The Call in Late March
Last March, on a Thursday afternoon, I got a call from a machinery builder whose new control cabinet had failed at a customer's site. It was the third one that month. Or rather, the fourth, if you counted the one the customer had already 'repaired' themselves. Their packaging line kept dropping out mid-cycle, and one cabinet had actually let out the magic smoke. They'd built twelve of these units, and three were down. They asked us to come audit the design before the customer canceled the whole order.
I'm a quality manager. I've spent the last four years reviewing BOMs, checking components, and rejecting parts that don't meet spec. My job is to catch problems before they reach the field. But this one got through, and I wanted to know why.
The Machine Builder's Background
The builder had been using Siemens contactors in their machines for years. Their senior engineer retired in 2024, and a new engineer took over the electrical design. In one of those cost-reduction reviews, the new engineer replaced several specified components with 'equivalents.' The IGBT was one. The remote IO module was another. The contactor stayed Siemens, but the coil voltage was changed from 230 V AC to 24 V DC to simplify the control wiring. That decision, on its own, wasn't wrong. It just created a new set of constraints that nobody checked.
What I Found in the Cabinet
The first thing you notice when a cabinet fails in the field is the smell—burnt enamel and epoxy. The customer had already cut power, but you could tell which module had died. In this case, it was the IGBT in the motor drive. The drive's main switching device was an IGBT IC, specifically an IGBT H20R1203, a 20 A, 1200 V device in a TO-247 package.
The BOM showed something interesting: the original design specified the H20R1203 from a qualified supplier, but the builder's procurement team had swapped it for a lower-cost replacement from a broker that wasn't authorized to sell the part. The part number was the same. The marking looked the same. But after a few hours of runtime, the case temperature on the replacement was about 20 degrees Celsius higher than the original on the same heat sink. What most people don't realize is that static ratings are only the beginning. Gate charge, switching losses, and thermal impedance matter just as much.
I don't have hard data on how many field failures trace back to component substitution, but based on our audits over the past four years, I'd guess about a third of small panel builders have at least one 'equivalent' that isn't. The surprise wasn't the replacement, though. It was how many other pieces in that cabinet were connected to it.
The Small PLC, the Touch Screen, and the Analog PLC
The control system used a small PLC on the enclosure door, paired with a PLC touch screen for the operator. The small PLC was actually an analog PLC—four analog inputs, two analog outputs, and sixteen discrete I/O. It handled a temperature loop, a few interlocks, and the main motor start command. Nothing exotic. But the touch screen was flickering under the shop lights, and the integrator had written that off as a grounding problem.
It wasn't a grounding problem. It was a voltage drop on the 24 V DC control rail, and it was affecting more than just the display.
Testing the Whole Circuit
We started with the obvious suspects: the gate resistor, the gate driver chip, the snubber circuit. Those were fine. The failure was in the IGBT itself, but the trigger was elsewhere. So we moved to the control circuit.
We set up a meter and an oscilloscope on the coil of the main Siemens contactor (a Sirius 3RT series, 24 V DC coil). When the small PLC commanded the contactor to close, the coil voltage dropped from 23.8 V to 20.4 V. That's right at the pickup limit for most 24 V DC contactors, around 85% of nominal. But the remote IO module that drove the coil didn't have enough output capacity for the coil's inrush current. The voltage sagged just enough to make the contacts chatter.
And here's the part that surprised me: the IGBT failure was downstream of that chatter. When the contactor reopened under load, the arc re-struck and sent a current spike through the drive. The replacement IGBT, already running hot from higher switching losses, didn't survive the resulting 300 A spike. No IGBT does.
The chatter also explained the flickering touch screen. Every time the contactor dropped out, the voltage transient on the DC rail created noise that the analog PLC's input stage couldn't filter out. The thermocouple reading jumped by about 20 degrees. The operator thought the heater was broken. It wasn't.
The Fix
We rebuilt the BOM with three changes:
- Switched back to the original IGBT H20R1203 from a qualified distributor.
- Added a separate DC supply rail for the contactor coils, with a proper relay buffer instead of driving the coil directly from the remote IO module.
- Changed the Siemens contactor to a version with a lower inrush coil.
I called the component broker before we made the changes. They said, 'it's the same part number, what's the problem?' I sent them a scope capture showing the gate waveform and the current spike. They stopped returning my calls.
After we implemented the fixes, the next twelve cabinets powered up with zero IGBT failures. There's a quiet satisfaction in watching a ten-second power-up test pass without sparks.
What I'd Do Differently
The real issue wasn't any single component. It was the assumption that components are interchangeable if the static ratings match. A contactor is not just a contactor. An IGBT is not just a switching device. The dynamic behavior—gate charge, switching losses, coil inrush, output current—affects the system around it.
Reference: IEC 60947-4-1 covers contactor performance requirements, and IEC 61800-5-1 covers drive system safety. Both assume the designer validates the circuit, not just the part number.
There was a moment during testing when I doubted myself. The replacement IGBT measured fine on the curve tracer. The static parameters were all within the H20R1203 datasheet. If I had stopped at the component test, I would have signed off on a cabinet that would fail again. That's why we don't stop at component tests anymore.
What I wish I'd done differently: insisted on a full circuit review before production, not a quick component check. The piece-parts were mostly fine individually. It was the interaction that killed it.
Now, every cabinet we certify gets a dynamic test: power up, cycle the contactor ten times, measure the coil voltage under load. It takes twenty minutes, and it's saved us from at least three more field failures this year. The efficient path isn't skipping the test. It's doing the test before the customer does it for you.
If you're reading this because you have a similar cabinet on your bench, start with the DC supply. Measure the coil voltage while the contactor pulls in. Measure the IGBT case temperature after ten minutes at full load. And check whether the remote IO module is rated for the inrush current, not just the steady-state current.
