Share your interesting engineering stories

A few days ago, I troubleshooted a really unusual and unexpected problem with a grid tie inverter I designed for senior design class. It had two large 220uF, 450V caps on the high voltage DC input. In my current configuration, the DC bus runs at 270V. I was powering it up for testing when I heard a startling bang and a flash of light. I was a little confused when nothing seems to have visible damage. After two more attempts to power it up, I traced the arcing to under one of the caps. So I discharged the cap and removed it, thinking something must have gotten under it. Turns out the via plating cracked and started arcing. I cleaned it up, added wires to jumper over the bad vias, and it worked just fine after that.

I have had plenty of other problems with the vias on that board and other boards. Other students have seen the same problem. Texas A&M's PCB service is so terrible, it's no wonder the Aggies are made fun of.

What is interesting is that it resembles a common physics problem where two capacitors are connected in parallel after one is charged and the other discharged. Solving that problem shows that the final energy is somehow half the initial energy. That unintended "experiment" clearly showed what happened to the other half!

And here's a picture of the board mounted in a case, with a tablet for scale:

Texas A&M has a PCB service? That's not so terrible.... :)

-- The Ruggeduino: compatible with Arduino UNO, 24V operation, all I/O's fused and protected

Not a very good one. Apart from the via issue, it only does two layers, no solder mask, and the finest pitch it can do is SOIC. That makes it completely unusable for modern logic designs since almost all of the really interesting digital parts (FPGAs/CPLDs, 32 bit microcontrollers, RAM) are smaller than SOIC. The yield I have seen is less than 25%, really bad considering that even the really cheap PCB manufacturing gets a yield close to 100%.

Any ideas what the PCB service is like at other universities, UT in particular? I would have asked my friends, except I don't know anyone at UT who has made a PCB or if such a service is even available there. I have seen a few boards by Oregon State that are really good quality.

At our local university (GVSU) they send boards out to Advanced Circuits who has a $33 special for students. Not as cheap as the aggregators/offshore services but boards come back in about 8 days. Don’t know about UT, but it seems to me like it would take a lot of $33 orders to justify running your own board shop.

The Gadget Shield: accelerometer, RGB LED, IR transmit/receive, speaker, microphone, light sensor, potentiometer, pushbuttons

Congratulations Mike! You reinvented the arc lamp :) Now you can proudly say you understand the unit of electric field is volt/m, high volt and small distance makes the field very strong and air is not very good insulator at that kind of voltage.

Regarding PCB, have you tried those services at seeedstudio or iteadstudio? They are cheap (slow though).

The project is just ending, just over a week left for more debugging. The hardware runs great and the firmware runs fine. I would keep those services in mind if I need to make a board in the future, though.

I've ordered from them several times each and they provide decent quality PCBs, just the waiting time is long, around 20 days.