The setup: a generator, a power grid, and a blind spot

In November 2019, I signed a $2,400 contract for home generator services in Petaluma, CA. Our small transcription company was getting hammered by PG&E's rolling blackouts—two shutdowns a week, minimum. The generator was supposed to be the answer.

Six weeks later, it was installed and humming. Problem solved, right? Not even close.

Here's what I didn't know: residential-grade generators take about 15 seconds to kick in during an outage, and even when they do, there's a brief voltage sag during the transfer. That's harmless for a refrigerator. For four transcription workstations and a recording array, it's a death sentence.

That's when I learned I needed a UPS—uninterruptible power supply—to bridge the gap. What followed was a $340 mistake I've documented so you don't repeat it.

Mistake #1: "A UPS is a UPS"

I grabbed a $58 standby UPS from a big-box store. My thinking at the time: battery is a battery, right?

Wrong. Standby units have a transfer time of 6–10 milliseconds and usually output a simulated modified sine wave. Our server PSUs didn't care about the "power good" LED on the front panel—they cared about what the waveform actually looked like when the grid dropped.

In June, a half-second voltage sag hit our building. The $58 unit tripped. We lost a PSU mid-afternoon and a chunk of client work with it. Data recovery alone cost $340. Client deadline? Missed. My credibility? Bruised.

I taped a note to my monitor: "Buy the right UPS."

Mistake #2: I did the research—and still got it wrong

After my self-flagellation phase, I actually started reading. I read maybe fifteen reviews that included ups cyberpower ut1000g 1000va/500w reviews in the mix—the UT1000G is genuinely well-reviewed for its price point. And honestly, it's a solid standby unit.

But when I ran the numbers for our actual setup, the UT1000G couldn't sustain a workstation plus a recording array back-to-back. I'd need two of them, which meant either a daisy-chain headache or a single larger unit.

This is where the math got uncomfortable. I'd been comparing sticker prices, not total cost of ownership. Once I actually added up the cost of my "cheap" approach—including the $58 unit, the $340 data recovery, the replacement PSU at $90, and about eight hours of lost labor at roughly $80/hour—the total came to about $1,186.

Compare that to a single cyberpower online ups (dual-conversion, zero-transfer-time) at roughly $680, plus battery replacements every 3.5 years at about $120 each. Over five years, that's roughly $920. Cheaper than the "bargain" route by a wide margin.

To be fair, not everyone needs an online UPS. If you're running a single desktop and a router, a decent standby unit is usually fine. Our situation was different—we had sensitive audio equipment running through multiple PSUs, and the cost of a failure was high enough that the TCO math flipped fast.

Mistake #3: I never checked compatibility

A friend who does mechanic work called me during that same month. He'd ordered a 97 Chevy 1500 fuel pump for a customer's truck—except the customer's truck was actually a '96. The pump wouldn't fit. The order was non-returnable. That was $180 down the drain, plus two days of an angry customer.

"I should've checked the part number before ordering," he said. That was my exact problem, just at the UPS level. Our server PSUs had a specific sine-wave requirement, and I hadn't checked the specs closely enough before buying. Compatibility isn't glamorous. It's just cheaper than mistakes.

What I learned about testing

My nephew recently searched "how to test spark plug with multimeter" because his truck was idling rough. He assumed you just touch the probes and look for a reading. Turns out you need the right setting (resistance/ohms), the right probe points, and a comparison to the manufacturer spec. Without that, you're guessing.

I was doing the same thing with my UPS. I'd push the test button, see a green light, and assume everything was fine. That only proves the battery has a charge—it says nothing about whether the PSU will survive the transfer. What I should've been doing was simulating a voltage sag to see how the system actually behaved under load.

What I do now

I ended up going with a CyberPower online UPS from the cyberpower-ups product line—dual-conversion design, zero-transfer-time, pure sine wave output. It's overkill for some setups, but for our use case, the TCO math actually favored it once I stopped comparing sticker prices.

I won't claim I never make mistakes. I just make different ones now. But the checklist I keep looks like this:

  • Calculate TCO before comparing any vendor quotes. Include the cost of the failure you're trying to prevent—recovery costs, downtime hours, replacement parts.
  • Check compatibility specs line by line. Transfer time, waveform type, VA rating at load. Not just "is the plug the right shape."
  • Test under realistic conditions. A green light isn't a test. Simulate the failure and watch what happens.
  • Ask someone who knows more than you. I'm not an electrical engineer—anything beyond basic selection criteria, I call a consultant. That's not weakness. It's cheaper than guessing.

That $340 lesson stung. But it taught me something the spec sheets never would have: the cheapest UPS is the one you only have to buy once.