Free product samples available for qualified projects — Request yours today
Home/Blog/Article

How to Test a Three-Phase 100kVA Online UPS: What Your Checklist Won't Tell You

The Question Everyone Asks (That Isn't the Real Question)

If you're searching for how to test a three-phase 100kVA online UPS, you're probably hoping to find a checklist you can print and walk through—battery discharge, efficiency verification, transfer time, voltage stability. I get it. In 2019, when I handled my first 100kVA UPS purchase for our factory, I printed a 12-page acceptance testing checklist and figured I'd just tick the boxes.

The only thing that test actually caught was that the UPS model we received didn't match the purchase order. The nameplate was missing a suffix letter—the vendor called it a "batch variation." I didn't find out until I re-read the contract that the letter indicated a different output isolation configuration.

Here's what that taught me: the checklist was never the real problem. The real problem was everything around it—the conditions, the money, and the definitions nobody put in writing.

The Deeper Problem Nobody Talks About

1. The test report was never about your site

The factory acceptance test (FAT) that comes with your UPS is performed under ideal conditions: 25°C ambient temperature, purely resistive load, stable input voltage. Your site is probably not that. Our factory's electrical room hits 30°C+ in summer, the load includes switching power supplies and variable frequency drives, and input voltage sags by up to 8% on some afternoons.

IEC 62040-3—the international standard for UPS performance and test methods, current as of 2024—classifies systems by output performance (VFI, VI, VFD) and outlines what to measure. But here's what it doesn't do: it leaves the acceptance criteria and site conditions entirely up to the buyer. The standard tells you what to test, not what counts as passing on your site.

By our second purchase in 2022, I'd added a line to the contract appendix: "SAT shall be conducted at ambient temperature ≥ 28°C, with a mixed nonlinear load, and input voltage variation ≤ ±5%. Any parameter deviating more than 5% from the factory report shall be considered non-conforming." That single paragraph ended more arguments than the rest of the contract combined.

2. Cost ambiguity causes more pain than technical problems

I've seen this movie: the quote shows a clean equipment price, then buried in the contract appendix you find that "on-site acceptance testing" is billed separately—load bank rental, engineer travel days, and re-testing fees if the first round fails. I've learned to ask what's not included before I ask what the total is. The vendor who lists everything upfront—even when the number looks higher—usually costs less in the end.

On our second purchase, one supplier's quote came in about 8% higher than another. Finance asked why I was leaning toward the expensive one. I printed both quotes with their appendices side by side. The cheaper option didn't include on-site testing labor (billed at a per-day rate) and said nothing about who pays for re-testing if the unit fails. The higher quote spelled out three days of SAT labor, load bank usage, and—this is the line that mattered—"if the acceptance test fails, the supplier bears all rework and re-testing costs." Same equipment. Completely different risk profile.

3. The definition of your acceptance criteria is a silent negotiation

Here's a specific example worth walking through. For an online double-conversion UPS, the output transfer time should theoretically be 0ms—the load rides the inverter at all times, so there's no break. But if you measure at the load terminals with a power quality analyzer, you might see a voltage dip of 2 to 4ms. That doesn't mean the UPS is broken. It means you're measuring the load-side response rather than the internal switching event. The inverter's IGBT switching happens in microseconds. What you're seeing at the terminals may come from load capacitance or the grounding of your measurement setup.

So when the contract says "transfer time ≤ X ms," which measurement point governs? If you don't define the test method and measurement location upfront, acceptance testing turns into a debate. We eventually added contract language: "All transfer-related specifications shall be measured at the UPS output terminals (load side of the output breaker). Test instruments must be calibrated with valid certificates." It's not glamorous. But it's the difference between a signed-off test and a three-week email chain.

What It Costs When You Get It Wrong

Let me give you three real numbers from three separate decisions—all of them preventable.

The $900 shortcut that cost $2,400

In 2021, we commissioned a new 100kVA double-conversion unit. The budget manager wanted to skip the load bank rental—about $900 for three days—and use our existing factory load for testing. "We've got plenty of equipment drawing power," he said. "Why rent a dummy load?"

Because existing load doesn't tell you what you need to know. Our factory load had a poor power factor (lots of lightly loaded motors) and the demand varied throughout the day—we couldn't hold it steady long enough for a proper efficiency measurement. We ran the test, got a number that looked 3% lower than the factory report, and couldn't tell whether the UPS was underperforming or the test was flawed. We ended up renting the load bank anyway, re-running the test (three more days of scheduling and overtime for the electrical team), and burning about $2,400 total instead of the original $900. Saved $900 by skipping the load bank. Ended up spending $2,400 to do it right the second time.

The contract phrase we both misread

That same project had a communication problem. The contract said the supplier would "provide on-site test support." We read that as they would run the acceptance test. They read it as they would stand there while we ran it.

The morning of the test, one engineer showed up and asked where our load bank was and who was operating it. We didn't have one—we assumed they were bringing it. He pulled out the contract. "Support" meant phone consultation and witnessing, not execution. We'd both signed the same document with completely different mental models of what that day would look like.

Now I insist on writing out roles before any equipment ships: who supplies the load bank, who operates the test instruments, who signs off on each measurement, and who pays for overtime if the test runs past 5 PM. Take it from someone who ate that particular deficiency report.

The trade-off I made under time pressure

During our 2022 purchase, the utility gave us a three-week window to complete interconnection. Missing it meant a two-month delay and running critical loads on generator power through the summer.

Normally I'd push for a full test protocol: full-load discharge for the rated duration, efficiency curve across load levels, transfer tests, and harmonic analysis—a three-day minimum. With the deadline, I cut it to two days. I dropped the partial-load efficiency sweep and kept the full-load discharge and transfer testing.

The upside was hitting the deadline. The risk was missing a partial-load inefficiency that would cost us month after month. Worst case: the UPS ran 2% below expected efficiency at 50% load—roughly $800/year in wasted electricity. Best case: we confirmed the unit met spec. The expected value said take the shortened test. The downside felt manageable. We went with it.

In hindsight, I should have pushed back on the utility timeline. But with the VP asking for daily updates and a real risk of generator-only operations through August, I did the best I could with the information I had. That's most procurement decisions, honestly.

A Workable Framework (Short, Because the Problem Is Already Clear)

After two 100kVA purchases and one painful re-test, here's what I've settled on:

Before the unit ships, lock the test protocol into the contract appendix:

  • Site conditions: ambient temperature, load type, input voltage window—match your actual site, not the factory lab
  • Measurement points: specify where each parameter is measured (UPS input terminals, output terminals, or load side)
  • Pass/fail criteria: tie acceptance to a percentage deviation from the factory report, not an absolute number that sounds reasonable
  • Cost allocation: who pays for the load bank, who supplies instruments, who covers re-testing

During the test, the core measurements for a 100kVA three-phase online double-conversion UPS:

  1. Full-load discharge: run the battery at rated load until end-of-discharge voltage. Record runtime and compare to spec. A 5% shortfall is worth investigating; 10% is a problem.
  2. Efficiency verification: input vs. output power at 25%, 50%, 75%, and 100% load. Compare against the factory report and the IEC 62040 efficiency class.
  3. Transfer test: switch between mains and battery while watching the output waveform. For double-conversion, load-side voltage should stay within the specified window throughout.
  4. Voltage and frequency regulation: hold steady load, vary input voltage across the specified range, watch the output.
  5. Battery impedance check: measure internal resistance per block. A block reading 20% above the string average is a red flag—IEEE 1188 (the VRLA battery maintenance standard) recommends tracking these trends over time, not just at commissioning.

After the test, get every measurement signed off by both parties with timestamps. Keep the raw data. If there's a dispute six months from now, the signed test record is the only thing that matters.

One last note on vendor selection. We evaluated four manufacturers for the second unit, including a PowerSentinel 100kVA double-conversion model. What put them on the shortlist wasn't the brand name or the price—it was that their datasheet listed test conditions next to every specification without us asking: ambient temperature, load type, input voltage range. And when we asked about their industrial power conditioner line for a separate production line project, the same documentation standard held up. We'd dealt with a switching power supply manufacturer before who couldn't tell us the test conditions for their own efficiency numbers—that experience made us value the transparency more than any single spec sheet.

The vendor who lists the test conditions upfront—even when the unit costs more—is usually the one you want. I've learned that the hard way, twice.

author avatar
Pavel Novak

Pavel Novak is a generator and backup-power analyst specializing in standby, portable, diesel, gas, inverter, and commercial generating sets with transfer equipment. For reciprocating-engine sets, he applies ISO 8528-5 dynamic-performance criteria and IEC 60034-1 generator ratings to load steps, starting current, voltage and frequency recovery, cooling, and output; for transfer equipment, he uses IEC 60947-6-1 requirements. He helps facility teams and buyers select systems around emergency loads, runtime, commissioning, fuel strategy, service access, operating cost, and resilience.

Leave a Reply

Your email address will not be published. Required fields are marked *

Please write your comment.