Before you read further, I should put a label on myself. I've been handling UPS and generator service orders for eight years. In that time, I've personally made and documented eleven significant mistakes, totaling roughly $46,000 in wasted budget. I stopped counting after a while. I now maintain our team's checklist, and the only reason I'm writing this is so you don't repeat my errors.
The story starts with a phone call in January 2025. A regional data center had two Eaton 93PM UPS units in a 2N configuration. Both units are Eaton online UPS units—true online double-conversion topology. That means the load is always running on the inverter, and the DC bus between the rectifier and the inverter has to be healthy.
The alarm log showed the units transferring to bypass four times in two weeks. No utility event lined up with the timestamps. The previous technician had already sold them new battery strings. To be fair, that recommendation wasn't crazy: the batteries were over seven years old. But the transfer to bypass happened again three days after installation. That was the expensive clue.
I checked the new batteries anyway. Conductance, voltage under load, impedance. All fine. The UPS kept logging a DC bus undervoltage warning. The facility manager needed an answer within 48 hours because the maintenance window was already scheduled. Every report pointed at the battery string. The alarm even said 'DC bus.' My gut said the batteries were a scapegoat. I told them to hold any further component orders and let me run one more test.
Here's what I should have done earlier: check the DC bus capacitors. In an online UPS, the rectifier charges the DC bus, and the inverter draws from that bus on every switching cycle. When a capacitor loses capacitance, the bus voltage sags when the inverter pulls current. The UPS sees that as a fault and transfers the load to bypass. It's not doing anything wrong. It's protecting the load from a weak internal power path. But the result is still a visible, embarrassing transfer event.
People ask me what tool catches a bad UPS capacitor. The answer is usually a multimeter with capacitance mode—and a safe procedure. I locked out the UPS, waited for the DC bus to discharge, verified zero voltage, and connected a bleed-down resistor across the bus terminals. No shortcuts. A charged capacitor can hurt you badly.
Once the bus was proven dead, I set my multimeter to capacitance mode. The capacitor nameplate said 470 µF. The meter settled at 61.7 µF. The next cap read 468.3 µF, and the third read 452 µF. The bad cap had lost almost 90% of its rated capacitance. That was the root cause. It didn't show up as an open circuit or a short on an ohmmeter. It just couldn't hold energy anymore.
I know 61.7 µF doesn't sound dramatic. But the inverter is switching thousands of times per second, and it needs that capacitor to stabilize the bus between switch cycles. When it can't, the bus collapses at the worst possible moment. Checking capacitor with multimeter won't catch every failure mode—a capacitor can fail with high ESR and still read close to its rated microfarads—but it's the absolute minimum, and it caught this one.
The replacement caps came from Eaton's certified spare parts list. I've been tempted by cheaper options before. A capacitor that tests okay when it's cold can still die under load at the exact moment the load bank is watching. Not worth the drama.
After replacing the caps, we load-tested both Eaton 93PM units. No transfers, no alarms. The UPS side was solid. But the backup generator protecting the same data center failed its run test. It started, ran for about ten seconds, then shut down.
The facility manager looked at me and said, 'We changed the oil filter yesterday. Maybe the oil filter bypass valve is stuck open.'
That theory was worth testing. The oil filter bypass valve is a pressure-operated bypass inside the filter or filter housing. If it sticks open, unfiltered oil can flow and, depending on the oil pressure sensor location, the engine may see a low-pressure condition and shut down. We removed the filter, applied regulated shop air, and watched the valve open near the pressure printed on the filter. It seated again when I bled the air off. The bypass valve was fine.
To be fair, the aftermarket filter itself wasn't bad. But it wasn't our problem.
Next suspect: fuel. The generator had fuel, and the electric fuel pump would prime when the controller called for it. But after the engine started, the pump lost power. That's a classic bad relay—the coil works when there's no load, but the contacts don't hold under current.
Let me show you how to bypass fuel pump relay the safe way. I have a scar from learning this the hard way. In 2023, I used a paperclip to jumper a relay socket. The paperclip carried more current than it should have, got hot, and melted the socket in about four seconds. That mistake cost roughly $900 and a very quiet ride home. Since then, I use only a fused jumper wire.
You don't bypass the relay to hide a problem; you do it to isolate the pump. Remove the relay. Identify the battery feed terminal (usually 30) and the load terminal to the pump (usually 87). Install a fused jumper between 30 and 87. If the pump runs continuously, the pump side is fine and the relay is suspect. If nothing happens, check the pump, its ground, and the wiring on the load side. And don't jumper 85 and 86—that's the coil circuit, and it won't tell you what you need to know.
In this case, the pump ran the moment we bypassed the fuel pump relay. We put in a new OEM relay, restarted the generator, and it carried the full load for the rest of the test. The oil filter bypass valve went back in and stayed there.
This job could have gone the way the previous one did: replace the batteries, call it fixed, wait for the next phone call. Instead, the actual fix was one capacitor and one relay. The wasted battery money was painful, but the real cost was bigger. The data center's tenants didn't see a 'capacitor problem.' They saw a UPS that transferred to bypass at random moments. In data centers, reliability is the brand. Maintenance quality is what people remember.
Now, before I approve any battery replacement on an Eaton online UPS, the checklist looks like this:
I'm not saying every problem is a capacitor. I am saying 'replace the batteries' was wrong until I proved it. Eaton UPS systems are generally reliable, but they're not magic. The Eaton 93PM UPS data center setup was solid. My diagnostic process wasn't. Check the capacitor first. It's a ten-minute test that can save a very expensive phone call.
Oh, and use a fused jumper. Always.