Sometimes the facts in a power quality problem lead to a conclusion too implausible to believe. As a PQ consultant, it's important for me to approach a situation logically — no matter how illogical the explanation may initially seem to be. Many times the evidence may lead me to believe one thing, but logic and reason inevitably win out. And in the end, the conclusion is never as strange as it first appears.
As manager of information systems at a well-known medical learning center, my friend Joseph had recently completed a systems installation in the computer lab used by the center's students. The large networked system employed advanced technology, and the designers had paid special attention to the power requirements by providing backup systems and preplanned shutdown schemes in the event of a power outage.
The equipment worked flawlessly — except on cloudy days. As soon as a storm began to brew, the system failed. I had a hard time imagining dark clouds could be responsible — the bad weather must be causing a power disturbance. His team had installed monitoring equipment for several weeks, yet they had not detected any changes to the system. During the next storm, the team looked on as the system crashed again, but the monitoring equipment failed to uncover the cause.
I planned to visit the center during the next forecasted storm. When the weatherman predicted afternoon showers one day in April, I set out with a bag of test equipment in hand, determined to put an end to my friend's power quality problems. As the storm approached and the sky grew dark, our anticipation grew. With an oscilloscope connected to the mains for the server, I kept one eye on the storm through the window and the other on the scope display. Then, just as Joe had predicted, the server began acting up and shut down, yet I saw nothing on the scope! He was right — the cloudy sky appeared to have shut down his server.
As Joe began the process of rebooting the system, I started walking around the room looking for something — anything — that might have caused the server to fail. The lights were still on, so I knew there hadn't been a power disturbance of any kind. The only sound in the room was the buzzing of the fluorescent lights. As I listened to them, I realized the problem.
When the sky had darkened, the fluorescent dimming system had kicked in and brightened the lights. I pulled out my $5 AM pocket radio and walked around the room, listening for the static that electromagnetic interference (EMI) would produce. When I got to the wall near the server, the static increased and became even stronger as I moved the radio up the wall. The answer was behind the wall next to the server where the broom closet was located. Before we opened the door, I had an idea what I might find.
Looking into the broom closet, I could see a ceiling-mounted lighting transformer mounted next to the panel for the dimming system. I pointed out the problem to Joe with my flashlight. The transformers were not covered, and the unit had little or no shielding. As the sky outside darkened, the dimming system would increase the load on the lighting transformer.
The system's unusual oscillations probably also contributed to the problem. Further investigation revealed the dimming system went to full power at night, minimizing the interaction between the transformer and controller.
Once we put the covers back on the transformer, the server was no longer affected by the interference. Believe it or not, the server really didn't like rainy days.
EC&M Needs You!
In our next PQ Advisor, we'll have two opportunities for you to share your PQ knowledge with fellow readers. We're looking for PQ troubleshooting tips and simple solutions to specific PQ problems. We'll publish the solutions in a Q&A format. Questions are acceptable, too.
You can submit a total of four items. Send all submittals to [email protected]. We'd like a small e-mail message, per item, of no more than 250 words. We may contact you for more details. Of course, we'll give credit for any tip or solution we use in the final article.