Your 2020 Backup Power Spec Is Probably Wrong — An Argument for Rethinking UPS and Generator Sizing

“We specced this in 2020, it should still be fine”

I hear some version of that sentence on almost every urgent call I take. Usually it's said by a facilities manager standing in a dark control room at 7 a.m., phone in one hand, coffee in the other, genuinely believing the problem is a dead battery.

Most of the time, it isn't. The hardware is doing exactly what it was designed to do. Nothing failed. The design just aged out.

My position: if your backup power plan was drawn up before 2022 and you haven't revisited it since, there's a real chance it's wrong — not because the physics changed, but because the equipment on both sides of the transfer switch did.

I coordinate emergency power deployments for industrial and commercial clients. Over the past nine years I've handled 200-plus rush orders, including same-day turnarounds for plants that were already dark and losing money by the hour. In my role triaging those calls, the pattern is almost boringly consistent: the failure isn't in the box. It's in the assumption behind the box.

When I first started doing this, I assumed the biggest number on the spec sheet was the safest number. That cost me — and one client — about six figures of downtime before I figured out I was solving the wrong problem.

Argument one: inverter generators quietly rewrote the “dirty power” assumption

For years the default logic went like this: generator output is messy, so put everything behind a double-conversion UPS and oversize the whole chain by 30%. That was sound advice when the unit outside was a conventional synchronous generator running at fixed RPM with whatever load you threw at it.

Then inverter generators got cheap enough to matter.

So what is inverter generator technology, exactly, and why does it cost more than a conventional unit of the same wattage? Short version: the engine spins an alternator, that raw AC gets rectified to DC, and an inverter rebuilds a clean sine wave at exactly 50 or 60 Hz regardless of engine speed. Because output frequency is decoupled from RPM, the engine can idle down when load drops. Lower total harmonic distortion, tighter voltage regulation, less fuel burn, and a noise profile you can actually hold a conversation over.

Practical consequence: a 10000 inverter generator — a 10 kW-class unit — can now carry a small server room, a bank of PLC panels, and the emergency lighting circuit without the voltage sag that used to force people into oversized UPS chains. THD on decent units typically lands under 5%, which is well inside what most line-interactive UPS systems will accept without complaint.

But here's the catch nobody mentions in the brochure. Inverter generators have surge behavior that doesn't look like conventional units. Motor starting — compressors, pumps, anything with inrush — can trip a unit that looks generously sized on paper. That's not a UPS problem. That's a load-study problem, and it's why I still ask for a startup sequence before I'll sign off on a size.

I didn't fully understand that until a mid-size bottling plant called me on a Tuesday morning in February 2024, roughly 30 hours before a production run they couldn't move. Their new inverter generator was fine. Their sequencing wasn't. Two compressor starts overlapped by 400 milliseconds and the whole thing went down. We fixed it with a startup interlock and a bigger battery bank on the UPS — not a bigger generator.

Argument two: right-sizing beats up-sizing, and the product range finally supports it

Here's the counterintuitive part. The correct response to most of these calls is smaller, not bigger.

Oversizing a UPS isn't free. You pay for it in battery replacement cycles, in cooling load, in idle losses, and in the false confidence of a runtime number that was calculated at 40% load and will never be true. As of early 2025, replacement battery cartridges for 1500–3000 VA class units generally run in the low-to-mid hundreds of dollars per unit depending on chemistry and supplier — verify current pricing with your distributor, because that number moves with lead and lithium markets.

The APC-UPS range splits cleanly enough that you can stop guessing. The APC Back-UPS 1600VA is the unit I reach for when the load is a network closet, a handful of POS terminals, or a small edge rack — line-interactive, transfer typically in the 8–10 ms range, and enough headroom that you're not running at 90% utilization on day one. The APC UPS 1350VA covers the tier below it: a workstation plus router, a branch office, a security head-end. And when the load genuinely matters — a server room, a control room, anything with a PLC panel sitting behind it — you're into Smart-UPS territory, where double conversion and external battery packs let you buy runtime instead of buying VA you'll never use.

If you want the formal language: IEC 62040-3 classifies UPS topologies as VFI (online, double conversion), VI (line-interactive), and VFD (passive standby). Most people overbuy VFI when VI would do, then underbuy runtime on the one load that needed VFI all along.

Why does this matter? Because during an outage, the question isn't “how many VA do I own.” It's “how many minutes do I have to land this process safely.” Those are different questions, and only one of them shows up on a quote.

Argument three: PLC panels are the load nobody budgets for

This is the one that keeps me up.

Walk into almost any plant and you'll find the IT rack on a UPS, the phone system on a UPS, and the PLC panels on raw utility power — because “the panel has its own power supply.” Technically true. Practically meaningless.

A PLC panel doesn't need to lose power for minutes to cause damage. A sag of a few cycles can drop the power supply below threshold, reset I/O modules, and send every valve and actuator to its fail-safe position mid-batch. That's a lost batch, a lost recipe, sometimes an unplanned mechanical stress event. And from a code standpoint, most PLC panels sit on an optional standby branch under NEC Article 702 — not Article 700 emergency or 701 legally required standby. So nobody designs ride-through for them. They're compliant. They're also unprotected.

Consequence anchor for you: a client of mine lost a $180,000 production run in 2023 because a 300-millisecond dip reset a PLC panel that had been “fine for eleven years.” Eleven years of not being tested is not the same as eleven years of being adequate.

And here's the communication failure I see constantly. I say “clean power.” The maintenance manager hears “surge protector.” Different words, same sentence, completely different systems. We discovered that gap the hard way when an installation came back with a $40 strip installed where a line-interactive UPS was supposed to go. Nobody lied. Nobody misunderstood their own job. The vocabulary just didn't line up.

The pushback I always get

Fair challenge: isn't this just an argument for buying bigger, more expensive systems? I get why that's the read. Vendors do have an incentive to sell you capacity.

But look at what I actually recommended above. Smaller line-interactive units on the loads that don't need double conversion. A load study instead of a bigger generator. A startup interlock instead of a second machine. Most of my recommendations save money up front and cost money only in engineering time.

Second pushback: “Our electrician signed off on it.” Granted — and that's not a criticism of your electrician. Code compliance and process ride-through are two different standards, asking two different questions. NEC Article 700 asks whether the egress lighting stays on. IEEE 446, the Orange Book, asks the harder question about what your operation actually needs to survive. Those documents don't argue with each other. They just aren't about the same thing.

Third pushback, and the honest one: “We've never had a problem.” I hear that a lot. Almost never from someone who's had an outage longer than four hours. Reliability that's never been tested is usually a scheduling coincidence, not a design outcome.

What I'd actually change before next quarter

• Map your loads by consequence, not by location. Which circuits, if they drop for 500 ms, cost you money? Those get protected first. Everything else can wait.
• Test the transfer, not the battery. A UPS that holds up the IT rack flawlessly and fails to coordinate with the generator is a UPS that bought you 12 minutes and a false sense of security.
• Put a UPS in front of every PLC panel that runs a batch process. Even a modest line-interactive unit will bridge the gap between “dip” and “catastrophe.”
• Re-read the 2020 spec with 2025 equipment in mind. The inverter generator you'd buy today is not the generator you priced five years ago, and the UPS that pairs with it isn't either.

The fundamentals haven't changed. Power still has to be available, clean, and long enough to land the plane. What changed is the execution — and the gap between a spec written in 2020 and the hardware available in 2025 is wide enough that “it's still fine” has quietly become an expensive assumption.

Take it from someone who's spent nine years getting called after the assumption fails.