By Daryl Gerke, Contributing Writer
We were called in by a client to investigate a suspected EMI problem at an industrial facility in the deep South controlled by a single, large computer center.The processes were monitored with lots of analog and digital input sensors, lots of processing, and of course output control. The plant was experiencing high rates of error in its process control, which occurred whenever the door to the computer room was opened. But here’s the wrinkle: the problem started in the late spring, continued through the summer, and started up again the following spring.
The computer room opened immediately to the outside and directly faced the county airport a short distance away, but not line of sight. We immediately focused on it as prime suspect, although it was not at all obvious why such a problem would occur only in the summer.
Loaded up with instruments, including a spectrum analyzer and a power line monitor, we trooped over to the facility. It was a metal building, and the air-conditioned control room opened directly to the outside. There was a large UPS installed in the room, and the equipment was set on a false floor.
We set up our power line monitor to record power quality (we never assume the UPS will give clean power) and the spectrum analyzer and antennas to record RF. We measured ambient RF levels inside and out, with the door open and closed. As expected, we found appreciable differences inside and out, but no levels were high enough to raise a flag. The errors did not occur continuously, rather they surfaced at about the same time every day.
The CRT monitors were programmed to display the status of the various sensors. Data were displayed in green for nominal, amber for marginal, and red for out of tolerance, so it was easy to spot errors. At approximately the designated time, the errors started to occur. But our instruments did not detect any significant change from previous readings. Sadly, we are often in the position where our instruments don’t provide any significant information, leaving us to work by hunch.
We borrowed a handheld radio (about 15 MHz) from the maintenance guys. Keeping an eye on the monitor, we keyed the transmitter in the proximity of the sensor electronics (we estimated an electric field strength in excess of 10 V/m) with no apparent effect. Subsequently, we keyed the transmitter in proximity to all the electronics, again with negative results. We then tried installing additional ground straps to the false floor stringers and improved the various module and enclosure grounds, too. Nothing seemed to make a difference.
It was time to sit back and reflect on the symptoms. As we watched the errors, we became more aware of the time delay in the error indications. It would take about 30 seconds to ramp up to a full set of errors after opening the door and several minutes for the errors to go away after closing the door. Averaging by the software before reporting an error might be a cause, but it didn’t account for the source of interference. Stuck for an answer, we asked “What if it is not EMI at all?”
We had been looking at the door as an electromagnetic shield, but suddenly we saw it as an environmental shield. Once the light bulb went off, the answer came immediately: humidity.
The room’s air conditioner was positioned so that air flow was directed at the enclosures. When the door was open, the cold air tended to flow outside and was replaced by warm, moist outside air. Unfortunately, the analog circuit boards were located so as to get more than their share of the air flow, although they actually needed less cooling. Upon closer inspection, we could see condensation forming on bare aluminum members near the analog devices, which were conducting heat away. Somewhere in the high-impedance, low-level analog path, condensation was also occurring, effectively reducing signal levels.
We installed a cardboard baffle to divert the air flow away from the sensitive analog circuits. While this crude fix didn’t cure the problem entirely, it did provide tangible improvement. And since we used cardboard, we ruled out shielding as an issue.
The lesson here is that when looking for an EMI problem, don’t overlook other possibilities. Humidity, mechanical vibrations, and shock may all look like EMI.
Daryl Gerke is an EMI expert and co-owner of the consulting firm Kimmel Gerke Associates. This case was adapted from the EMI Suppression Handbook, by William Kimmel and Daryl Gerke.