entire inverter. It also causes a loss of HV isolation, a Vehicle Off Road (VOR) failure that forces the driver to pull over immediately.

To tackle this, I spent extensive time creating a protocol to artificially generate TPAK failures that matched what we were seeing in the field. Using those failed units, I began developing a characterization test that could catch failures that previously would have passed the manufacturing process. The test relied on the steady state of a high-voltage polarization test, which wasn't yet viable for a production line. I also supplied other engineers with artificially failed TPAKs so they could test in parallel.

I then designed PCBs that mechanically clamp these devices and hold them at high voltage during environmental stress tests such as high temperature/high humidity and thermal shock. The results helped shape future tests and how testing might be integrated into the production line.

What I really enjoyed about Tesla was the availability of resources. I was interested in running tests on the semiconductors that would be prohibitively expensive at most companies. Tesla already had so much equipment and expertise that I could usually complete these projects with what we had in house.