I thought I'd share a bit of personal experience with designing products for harsh environments. I worked in telecom for several years, designing equipment for last mile applications. We designed central office racks as well as pole mount remote equipment. Telecom industry standards are very stringent and require a lot of safety and environmental protections. I thought I'd share 2 scenarios, one in which I was able to modify a test procedure in order to pass a stress test and another where I thought I was done with validation testing, only to realize my setup was flawed and I had to redo it all. Ironically both of these scenarios involved heat related failures.
For the first entry let's talk about power cross testing (GR-1089-CORE). This standard is to ensure safety for equipment exposed to lightning strikes as well as AC faults (think power lines "crossing" telephone lines). Essentially the equipment has to survive different levels of fault and either survive and keep working or at least not catch fire. Several of the tests are repetitive, having to survive "x" number of strikes within some time frame. That time frame became key for me in one design where we were failing test. In this particular system we were receiving both xDSL communication and line power over a single twisted pair wire. There was a communication transformer attached to the line to transmit the xDSL signalling. Below is a capture of the circuit.
During this particular test, CR1 and CR6 would activate and short to ground during the test to protect the downstream circuit. During the test one device would fire before the other, causing current would flow through T10 for a short period until the other voltage suppressor activated. When that happened once it wasn't that big a deal. However, during the repetitive test what ended up happening is that around the 5th time you hit it the transformer was shorting. Each time you hit it the transformer temp spiked from the current flow through the primary winding. After a few consecutive hits the isolation material inside the transformer between windings began to melt. Once that happened the windings shorted and the board was dead. Because I wanted to get the testing completed as quickly as possible it was set up to keep hitting the device in fairly rapid succession the maximum number of times required to be compliant. I realized what was happening and why the transformer was failing by using high voltage probes that showed the ringing and thermocouples that showed the temp rise higher on each strike. I was able to stretch out the recovery time between hits to a point where the temp rise of the transformer wasn't increasing on each strike. By doing so it still fully met the requirements of the safety specification and it gave the transformer enough time between strikes for the isolation material to cool down enough that it didn't melt.
Next month I'll share how I once wasted a week of thermal testing when I discovered a variable I didn't account for.
