Eaton UPS Sizing, Transfer Switches, and Breaker Testing: Three Scenarios, Three Different Answers

The Short Version: There Are Three Different Power Problems

I coordinate emergency power deployments. I have handled 200+ rush orders in nine years, including same-day turnarounds for data center and light industrial clients. The calls come in roughly the same shape every time, and they usually start with four words: we need a UPS.

Nine times out of ten, that is not the actual problem.

Sometimes the person on the phone needs to size a new unit. Sometimes they need a switching path — a transfer switch, a charger, a start battery that will actually turn over at 3 a.m. And sometimes the gear is already installed and nobody can prove it works, which is a completely different job with a completely different set of tools.

Those three jobs get answered with the same generic advice all over the internet, and that advice is wrong for at least two of them. So here is the split I use when I am triaging.

What was best practice in 2019 is not automatically right in 2025. NFPA 70B changed status, lithium moved into the rack, and the sizing calculators got genuinely good. The fundamentals did not move. The execution did.

Scenario 1 — You Are Sizing a New UPS

This is the cleanest scenario, and the one people overthink the most. Rack UPS selection is not that complicated once you have one number: the measured load, in watts.

Not the nameplate load. Not the number the server admin remembers from the purchase order. The measured load, taken at the panel or from the old UPS display during a busy period.

The Calculator Is a Starting Point, Not an Answer

Eaton's UPS calculator is a decent tool. You put in your load, your runtime target, your voltage, your redundancy preference, and it hands you candidate models and an external battery module count. It saves you twenty minutes of spec-sheet cross-referencing.

What it cannot know is your growth plan. If you feed it today's 6 kW load and ask for 10 minutes of runtime, it will size something that gives you 10 minutes forever. Add one more row of servers next spring and your runtime quietly halves. No alarm goes off. You find out during the next outage.

My rule: size the runtime for the load you expect in eighteen months, not the load you can measure today.

Rack-Mount vs. Modular

Rack-mount is the right answer below roughly 20 kVA — you buy the box, you mount it in the rack, you plug in the battery modules, done. Modular makes sense when you are past that point and want to grow without swapping the whole unit out. The 93PM class of gear exists for exactly that reason.

The temptation is to buy modular because it sounds more serious. Do not. Modular buys you scalability and serviceability and it costs you floor space, heat, and a bigger commissioning effort. If your load is not going to grow past the chassis you can afford, rack-mount is not a compromise. It is the correct call.

Where the 9PX Fits

The Eaton 9PX is a double-conversion online unit in the 5 to 11 kVA range, unity power factor, rack or tower convertible. In practice that means watts and volt-amps track together on the UPS side, which removes one whole category of sizing confusion that older VA-rated gear created.

Two things I would flag if you ask me:

• Efficiency numbers carry asterisks. Eaton rates the 9PX in the mid-90s for double conversion, higher in energy-saver mode. That number moves with load percentage and battery charge state. Check the spec sheet for your exact model and load.
• The internal charger is sized for the internal battery plus a handful of external modules. Stack too many battery modules on and your recharge time stretches well past the window you designed for. If your recovery time matters more than your runtime, look at the charger spec before you look at the runtime table.

If you only do one thing here: put a clamp meter on the actual feed during peak hours. I have never once seen a client overestimate their load. Every single time, it is lower than they feared and higher than they documented.

Scenario 2 — You Are Building or Replacing the Transfer Path

This is where the expensive mistakes live, because the transfer path is the part of the system nobody thinks about until it fails.

Manual Transfer Switches: Cheaper, and Usually the Wrong Tool

The advice you will read is that a manual transfer switch is simpler and cheaper than automatic, and that simplicity is a virtue. That is half true.

It is tempting to think the manual-versus-automatic decision is about budget. But the real variable is who is physically on site when the power drops. A Zenith manual transfer switch (the brand has changed hands over the years — verify whose logo is on the current catalog before you spec one) with a proper kirk-key interlock is a perfectly legitimate solution for a shop with a maintenance tech on every shift who has been trained on it and practices twice a year.

Put that same manual switch on a colo cage three time zones from your nearest engineer and you have built a very reliable mechanism that requires a human being who is not there.

And the cost gap is narrower than people assume once you factor in the annual testing, the training, and the interlock hardware. Automatic transfer switches are not cheap. Neither is an outage that nobody could respond to.

The Onboard Charger Is the Part Everyone Forgets

Whether the charger is built into the transfer switch or the UPS, it has one job: put back what the battery gave up, fast enough that the next event finds a full battery.

Two failure modes I see constantly. First, the charger is sized as a trickle maintainer and the battery is a deep-cycle start bank, so a long outage leaves the generator with a start battery that will not turn it over. Rough rule of thumb, take it with a grain of salt: a charger should be able to deliver somewhere around a tenth of the battery's amp-hour capacity. Undersized chargers work fine until they are asked to do the one thing they exist for.

Second — and this one stings — the charger dies quietly. There is no alarm on most of them. The battery just slowly stops being charged and everything looks normal until it is not.

Never expected the onboard charger to be the failure point on a critical site. Turns out a two-hundred-dollar charger took out a forty-thousand-dollar genset start sequence. We now put charger output on the quarterly inspection sheet, right next to battery voltage.

Scenario 3 — The Gear Is Already There and You Have to Prove It Works

This is the scenario nobody writes about, and it is the one I get asked about most: how to test a circuit breaker without power.

Start with the honest answer. You cannot fully test a circuit breaker with the panel de-energized. What you are testing without power is the mechanical linkage and the conductor path. What you are not testing is whether it will trip at the right current in the right time. Those are different questions.

What You Can Actually Test With the Breaker De-Energized

Before any of this: lock out and tag out per OSHA 29 CFR 1910.147, and verify absence of voltage with a meter rated for the category you are standing in front of. Test the meter on a known live source first. A meter that reads zero because the meter is broken has killed people.

• Visual inspection. Scorching, arcing residue, discolored lugs, a breaker that looks like it got hot. It takes two minutes and catches more than any instrument.
• Manual operation. Cycle it. It should feel firm and positive, not mushy. A breaker that will not reset cleanly is telling you something.
• Low-resistance ohmmeter check. A micro-ohmmeter across the poles shows you contact resistance. High resistance equals heat equals nuisance trips and eventually a fire.
• The test button on GFCI and AFCI breakers. Useful, but understand what it does. It exercises the sensing electronics. It does not verify the trip mechanism or the trip curve.

Ten years ago NFPA 70B was a recommended practice. As of the 2023 edition it is a standard, which mostly means the maintenance intervals now have to exist on paper. That changed the paperwork. It did not change the physics.

What You Cannot Test Without Power

The trip curve. The instantaneous element. Whether it trips at 6x rated current in under two cycles like the coordination study says it should.

That requires primary injection — a high-current test set pushed through the breaker with the load side isolated. For a handful of critical breakers, renting a set for a day is money well spent. I am not going to guess at current rental rates; they move and they vary by region, so call two suppliers.

The Two Checks That Catch Most Real Failures

If you cannot do injection testing on everything, do these two things instead.

First, torque the terminations to the label value. Loose lugs cause more failures than worn trip mechanisms, and nobody checks them because it is boring. Second, thermal scan the panel under real load. A breaker running twenty degrees hotter than its neighbors is a breaker with a story, and the camera finds it in seconds.

Both of those require the panel energized, which is why they get skipped. Get someone qualified, in the right PPE, to do them anyway.

A lesson learned the hard way: we did not have a formal pre-outage checklist for transfer tests. Cost us when a tech ran a manual transfer on a Friday afternoon and left the switch in the wrong mode over the weekend. Nothing failed. That is the worst kind of near-miss, because nobody learns anything from it. We have a checklist now.

How to Tell Which Scenario You Are In

Three questions, in order.

One: does the gear already exist? If there is a UPS or a transfer switch on the wall, you are in Scenario 3 regardless of what you think you need. Sizing a replacement before you have measured the existing load is how people buy the wrong box twice.

Two: do you have a measured number? If you cannot produce a watts figure taken during peak hours, you are not ready for Scenario 1. Everything after that is guesswork wearing a spreadsheet.

Three: who is on site at 3 a.m.? If the honest answer is nobody, then a manual transfer path is off the table no matter how much cheaper it looks on the quote.

Most people who call me believing they are in Scenario 1 are actually in Scenario 3. The unit they want to buy solves a problem they have not yet confirmed they have.

The One Thing All Three Have in Common

Every one of these failures is a documentation failure before it is an equipment failure. The load was never measured. The charger was never logged. The breaker was never scanned. The transfer switch mode was never written down.

If you only do one thing this quarter, do not buy anything. Measure the load, scan the panels under load, and write down the model number of every charger on site. That list is worth more than the upgrade you were about to spec.