Home › Blog › LED Flicker and Stroboscopic Effect
Published August 7, 2026 · By Industrial Lighting GR Editorial · ~9 min read
LED flicker is a periodic dip in light output set by the driver, not the diode. Above roughly 3 kHz it is harmless. Near 120 Hz with deep modulation it causes eyestrain and can make rotating machinery appear to stand still. Specify driver frequency and percent flicker before you buy.
Every LED array is a switch. It is either passing current or it is not. The driver decides how fast that switching happens and how deep the dips go, and cheap drivers do it slowly with a big swing. That is flicker: a periodic drop in light output riding on top of the light you think you are getting.
Two numbers describe it. Flicker percent is the depth of the swing, the difference between peak and trough divided by their sum. Flicker frequency is how many times per second it happens. A fixture at 8 percent flicker running at 30 kHz is invisible and harmless. The same 8 percent at 100 Hz is a headache generator on second shift.
Magnetic-ballasted metal halide and old T12 fluorescent flickered at 120 Hz, twice line frequency, and everyone lived with it. The frustrating part of a bad LED retrofit is that it can be worse. A poorly filtered driver can run 100 percent modulation at 120 Hz, meaning the light goes fully dark 120 times a second. Your eye does not register it. A camera does. So does a spinning shaft.
Here is the failure mode that turns a lighting spec into a safety issue. When light pulses at a frequency that lines up with the rotation of a machine part, the part appears to slow down, stand still, or turn backward. Anyone who has watched wagon wheels roll backward in an old movie has seen it. On a plant floor it is not entertaining.
A grinder wheel, a lathe chuck, a saw blade, an unguarded shaft coupling, a mixer paddle: any of these can be turning at full speed and read as stationary under 120 Hz light. The IEEE recommended practice lists the apparent slowing or stopping of motion as one of the documented consequences of modulated light, alongside headaches, eyestrain, migraine, reduced task performance, and seizure risk in photosensitive individuals.
The risk is not evenly distributed. It concentrates exactly where you would expect: machine shops, stamping, packaging lines with rotating star wheels, and anywhere a maintenance tech reaches toward equipment that was supposed to be locked out. If your facility runs robot cells or automated equipment, the lighting question sits next to the access control question, and both belong in the same walkthrough.
IEEE 1789, "Recommended Practices for Modulating Current in High-Brightness LEDs for Mitigating Health Risks to Viewers," is the document to point at when a vendor tells you flicker is not a real thing. It is a recommended practice, not a code, and that distinction matters when you are writing a purchase spec rather than citing a violation.
The practical takeaway is a frequency threshold. Above roughly 3 kHz, no evidence of human effects has been found, which is why the standard's guidance is often reduced to a single line: run the driver above 3 kHz and stop worrying. Below that, the allowable flicker percent tightens sharply as frequency drops, and the low-risk region below about 90 Hz is narrow enough that most commodity drivers cannot meet it.
The US Department of Energy's solid-state lighting program maintains a plain-language explainer on flicker that is worth forwarding to whoever signs the purchase order. It covers the metrics without the standards fee.
You do not need lab equipment for a first pass. You need a phone and five minutes.
Open the camera app, do not record, just watch the preview while pointing it at each fixture type in the building. Rolling-shutter sensors turn light modulation into moving horizontal bands. Strong bands mean deep flicker at a low frequency. A clean image means the driver is running fast enough that the sensor cannot resolve it. This is a screening test, not a measurement, and it will miss high-frequency low-depth modulation, which is the kind that does not matter anyway.
The second test costs nothing either. Spin something. A shop fan, a hand drill in a vise, a belt-driven pulley on a machine that is safe to run. Vary the speed and watch for the point where motion smears or stalls. If you find one, you have a stroboscopic hazard in that bay and a documented reason to replace the fixtures over it first.
For anything past screening, a handheld flicker meter reporting percent flicker and frequency runs a few hundred dollars, and most lighting contractors who take the topic seriously already own one. We bring one to every audit, and it goes in the report next to the foot-candle readings.
Most flicker problems are bought, not developed. They arrive with a pallet of fixtures that met every line of the spec because the spec never mentioned flicker. Three lines fix that.
Driver switching frequency above 3 kHz. This is the single highest-value requirement and it is cheap to meet. Reputable driver manufacturers publish it. If a vendor cannot produce the figure from a datasheet, that is your answer.
Percent flicker below 10 percent at full output, and at every dimmed setting you intend to use. The second half of that sentence is the one vendors skip. A driver can be clean at 100 percent and awful at 20 percent, because many dimming schemes work by pulse-width modulation, chopping the output into on and off periods. If you are specifying occupancy sensors or daylight harvesting, the dimmed condition is the normal condition, not the exception.
Flicker data reported per the IEEE 1789 framework, submitted with the quote. Not a marketing claim of "flicker free." A number, at a frequency, at a drive level.
Adding these does not meaningfully change fixture cost on a commercial-grade product. It changes which products can bid, which is the point. If you are building the payback case for a retrofit, note that these lines protect the payback: a fixture crew coming back to swap out headache-inducing lights is not in anyone's ROI model.
Three patterns show up repeatedly across Grand Rapids, Wyoming, Holland, and the Kalamazoo corridor.
The first is the mixed-vintage building. A plant retrofits the warehouse in one phase and the machining bay two years later from a different supplier. The bays look identical and behave nothing alike, and nobody connects the second-shift complaints to the fixtures because the first phase was fine.
The second is the dimming retrofit. Controls get added to an installed base of fixtures whose drivers were never specified for smooth dimming. The energy savings land, the light quality falls apart at low output, and the operators start overriding the sensors. That kills the savings and the controls project at the same time.
The third is the office-grade fixture in an industrial space. Low-cost troffers and strips designed for a break room end up over a bench where someone runs a bandsaw. The fixture is not defective. It is in the wrong room.
None of these require a full teardown to fix. They require knowing which fixtures are the problem, which is a measurement, not a guess. Our manufacturing facility lighting work starts with that measurement, and the audit report tells you which bays to address first and what it costs.
Yes. When light pulses at a rate that lines up with a rotating part, the part can appear to slow, stop, or reverse. This is the stroboscopic effect, and IEEE 1789 lists it among the documented consequences of modulated light. It is a real hazard around grinders, saws, and exposed shafts.
IEEE 1789 notes that no evidence of human effects has been found above roughly 3 kHz, which is why most specifications simply require a driver switching frequency above that mark. Below 3 kHz the acceptable flicker depth tightens quickly, and commodity drivers running near 120 Hz rarely stay inside it.
No. OSHA has no flicker-specific standard, and 1910 sets no maximum flicker percent. The exposure shows up indirectly through machine guarding and the general duty clause when a stroboscopic effect contributes to an injury. Treat IEEE 1789 as a purchasing spec, not as a citation you are avoiding.
Point a phone camera at each fixture type and watch the preview. Rolling horizontal bands mean deep low-frequency modulation. Then spin a fan or a drill under the light and vary the speed, watching for motion that smears or stalls. Both are screening tests, not measurements, but they find the worst offenders.
They can. Many dimming schemes chop driver output into on and off periods, so a fixture that measures clean at full output can flicker badly at 20 percent. If you are adding occupancy sensors or daylight harvesting, require flicker data at the dimmed levels you actually plan to run, not just at full.
IEEE 1789 lists headaches, fatigue, blurred vision, eyestrain, and migraine among the reported effects of modulated light, along with reduced visual task performance. Employees usually report the symptoms without connecting them to the lighting, which is why flicker complaints tend to arrive as vague comments about a bay feeling harsh.
Industrial Lighting GR's editorial is led by senior lighting designers with 15+ years of West Michigan industrial and commercial experience. We measure percent flicker and driver frequency alongside foot-candle levels on every audit, write flicker limits into retrofit specifications before fixtures are ordered, verify performance at dimmed levels when controls are part of the scope, and carry Consumers Energy and DTE rebate paperwork through pre-approval, install, and final payment. We serve Grand Rapids, Wyoming, Kentwood, Walker, Holland, Muskegon, Kalamazoo, and surrounding West Michigan manufacturers.