A Little Lesson in DC Power

This happened several weeks ago, but I have been pretty busy, and Conni is always on the laptop. In her defense, it is her laptop. On with the story...

Most of you know that I work in a lab with fairly large network equipment. Part of my testing involves power failures and recovery, as well as redundancy. On this lovely Monday morning, I was testing the power conditioning units in a chassis. All computers (including laptops, desktops, servers and network appliances) run on DC power. We don't have DC power feeds in our lab, because it is expensive, dangerous, and the line loss is pretty high over anything longer than about 30 feet. So we run AC to the rack, and have a rectifier turn the AC into DC. To give you an idea of the amount of power, the rectifier takes four (4) AC inputs, and has two DC outputs, one for each power conditioning unit in the chassis. I can picture everyone (but mostly brooke) with their eyes glazed over. On with the story!

I had to simulate a power outage, and remove the PCU from the chassis. Being a belt AND suspenders kind of guy around electricity (hi Matt!) I decided that I would not only turn off the switch in the PCU, but also disconnect the PCU from the rectifier. The smart course would have been to remove the power cord from the PCU, but I wasn't sure that the cord (DC, remember?) didn't have exposed prongs, and I didn't want to be waving live DC around in one hand while I fiddled around with the PCU with the other. So I disconnected power from the side of the rectifier which fed the PCU (which is about the size of a shoebox) and then turned off the PCU with a switch, and went to remove it from the chassis.

Now the exciting part--a very large arc as the PCU in my hand grounded itself to the chassis, about 6 inches in front of my eye. I couldn't see very well, as the spark left a fairly good-sized afterimage, so I exited the lab (in a bit of a hurry) and grabbed one of the other engineers to assist. I believe I said something like "I can't see very well, and we need to isolate power from rack 7 before we have a fire or someone gets hurt". We very carefully removed the remaining AC leads from the rectifier, which isolated the chassis entirely, and surveyed the damage. The arc had eaten a hole in the side of the chassis about the size of a dime.

In the post-mortem, we decided that the rectifier had a common bus. That means that all four AC power supplies lead to one common DC feed, which then split into two cables, one for each PCU. So my disconnecting two AC cords didn't really do anything. But, you ask, what about turning the PCU off at the switch? Surely that means there was no power to the PCU? Well, there was no power THROUGH the PCU, but there was still power to the switch. And it turns out that I brushed the hot power lead to the switch against the side of the chassis, which completed the circuit.

Something funny did happen later. My boss, his boss (a director named Scott, he reminds me of Dad) and I were discussing the event, and Scott said "Well, we aren't going to run this test anymore, let the vendors do it." My boss, Raj, said something like "After we initially qualify a chassis, there is no need to run this test again." Scott said "Didn't you hear me? We are NEVER running this test again. It isn't worth having an engineer smoking and twitching on the floor." Which still makes me kind of giggle. BTW, I stopped by the Lab Manager, and thanked him personally for ensuring the chassis were grounded to the racks, and the racks themselves grounded. Do you want to know why everything was grounded? I had picked up some stray voltage some months previously, questioned why the racks weren't grounded to spec, and stopped testing until they got fixed. True.