The phone rang at 2:22 on a Thursday afternoon in June 2024. The plant manager didn't say hello. He said the UPS was screaming and told me to get over there.
I'm a procurement manager at a 170-person metal fabricating company outside Cleveland, not an electrical engineer, so I can't explain the exact capacitor failure that ended the life of our old 150 kVA UPS. What I can tell you, from the budget side of the table, is that the next eleven minutes cost us about $18,000 in idle CNC machines, a missed shipping window, and the overnight freight we paid to make the customer whole.
Part of that number is on me. The old unit went in during a 2012 expansion, and the batteries were replaced once in 2017. In early 2023, our maintenance supervisor asked for another battery replacement. I deferred it to the next fiscal year to protect our spend. I'll spare you the spreadsheet logic. It looked responsible on paper, and it was not responsible in reality.
That June afternoon changed how I think about power protection. I used to treat a UPS like a fire extinguisher: buy it, mount it, forget it. It's not.
After the outage, I asked our electrical engineer to write a functional spec, and we sent it to three vendors. My instruction to each rep was simple: do not start with a price. Start with efficiency data, service reach, and the start-up procedure. We were replacing a 150 kVA three-phase unit, so high capacity Eaton UPS systems were in the conversation from the first round of proposals.
One of the quotes came from an Eaton channel partner. It covered a 150 kVA Eaton 93PM, a matching battery cabinet, remote monitoring, and commissioning. The other two proposals were from legitimate manufacturers, and I'm not going to tell you they were inferior—that kind of blanket claim is usually a salesperson's tell, not an engineer's answer. What I can say is that the Eaton proposal was the only one where the efficiency data matched the specific model we were quoted, not a general brochure.
I built a TCO model because that's what I do. Our average load measured about 90 kW. The old UPS was running at roughly 91 percent efficiency, and the 93PM's published data said 96.5 percent in double-conversion mode. The math: 90 divided by 0.91 is about 98.9 kW of input, and 90 divided by 0.965 is about 93.3 kW. That's a 5.6 kW difference, which at 8,760 hours per year comes to roughly 49,000 kWh. Our blended industrial electric rate as of July 2024 was $0.115 per kWh, so the efficiency gap alone was worth about $5,700 per year. Over a ten-year lifespan, that is not a rounding error.
The scope also included a few smaller loads. For two remote network racks near the loading docks, we ordered a pair of Eaton Tripp Lite UPS units instead of resurrecting old gear from storage. The Eaton Tripp Lite UPS line made sense at that size: lower upfront cost, straightforward warranty, and the same Eaton support channel as the big machine. I wouldn't use one as the backbone of a server room, but for a couple of edge switch stacks, they were the right call.
Here is where I made my first bad assumption. We have a 500 kW diesel generator on site, and I assumed we were set once the UPS and the generator were both new. I did not verify the connection between them.
The old building used a manual transfer switch, which is exactly as useful as it sounds. Somebody has to walk to the electrical room during an outage and transfer the load by hand—in the dark, under stress, while production is yelling. Our engineer's design included an automatic switch, and the line item was for a Russelectric transfer switch. I had never heard of the brand, and my first instinct was to mark it optional.
It was not optional. An automatic transfer switch is the bridge between the generator and the UPS. It watches both power sources and transfers without waiting for a human. More important for a maintenance budget: the Russelectric transfer switch included a bypass, meaning the UPS can be isolated for service without shutting down the whole building. I assumed we had a generator, so we were covered. That assumption was wrong.
The switch added cost and, more painfully, lead time. The UPS arrived in eight weeks; the Russelectric transfer switch took almost fourteen. That alone tells you how important it is to get the switch into the budget early.
Start-up was scheduled for September 2024. The Eaton field engineer arrived on a Tuesday morning, walked into the UPS room, and asked a question that stopped me cold: when was the last time the cooling unit's filter was changed?
I checked our maintenance log. The log literally said Honeywell air filter replacement—due April 2024. It was September. The filter in the unit wasn't visibly filthy, but it was restrictive enough that the cooling unit was cycling longer than it should. The engineer explained that battery life depends on room temperature, and most UPS battery warranties expect the room to stay between 20°C and 25°C. We were sitting at about 27°C under load. A $40 filter had been making a $100,000 battery decision look risky.
We replaced both cooling unit filters that afternoon. A Honeywell air filter replacement is not usually the part of a UPS project you remember, but it is now in my preventive maintenance calendar.
The same week, the warehouse manager forwarded an email with this subject line: why is my lithium battery charger blinking red. The photo showed the charger for our lithium-ion pallet jack battery, not the UPS. At first I assumed we would need to replace the charger.
We did not. On most modern lithium chargers, a slowly blinking red light means the battery management system is balancing cells after charging. It can be part of the normal charging cycle, not an error code. The light turned solid green about forty minutes later, and the charger has been fine since. The lesson: read the LED code table before replacing a $900 piece of equipment. A fast blinking red light that never goes green is another story, but a slow blink during charging is usually just the battery asking for a little more time.
Four months have passed since commissioning, and I've reviewed the sub-meter readings a few times. The energy savings are in the range I modeled, plus or minus the usual seasonal noise. I do not have exact numbers in front of me, but the pattern is clear.
If you are about to buy a large UPS, do not compare purchase prices. Compare the installed, commissioned, and maintainable system. Ask what the startup engineer will check, not just what the sales rep will sell. Ask where the transfer switch sits in the schedule. Ask what the room temperature will look like when the load is at 90 percent.
In my opinion, customer education is part of buying infrastructure. An informed customer asks better questions and makes faster decisions. I'd rather have a vendor spend forty minutes walking me through battery chemistries and transfer switches than have to explain to my CFO, six months later, why the cheap option did not include a bypass.
And if you see a blinking red light on a lithium charger, check the manual before you panic. Sometimes a red light is not a warning. Sometimes it's just the product telling you it's working.