Connected architecture: DALI-2 · 0-10V · occupancy · daylight harvesting

LED Fixtures Keep Failing? It Wasn't the Fixtures (It Was the Surge Path)

What You're Seeing: LED Fixtures Dying Way Too Early

In September 2022, we handed over 14 LED fixtures on a four-storey office retrofit. By November, three drivers had burned out. By early 2023, five more were gone.

The tenant complained. The tenant's complaint went to the building owner, who then went to me. The owner's exact words: "Did you use the cheap brand?"

My answer? "No — the fixtures are fine, it's the drivers." I sounded confident. Then I replaced six more drivers over the following five months. At one point, 11 of those 14 fixtures had been through at least one driver swap. Individually, the drivers weren't expensive. Collectively, with labor, return visits, and a lot of trips up and down a basement stairwell, the number started to get ugly.

For context: I've been handling lighting retrofit and maintenance orders for commercial facilities for nine years. I've personally made — and documented — five significant mistakes in that time, totaling roughly $8,400 in wasted budget. I maintain our team's pre-check list so nobody else has to repeat them. Every single one of those mistakes started the same way: I blamed the wrong thing first.

This one was no different. The fixtures weren't the problem. (Should mention: they were never the problem, at any point.) The surge path and the control circuit were. It took me a year to figure that out, and that delay was the expensive part.

The Part Nobody Was Looking At: Where the Surges Were Coming From

Here's the thing about surges in commercial buildings — most people mentally file them under "server room stuff." But an LED driver is a switch-mode power supply. It's more sensitive to transient voltages than the old magnetic ballasts we pulled out. Your legacy fluorescent fixture could shrug off something a driver will not.

IEEE Standard C62.41.2 classifies surge environments by exposure category — the service entrance being the harshest, then the panel, then the feeder, then the branch circuit. In nearly every commercial retrofit I've walked into, nobody checks whether the service entrance has surge protection. NEC Article 285 covers surge protective devices, and it's the kind of code section nobody reads until something starts failing.

That building had no service-level surge device. There was a small suppressor on one of the subpanels — the kind meant for a dryer circuit — that someone had pressed into duty years earlier. I found it during the postmortem. Not during the install.

So when the first drivers failed, what did I do? I replaced the drivers. I didn't chase the surge path, because I wasn't looking at it.

And Then the Second Mistake: The Three-Way Switch

Stay with me — this one's a wiring issue, not a product issue, but it compounds the surge problem.

The corridor was controlled from two ends by a pair of three-way switches, and during the retrofit those switches got wired through a lighting contactor.

Here's what I said to the electrician: "Standard three-way, just match the existing wiring."

He didn't mishear me. I misheard myself — what I meant was "follow the existing configuration," and what he heard was "wire it up any way that works." Those two instructions are very different once you're standing at a terminal block. A three-way switch uses two travelers and a common. Get the common upside down and you end up with a tiny amount of leakage on the switched leg — not enough to trip anything, not enough for a meter to scream at you, but enough to give the drivers a constant flicker of voltage every time the circuit "closed."

I later walked the circuit against Eaton's light switch wiring diagram — the one on eaton.com — terminal by terminal. That's how I found it. If you have a three-way switch in front of you, pull a wiring diagram before you start, not after. Trust me on that.

What did that wiring error cost? Somewhere around 40% of the driver failures in that building. Not all of them, but a meaningful chunk.

And the Third Mistake: The LED Load Was Never Sized for That Environment

We'd swapped the fixtures to LED — but the strip we chose was rated for a stable-temperature office environment. The corridor ran off a lighting contactor that shared a feed with a motor circuit. Every time the building's compressor kicked on, the voltage on that leg dipped. Nothing dramatic — you couldn't see it at noon. In the evening, when the corridor dimmed and the tenant walked through, they'd notice the one fixture at the end of the hall flicker. That's where the spotlight was, so to speak — every single person who walked past stopped and looked at it.

Put another way: the LED strips weren't bad. They were mismatched to the electrical environment, and nobody was checking that during the initial walkthrough.

What It Actually Cost: About $3,200

After the third rework in Q1 2024, I built a pre-check list. Should have done it after the first one. Apparently I needed three tries.

Here's the accounting that finally made me sit down with a notebook:

  • Replacement drivers: roughly $1,850 across the building. Two batches were supposed to be warranty, but "warranty" on a driver still means labor hours and two weeks of waiting — with the tenant still operating.
  • Extra labor: about $1,100, most of it return visits scheduled after hours because the tenant was running during the day.
  • Replacement LED strips for the corridor run, plus a matched dimmer: roughly $400.

That's $3,200-ish. And the dollars weren't the worst of it. By Q1 2024, the owner was genuinely asking whether our maintenance contract was worth renewing. That's the cost that doesn't show up on an invoice. That's the one that stays with you.

And Yes, This Is a TCO Conversation

In my first year (2017), I made the classic mistake — I compared three vendor quotes by unit price and picked the cheapest. I assumed the lowest number was the best deal. Two budget overruns later, I learned about total cost of ownership, and I've been calculating it before making any purchase decision ever since.

Run this job through a TCO lens and the numbers flip fast. The "cheaper" fixture package saved about $400 up front. The rework ate $3,200 and damaged the relationship with the client. If I'd listed the hidden line items before I bought — replacement parts, after-hours labor, downtime, the renewal conversation — the cheaper option stops looking cheap immediately.

What I mean is: TCO isn't just a nice phrase for a procurement slide. It's the total you're actually going to pay when the cheapest option doesn't work and you have to do it again. Unit price plus expediting plus labor plus rework plus the cost of your own time.

The Fix — Which Is Shorter Than the Problem

Here's what I actually changed. It's four items, and none of them are exotic.

1. Treat Surge Protection as Infrastructure, Not an Accessory

We ultimately installed the Eaton CHSPT2ULTRA. I'll be straight with you about why I picked it: it rated correctly for my panel, it was available, and I didn't want to spend a month evaluating options while fixtures kept failing. Check the current spec sheet on eaton.com rather than relying on a three-year-old forum post — model numbers and ratings have a way of drifting.

I read the reviews. The vendor pages are, predictably, glowing. What I noticed in the non-marketing write-ups is that the real debate is about installation — some people DIY it, some won't touch it without a licensed electrician. Both are defensible, but whichever you pick should be quoted alongside the device itself. That's a TCO point: an SPD that costs X uninstalled can end up at 2X once you've paid someone to put it in. Budget for the total or don't budget at all.

The device itself has been doing its job. No driver failures since it went in. That's not a product review — that's a data point from one building over roughly nine months. Your situation may differ.

2. Walk the Three-Way Switches Against a Diagram

Every job now, before we close up a wall, I pull the wiring diagram and verify the common and the travelers. It takes five minutes.

I'm not going to claim this is a gadget — it's not. It just means there's a wiring diagram on site and nobody is guessing. It's honestly the cheapest fix in the entire story.

3. Match the LED Load to the Environment

Not just the fixture to the ceiling. The load to the environment. If the circuit shares a feed with motors, you need to know that before you spec the strip. If the corridor runs off a contactor, ask what else runs off that contactor.

4. Run a Five-Minute Pre-Check, Every Time

Mine has four questions on it now: Is there a service-level SPD? Do the three-way switches have a verified diagram on site? What else is on this circuit's feeder? Is the driver rated for this temperature and load?

Five minutes. We've caught 47 potential errors with this list over the past 18 months — mostly wiring and load mismatches that would have shown up as callbacks a year later.

The Bottom Line

If your LED fixtures keep dying, the answer is almost never the fixture. Check the surge path first, then the control circuit, then the load sizing. In that order.

And do the TCO math before you buy — not after the third driver fails. The $400 you saved on the purchase order has a way of turning into $3,200 on the maintenance log.

I learned that the expensive way. You don't have to.