March 2024. 5:47 p.m. I was one step out of the office door when my phone rang—the kind of ring that makes you stop walking. It was a plant manager from a contract manufacturer outside Minneapolis. One of their packaging lines was down, they had a truck scheduled for 6 a.m., and he had already made up his mind about the cause.
"It's the TDK-Lambda power supply in panel C300," he said. "No DC on the bus. I need a replacement by morning."
I've handled these calls for a long time. In my role coordinating field support and emergency replacements for OEMs and industrial users, I've probably worked through more than 200 urgent requests. The exact number is in our CRM—I'd have to check, but the pattern is clear. When a power supply suddenly dies, the story isn't always about the power supply.
The part that wasn't dead
Line C300—that's the name on the equipment list, not a code name—assembles sterile medical procedure kits. It's not the most complicated machine I've ever seen, but it has enough moving parts: servo drives, a heat sealer, a label applicator, a camera check station, all running off one 24 V DC bus supplied by a single DIN-rail power supply.
The plant manager's first test sounded reasonable. He measured the DC output terminals and saw 0.4 V. His conclusion: dead supply. But he hadn't disconnected the load.
That's the first lesson. If a DC power supply reads near zero at its output while the load is still connected, you've learned that the voltage is gone—not why. To test the supply itself, you need to isolate it or at least check with the output disconnected. I asked him to do that before I panic-shipped anything. He didn't have time to argue. So I drove to the site, about 40 minutes away.
I brought a replacement TDK-Lambda power supply with me anyway. If the old one was truly dead, they needed it. The cost of the rush wasn't the issue. The issue was that the clock kept moving.
What the multimeter actually said
When I got there, the maintenance electrician had his 117 multimeter on the cabinet cover. A 117 is a solid, workhorse meter for this kind of troubleshooting—it measures true RMS AC voltage, and it has the continuity beeper that makes short-hunting easier. He handed it to me, and I started from the input side.
At the supply's AC input terminals: 120.2 V. Good.
At the DC output terminals, with the load cable still connected: 0.4 V. That's what the plant manager saw.
Then I shut down the incoming breaker, waited, disconnected the two DC output wires from the supply, and re-energized only the supply. The output jumped to 24.08 V. Actually, 24.08 V under no load—the key point is the power supply was fine.
The TDK-Lambda power supply wasn't dead. It was doing something almost as useful: it was refusing to power a load that had a short to ground somewhere downstream.
The cable that wasn't ready
Now, with the output wires free, the electrician and I started tracing the DC path. The supply fed a terminal block, and from there, a bundled cable ran up through the cabinet to the servo bus. In that bundle, we found the problem. It was a splice—a section of cable that had been joined years ago, probably when someone added a new sensor supply. The splice had been tucked behind a wire duct, close enough to a sharp cabinet edge that the insulation had been wearing for a long time. When the wires got warm and the cabinet vibrated, the cable shifted just enough to short against the frame.
The 117 multimeter told the story. At rest, the conductor-to-ground reading was open. When I wiggled the bundle, the beeper went off. Intermittent short. It didn't show up as a dead short because it wasn't one. It showed up as heat, intermittent OCP events, and eventually a supply that repeatedly latched off until the fault cleared.
The supervisor looked at the damaged splice and asked a question I still think about: "When was this cable ready for service?"
He meant it literally. The cable had been installed for years. It had carried current. But installed and ready for service are two different things. It was not ready for service on the day it was spliced, because nobody tested what would happen when the wire moved, warmed up, or rubbed against metal. It only became ready for service after we cut out the splice, ran a continuous cable, and verified continuity and isolation.
Why this wasn't a battery problem
At one point, someone in the meeting joked about whether TDK solid state battery technology would make these old power cabinets a thing of the past. I had to smile. I'm not a battery engineer; I can't speak to TDK solid state battery details or its product roadmap. From what I've read, the early work is aimed at compact, low-power applications, not at replacing an AC-DC power supply in an industrial control panel. More important, this failure wasn't about energy storage. It was about a cable that should have been tested years earlier. A better battery wouldn't have fixed it. A different power supply wouldn't have either.
Power supplies can be the victim
There's a common assumption in downtime situations: if the output voltage is zero, the power supply is the culprit. In my experience, the causation often runs the other way. A supply with overcurrent protection is supposed to shut down or latch off when something downstream tries to pull more current than it can safely deliver. The power supply didn't cause this failure. It was the first component to report it.
I'm not saying every power supply that reads zero is fine. Sometimes the supply really is dead. But before you spend money on a rushed replacement, spend ten minutes on diagnosis. If I can leave you with one thing from this rush call, it's this:
- If a DC supply reads zero with the load connected, disconnect the load and test again before ordering a replacement.
- Use a multimeter with a continuity mode to check output cables for intermittent shorts. Wiggle the wiring while you test.
- When an old machine suddenly stops, look for cable damage near sharp edges, wire ducts, and vibration points.
The end of the story is boring, which is how I like it. We replaced the damaged cable, reconnected the TDK-Lambda power supply, and the line started up. By 2:10 a.m., the supervisor wrote "C300 ready for service" on his work order and initialed it. The truck left before 6 a.m. No penalty. No emergency replacement was needed, and the customer was glad we didn't sell them one.
Later that week, the plant manager asked me whether they should upgrade the panel to a newer TDK-Lambda power supply anyway. My honest answer: the existing unit was still within specification, and the failure wasn't its fault. If they wanted a spare on the shelf, that was a reasonable call. But replacing it wasn't the fix. The fix was the cable that hadn't been ready for service for years—and only became ready when someone tested it, instead of assuming it was fine because it was connected.
