-
Before I drove out, I asked for two documents
-
First round: the Legrand dimmer wasn't the problem
-
The Zigbee piece isn't usually the villain
-
The flashing emergency light was the tell
-
Sunday night, the system ran — but the math wasn't pretty
-
What made us change our policy
-
The one question I keep asking now
It was 9 p.m. on a Friday in November 2024, and my phone lit up with a call I knew would eat my weekend.
"The lights downstairs flicker whenever someone walks through. The Legrand occupancy sensor switch on level two stopped responding. And one of the emergency lights in the back warehouse keeps flashing like someone's playing with the switch. We've got the walkthrough at 8 a.m. Monday."
I've been doing emergency lighting commissioning for 11 years. In my role coordinating rush jobs, I've handled roughly 200 after-hours calls like this one. Forty-eight separately controlled zones, three symptoms, 36 hours. Tight but doable — normally.
What made this one different: the general contractor had swapped out the fixture drivers to save $3,200. That detail won't seem important until later in this story, but it's the reason the job nearly blew up.
Before I drove out, I asked for two documents
Not five. Two.
First: a copy of the Legrand occupancy sensor switch manual for the exact model. Second: the driver spec sheet for every fixture on that floor.
The first one he had. The second one didn't exist. No spec sheet, no compatibility list, no manufacturer rating — just a generic line of "dimmable LED downlights" bought from an online reseller.
This is the blind spot I see over and over. Most buyers focus on the brand of the gateway and the switch and completely miss the driver. The switch sends the command. The dimmer handles the 0-10V curve. The driver decides what the light actually does. Mismatch those, and nothing else matters.
I told him that before I arrived. He said, "Just get it working by Monday." Fair enough. That's what I do.
First round: the Legrand dimmer wasn't the problem
We arrived around 11 p.m. First thing we checked: was the LED dimmer Legrand rated for the load? It was a wireless dimmer, properly rated, wired correctly. Its communication with the gateway was clean.
So we moved down the chain. We scoped the dimming curve on a sample fixture. It was supposed to climb through each step. Instead it dropped hard at every step. Not a gentle curve — a cliff. At minimum load, the current reading jumped erratically instead of holding steady.
That's a mismatched load impedance. The dimmer is doing exactly what it's told. The driver is arguing with it.
Not ideal, but workable — if we had a driver that matched the spec. We didn't.
The Zigbee piece isn't usually the villain
Then we moved to the occupancy sensor on the second floor. This one's a Legrand occupancy sensor, battery-powered, paired according to the manual's procedure. Pairing kept failing under the Zigbee HA configuration.
Here's what most people don't realize: the sensor is only as reliable as the MCU Zigbee firmware on the gateway it talks to. When we checked, the gateway was still on its factory firmware — not the version listed in the manual we'd pulled earlier.
We reflashed the gateway following the exact steps in the manual, then re-ran the pairing. It took about 25 minutes. Clean pair, stable link.
Almost fixed. But "almost fixed" isn't a Monday-morning pass, and the driver issue was still sitting in the room.
The flashing emergency light was the tell
The warehouse emergency light was the weirdest one. It's a UL 924-rated fixture, which under UL 924 must provide at least 90 minutes of emergency operation after normal power is lost. A UL 924 fixture should not flicker at all under normal conditions.
We started at the power source. The fixture was sharing a circuit with one of the dimmed fixture branches, fed by an off-brand driver. When we measured during the fixture's self-test cycle, the voltage on the shared branch was surging and dropping in rhythm with the dimming cycle.
So it wasn't the emergency light. It was the shared circuit being dragged around by a cheap driver. We isolated the fixture onto its own branch, recalculated load for that circuit, and left it running overnight. Stable. No flicker for 14 hours.
Sunday night, the system ran — but the math wasn't pretty
By Sunday night, everything was live: occupancy sensors behaved, dimming was smooth across all 48 zones, emergency lights held. The client passed the walkthrough Monday morning at 8:14 a.m.
Now let's talk about what that weekend actually cost.
- Saved on cheaper drivers: $3,200
- Emergency weekend labor (two technicians, 22 hours): ~$2,900
- Replacement drivers pulled out and written off: ~$1,900
- Contractor's holdback at risk if the walkthrough failed: $12,000
The "saving" turned into at least $4,800 in extra spend — and that's before counting the two other rush jobs I had to turn down that same week.
This is the total cost of ownership problem in a nutshell. The unit price is the easiest number to compare, so it's the number everyone compares. It's also the number most likely to lie to you.
When I calculate TCO now, I add: unit price + expedited shipping + the cost of a redo + the cost of a delay + the cost of my team un-burying a problem that shouldn't have existed. That's not an invoice line item. That's the project.
What made us change our policy
The policy didn't come from a spreadsheet. It came from a redo. About a year and a half ago, we ate an $8,000 driver reorder on a job that nearly pushed a client past their grand opening. A seasoned contractor had warned me about off-brand drivers months before. I didn't listen — the spec sheet said "dimmable."
"Dimmable" is a marketing word. "Compatible with this dimmer at this load" is the only thing that matters.
Now our internal rule is simple: no driver compatibility sheet, no project sign-off. It has cost us a couple of bids from clients who wanted to trim fixtures from the spec. It has also saved us from at least four weekends like the one above.
The one question I keep asking now
When I get a rush call today, the first question isn't "what's broken?" The first question is:
"Where's the driver compatibility sheet, and what dimmer was it tested against?"
If the answer is "it's fine, they're all dimmable," I already know what kind of weekend I'm walking into.
The cheapest part in the room is almost never the cheapest part of the project. And when an emergency light is flashing, the real problem is rarely the light.
It's the thing you decided not to look at six months earlier.

