September 14th, 2010
At
my old company in one corner of the engineering lab there were a few
tables with a 25 amp inverter setup for testing. Over the years I had
laid claim to this area for my embedded controls development. For the
company's main product line I had designed one universal control board
which we used for all the models from a few amps at 120 volts single
phase all the way up to 2400 amps at 480 volts three phase. The
single-phase, 120 volt, 25 amp unit was perfect for development. It was
big enough that it sort of behaved like the larger units but small
enough that when it exploded it didn't make your ears ring.
I
spent most of my lab time banging away on the keyboards of my
development machines in front of that 25 amp test system. So much time
that I got a bit too familiar with the setup honestly. I became a bit of
a cowboy when it came to operation of the unit and reconfiguring it.
The
topology of the product was such that the inverter was normally series
connected with an input source voltage. I used a 120 volt manually
operated variac to act as my test input source voltage. One day I was
working on a proof of concept idea where we reconfigured the inverter
dynamically from series to parallel connection and back. There were a
few technical issues I was working through in attempting to do that. I
was trying out lots of ideas at that time, one of which was adding some
series resistance to a power connection to damp out some voltage ringing
that was occurring during the series to parallel transitions.
The
unit was connected to the variac with the standard two wires: line and
neutral, just like the common two wire power cord for household
electronics. I had a separate ground wire for the unit directly
connected to our lab electrical grounding system. I can't remember the
exact reasoning behind the connection methodology but I remember wiring
in the damping resistance from the inverter neutral to ground.
Anyway,
I was working away with the damping resistance installed and nothing
seemed out of the ordinary. I felt like a taking a soda break so I
turned off the input variac using its built-in on/off switch and walked
out of the lab over to the vending machines in the break room. I headed
back to the lab with my Diet Dew and walked into some sort of commotion
in the vicinity of my test setup.
One of the other engineers
(Don) was standing over my test setup looking rather agitated. Evidently
within seconds of me leaving the lab, smoke started pouring out. I had a
Plexiglas shield placed over my unit which acted like an upside down
fish tank, collecting the smoke as it rolled off the inverter. That
penned up smoked made it impossible to see what was getting hot inside
the unit.

Don had been in the lab working on a different test
setup. When he noticed the smoke and me nowhere to be found he ran over
to my bench and frantically looked for the off switch. To my credit I
had it clearly labeled with a sign and red electrical tape. However he
quickly found the switch was already in the off position. Don, who I
always admired for keeping his cool in situations like this, quickly
located the upstream connection of the source variac and yanked out the
cord. He confirmed the smoke was in remission, then started checking my
setup to verify the power really was off. That was when I walked back in
the lab.
When Don noticed I was back he told me what happened. I
had just started telling him that he must be confused because my test
setup was switched off when I left. Then I got a whiff of the burnt
electronics smell and noticed the blackened resistor in my test setup,
the same damping resistor I had just wired in about ten minutes ago. Not
only was it completely charred but it had started to roast all the
components around it. I think I set a personal record that day on how
quickly I had to eat my own words. Explosions and burning parts are par
for the course in a power electronics lab but what was disconcerting in
this instance was the variac was switched off but the power was
obviously still flowing into the unit. It was like a zombie variac:
supposedly dead but still very much alive.
Once Don knew the
situation was back under control he went back to his setup and left me
to figure out what happened. It took me a few minutes but I eventually
figured out that I had miss connected line and neutral after I rewired
the unit to add the damping resistor. If you remember I had placed the
resistor neutral to ground. Normally those two connections are at the
same voltage potential because they are bonded together at the breaker
box in the electric service to the lab. However I had flipped neutral
and line.
On top of that, the off switch I was using was not a
full two-pole disconnect, but instead a simple one-pole on/off switch.
The switch was meant to open up the line connection but the neutral
always remained connected. In my miss-wired state the switch was opening
the neutral with the line connection remaining connected to the unit.
All of this meant the damping resistor I added ended up with the full
source input voltage across it. I had sized the resistor for a few volts
but inadvertently applied 100 times that. Joule's combined with Ohm's
Law: voltage squared times resistance equal power or in my case smoke.
In
retrospect I was surprised the resistor didn't instantly burst into
flames. I tend to use over-sized components during development, which is
why it probably held on till I left for my soda break unfortunately.
From that day on whenever I left the lab I yanked the cord of the varic
before leaving.
The whole episode reminded me of a quote by Mark
Twain that went something like "loaded guns are perfectly safe but
unloaded guns are the most deadly and
unerring things that have ever been created by man." That
sentiment is timeless.