How to Check a LED Driver: The Flicker That Wasn't the Driver's Fault

It was a Thursday afternoon in early November when I caught the light in conference room 3B doing something it shouldn't have been doing.

Not a dramatic failure. No buzzing, no dead fixture, no strobe effect. Just a subtle pulse—the kind of rhythm that makes you squint, look away, look back, and wonder if you were seeing things. The light would sit at full brightness for a few seconds, dip maybe ten percent, then climb back up. Over and over.

We were close to finishing a lighting retrofit at a 30,000-square-foot office building. 140-odd LED fixtures, Leviton smart switches with WiFi through the whole floor, Zigbee motion sensors in every room. My job on these projects is the final quality walkthrough: I review every installation before it reaches the client—roughly 200 unique systems a year. This one was about to teach me something.

I'd been through three zones already. Conference room 3B made it three and a half, because I stopped at the doorway and watched that light pulse for a solid twenty seconds.

I called the lead electrician over. He looked up, watched it for maybe three seconds, and said, "It's the driver. I'll order a replacement."

"Hold on," I said. "How do you know it's the driver?"

He gave me a look. The kind of look you give someone who just asked why the sky is blue. "Because that's what flicker means," he said. "The driver's the most common failure point in LED fixtures. Everyone knows that."

Everyone knows that.

Well, everyone was about to be wrong.

Why we test before we swap

Look, I'm not going to argue with the statistics. LED drivers fail more often than any other component in a luminaire. They handle heat, current, and voltage conversion, and they're a common culprit in the "how to check a LED driver" searches plenty of contractors end up running. But "most likely" is not "guaranteed." And replacing a part you haven't tested is a gamble—one that almost cost us a week on this project.

Here's what we did instead.

Step one: look at the driver

We pulled the fixture down and opened it on the workbench. The driver casing was clean. No swelling, no burn marks, no melted or discolored wiring. The electrolytic capacitors weren't bulging. For an electrician, this is the equivalent of taking a patient's temperature: it won't catch everything, but it rules out a lot.

Step two: verify the input side

Multimeter on the line wires. The driver was rated for 120–277V AC. We measured 121.6V at that junction box. Within spec, no flicker or sag on the supply side. If this reading had been low or unstable, the problem would have been upstream—something for the branch circuit or the utility, not the driver.

Step three: measure the output under load

This is the critical test. The driver was a UL 8750-listed Class 2 unit, rated at 54V DC and 700mA constant current. We connected it to a known-good test fixture we keep on the truck, put the probes on the output wires, and set the multimeter to DC volts.

54.2V. Steady. No drifting. No sag.

The electrician stared at the multimeter like it had lied to him. "So the driver's fine?"

It was. It was doing its job perfectly.

Step four: trace the control chain

If not the driver, then what? We worked backward from the fixture. First, the Leviton plugs and connectors that tie the driver to the LED board—all seated, no corrosion, no loose pins, the kind of neat terminations the National Electrical Code expects. Then the Leviton smart switch WiFi unit that controls the zone. The app showed 100% brightness, firmware current, no error codes. Then the Zigbee sensor in the corner of the room, which was reporting occupancy correctly and responding to commands.

Everything checked out individually. That should have been our first clue.

(note to self: when everything passes one by one, the problem is usually in how they talk to each other.)

The discovery that reframed everything

We finally dug into the settings—the part of a lighting system that has no smell, no heat signature, no visible failure mode. The Zigbee motion sensor had a timeout set to five minutes. This conference room sat empty for long stretches between meetings. So every five minutes, the sensor sent the switch a command: drop to 10%.

Now, the Leviton smart switch was configured with a low-end trim of 5%—meaning it would happily dim down to 5% if a sensor or a manual press asked for it. But the LED driver was only rated to dim down to 10%. Below that, its control circuit started oscillating.

That oscillation was the flicker.

The driver wasn't misbehaving. It was following instructions it should never have been given.

People think a flickering LED equals a failing driver. The reality, in this case at least, was the reverse. The driver was responding exactly as designed to a command that the rest of the system shouldn't have sent. Blaming the driver for a low-end trim mismatch is like blaming the engine for being floored in third gear. The causation runs the other way.

We changed two settings: raised the low-end trim from 5% to 15% on the smart switch, and extended the sensor timeout from five minutes to ten. The flicker disappeared. Nothing was replaced. Not a single component.

What this taught us about Zigbee systems

Field experience has given me a complicated relationship with Zigbee. On one hand, the protocol is mature and stable—it's the backbone of a huge share of commercial lighting controls. On the other, "Zigbee compatibility" is one of the most overused scapegoats I know. When a Zigbee sensor and a Zigbee LED controller or fixture misbehave, the first suspect should be configuration, not the technology. We use plenty of Zigbee LED controllers in our projects, and the ones that pass real-world testing are reliable.

That day was proof: the system's failure wasn't a hardware failure at all. A sensor, a switch, and a driver—each doing exactly what it was programmed to do. The only real flaw was an untested assumption about how they'd interact.

Five minutes of verification beats five days of correction

That project finished on time, but it changed how we work. Every installation now ends with a commissioning checklist, and it's the most valuable piece of paper we carry:

• Driver: visual inspection and live output test before connecting the fixture.
• Connections: verify Leviton plugs and connectors, and any other splice, are seated and secure.
• Dimmer/trim: confirm the switch's low-end trim falls inside the driver's published dimming range.
• Sensor behavior: set timeouts and thresholds to match how the room is actually used.
• Chain test: exercise the full path—sensor, switch, driver, fixture—before sign-off.

Since we started running this list, we've avoided at least six unnecessary driver replacements. Each one would have meant a backordered part, a return visit, and a client wondering why the fix was taking so long. The direct savings in labor and parts are real. The indirect savings—the trust we don't lose, the timelines we don't blow—matter even more.

Five minutes of verification beats five days of correction. I believe that more strongly after that afternoon in conference room 3B than after any other project I've worked on.

So if you're staring at a flickering LED right now, your next step isn't to order a new driver. It's to get your multimeter and work through the chain. The driver might actually be bad—and if it is, you'll know. But you'll be surprised how often the fault shows up in the settings, the connections, or the assumptions you brought into the room.

It was, for us.