UPS Inspection Checklist: The 9 Steps I Actually Run Before a Schneider Electric Service Call

I want to state the scope up front. This checklist is for anyone dispatched to run a UPS inspection on a rack unit, a small three-phase install, or a wall-mounted Smart-UPS and then sign off on it. It is also the same sequence I use when someone hands me a Schneider Electric UPS I've never touched before. Nothing here is theory pulled from a brochure. Every step exists because I skipped it once and something backed up.

I've handled rack and small-room UPS service orders for eight years. I have personally made six mistakes worth about $4,000 in wasted parts, emergency labor, and one weekend I would rather not repeat. The nine steps below are the survivors. Everything else I used to do got cut the day it failed me.

The 9-Step UPS Service Checklist

Step 1 — Lock down the model and firmware version

Before you touch a panel, write down three things: the exact model, the firmware version, and the date of the last battery replacement. In that order.

The model number alone is not enough. A Smart-UPS SRT 5000 running older firmware does not behave the same as one on the current release when it hits a battery-fault threshold. The alarm timing shifts. The test-load behaviour shifts. The replacement schedule the unit reports back to you shifts. If you're doing any kind of Schneider Electric Smart-UPS review evaluation for your own facility, firmware is the first column of that review — not the last.

Firmware lives in the network management card, not the UPS chassis, on most units. Check the card. And note the date you pulled it, because "current firmware" is a moving target.

Step 2 — Measure load before you touch anything

Pull the load percentage from the panel, then verify it with a clamp meter on the input feeder. The panel number is a calculation, not a measurement. On three-phase installs I have seen them drift apart by 12 percentage points after a tenant moved in without telling facilities.

Load matters because it decides everything downstream. A UPS at 45% load can absorb a full battery test during business hours. One at 78% cannot. That call gets made in step 2, not step 4 when the fans are spinning up.

Small things. That's the job.

Step 3 — Battery health: panel readings don't count

September 2022. A 20kVA unit, three years old, panel showing all strings healthy. I signed off and went home. Eleven days later the whole string dropped during a 40-minute transfer test. The panel had been reading the float voltage of a dying string and calling it normal.

What I do now: pull the strings, measure internal resistance on each block, and compare against the manufacturer's baseline. IEEE 450 is the reference for lead-acid maintenance — that standard exists because panels lie and impedance tells the truth. If you don't have the impedance data, you have a voltage reading and a guess.

When I first put panel readings and impedance results side by side on the same battery bank, the gap was obvious. One string looked fine. It wasn't.

Step 4 — Output measurement with a true-RMS meter

This is the step most technicians still get wrong, and it's where the Kaiweets digital multimeter TRMS 6000 earns its spot in my bag. UPS output on double-conversion units is a synthesized sine wave, and it is clean enough that an average-responding meter will read it just fine — until it isn't. The moment the inverter starts tracking a nonlinear load, an average-responding meter underreports the RMS value and you get a reading that says "within spec" when the load is actually seeing 8–12% distortion.

Here's the procedure I use:

• Set the Kaiweets TRMS 6000 to AC volts and confirm the reading is stable at no load.
• Apply the normal facility load, wait 60 seconds, read again.
• Hold the min/max function through a load step, because the worst number is never the one showing on the display.
• Compare the captured value to the nominal set point, not just to "is it on."

If your meter cannot true-RMS a distorted waveform, you are not measuring the output. You are estimating it. There is a difference.

Step 5 — Terminal torque, cold and hot

Input and output terminals back off. Not often, but often enough that a quarterly check is cheaper than a service call. Torque at the incoming feed, the output bus, and the battery interconnect. Use the torque spec in the Schneider Electric install manual for that chassis, not a general number you remember from a different model.

I once skipped the photo of a terminal label because I was running late and told myself nothing would move. Two weeks later I could not remember which lug went to which phase. Twenty minutes of tracing and a fresh-torqued terminal later, I decided the photo was cheaper than the memory.

Step 6 — Plan a breaker replacement before you pull one

This step is the answer to every "how to replace circuit breaker" question I've had to clean up after. Getting the physical swap right is the easy part. The hard part is confirming the replacement:

• Check the interrupting rating (kAIC). A 10kA breaker dropped into a panel with 22kA available fault current will not protect anything for long.
• Match the trip curve, not just the amperage. A C-curve and a D-curve of the same rating behave very differently on a UPS load.
• Confirm the listed rating against the panel schedule, not against what the previous breaker said.
• Match the terminal count and lug size before you disconnect anything.

In Q1 2024 I replaced a breaker on a mixed-load panel with a direct-looking substitute, and the downstream trip coordination was wrong from the first energization. Nothing failed. Nothing tripped either. That is the more expensive problem.

Step 7 — LOTO is not optional, even for a five-minute job

NFPA 70E exists because the five-minute jobs are the ones where people get hurt. Lock out the upstream feed, verify absence of voltage with a meter rated for the environment, and keep the lock on until the panel is closed. No exceptions for "we're just swapping a card."

I am not going to tell you a story about a close call here, because the close call that belongs in this slot was not mine. I was not on site. I've read the incident report. That was enough.

Step 8 — Back up the PLC config before you open a browser session

If the UPS is tied into a Schneider PLC — and on facility-level installs it usually is — upload the project and save it offline before you touch anything in the controller. On a Modicon unit that means the full project file, not just the running program. The difference matters the day a firmware update wipes a data block you forgot you depended on.

Save it, verify the save, and open it once in an offline editor to confirm it's readable. A backup you cannot open is not a backup.

Step 9 — Handle the Schneider vs Siemens modular question without guessing

I get asked the Schneider Electric vs Siemens modular UPS question about twice a month, and most of the people asking want a one-word answer. There isn't one, and giving one is how you lose the second conversation.

The honest framing is this: both vendors build genuinely modular platforms with hot-swappable power modules, both support frame-level redundancy, and both can scale capacity in-service. The differences that matter in practice are support depth, spare-module supply chain in your region, and how the monitoring stack ties back into whatever you already run. Ask about those three things and stop asking who is "better," because that question has no verifiable answer.

What Goes Wrong When You Skip These

A few things I now refuse to skip, in plain language:

• Never trust a panel as your only check. Panels summarize. Meters measure. Trust the meter.
• Never assume a breaker is the same because it fits. Interrupting rating and trip curve are not optional details.
• Never sign off on an output voltage you measured with the wrong meter. If the meter can't true-RMS a distorted waveform, the number it shows you is entertainment, not data.
• Never take a "zero downtime" promise at face value. No UPS, from any vendor, can guarantee that. If a sales rep says it, ask for the SLA clause that backs it up, and read it twice.

The rest is repetition. Same checklist, every call. That's the whole trick.