IGBTs Transmit Woe to Analog Signal Conditioning

Back in the mid to late '90s, I worked for a small engineering firm that makes automated testing equipment for the fluid power industry; test stands for things  like hydraulic pumps, motors and valves.  In combination pump/motor test stands, we  typically used Variable Frequency AC drives for the prime mover and loading device. On one particular machine, we selected a new and new-to-us VFAC drive based on its price, small size and advanced capabilities. This one was rated for 100 HP (74 kW)  and ran on 480 VAC, three phase power.

This drive was working out nicely, but for one BIG problem, noise. Our single-ended (long story) analog  signal conditioning for flows, pressures and torque were picking up noise in a big way.   Intermittent, irregular and relatively infrequent, HUGE spikes were swinging the proper signal either high or low by as much as 25% full scale.  It didn't take long to realize that the VFAC drive was the source.  Stop the drive and everything became peaceful again.

Though I studied EE some at school, I'm primarily a software guy.  I tried ferrite chokes, filters, and variations on signal cable and shield termination.  Nothing seemed to help much and I had pretty much used up my knowledge on the topic. Troubleshooting this problem quickly became a collaborative effort between me, the technicians I worked with, the North American drive reseller and the Japanese drive manufacturer.  By this time, the test stand was late and our customer was in a bind.  They already had an assembly line piling up parts that, by their certified ISO process, had to be tested before they could be sold.

The reseller sent their resident EE to our plant to have a look. Our effort at chokes and such weren't working because the very air was filled with high frequency pulses. One could hold a scope probe in the air and see a classic ringing waveform on the scale of tens of nanoseconds. Every wire that wasn't transmitting this was sure as heck receiving it.  The drive OEM finally sent their senior design engineer from Japan to our little factory here in Nowhere, Oklahoma.

After much probing and brow-wiping, we were told that the IGBTs used in this drive were 5th generation, the newest at the time, and offered great benefits from the efficiency gained by their high switching speed.  Well at 600 Volts on the DC buss, there's a lot of potential being modulated at a very high speed.  Every time the  IGBTs fired (about 14,000 times per second per IGBT), the airwaves lit up.  We only saw irregular and infrequent spikes in our analog signals because our sample rate was so low and the duration of the pulses so short. But when we did see a spike, it swamped our signal.

A few tweaks were made to the drive's internal settings, to lower the modulation frequency and stretch the switching time a little. We even added a filter between the drive output and the motor leads. In the end, however, optical isolation was the real winner. Every interface wire that connected to the drive was acting as an antenna to transmit that noise throughout the entire test stand. Full credit goes to the reseller's EE who tried optically isolating the digital interfaces to the drive and the result was dramatic; we had a  solution. We wound up designing an optical isolator circuit board of our own that handled all digital and analog interfaces to the drive, which we used in all  relevant machine designs going forward.

Reagan Thomas studied EE and EET in school, but wound up becoming a software guy who dabbles with electrons. He worked 18 years designing software (and some hardware) for automated testing equipment. He now deals with software for radar video distribution equipment.