A rectified mains supply ripples at twice the line frequency, so on the Saudi 60 Hz grid an LED's output pulses at 120 Hz. Done well you never notice. Done badly it bands on camera, and it is measurable: IEEE 1789 grades depth against frequency, CIE TN 006 defines the modern metrics, and IEC TR 61547-1 and IEC TR 63158 give the methods for measuring Pst LM and SVM.
Four ideas to carry into the next specification conversation you have.
IEEE 1789 grades the pair, not either alone. On the Saudi 120 Hz ripple the low-risk ceiling is 9.6% modulation and the no-effect ceiling is 4.0%. At 60 Hz the same two lines allow only 1.5% and 0.6%.
Percent flicker = 100 × (max − min) / (max + min). Flicker index = the area of one cycle above the average light level, over the total area. Neither carries any frequency information at all, which is exactly why neither is enough on its own.
Stroboscopic Visibility Measure, defined in CIE TN 006. SVM = 1 is the visibility threshold, where an average observer detects the effect half the time. The EU's SVM ≤ 0.4 is a legal ceiling well below that, not a "no effect" line.
Measured by the light flickermeter method in IEC TR 61547-1, which sets no limit itself. Pst LM = 1.0 does not mean imperceptible: it means the source is as flickery as a 60 W incandescent lamp, which about half of viewers notice.
The Saudi grid runs at 60 Hz, so the ripple to design against is 120 Hz, not the 100 Hz behind most published rules of thumb. IEEE 1789 supplies the depth-against-frequency lines; the Pst LM and SVM ceilings are EU law, not measurement standards.
| Metric | Threshold | Affected scenario |
|---|---|---|
| Modulation depth at 120 Hz | < 9.6% | IEEE 1789 low risk, 0.08 × f, on the Saudi ripple |
| Modulation depth at 120 Hz | < 4.0% | IEEE 1789 no observable effect, 0.0333 × f |
| Modulation depth below 90 Hz | < 5% | IEEE 1789 Recommended Practice 3, the only unconditional one |
| Pst LM | ≤ 1.0 at full load | EU ecodesign limit. Pst LM = 1 is the 50% detection point, not silence |
| SVM | ≤ 0.4 at full load | EU ecodesign since 1 Sep 2024. Visibility threshold is SVM = 1 |
| Percent amplitude modulation | < 30% below 200 Hz | California Title 24 JA8, required at full output and at 20% |
Five points the lighting industry only agreed on after 2015.
Magnetic-ballast fluorescent measured roughly 30% to 49% modulation, not the 100% often quoted. A 1989 double-blind cross-over study in occupied offices more than halved headaches and eyestrain by moving from switch-start choke ballasts at 43% to 49% modulation to 32 kHz electronic ballasts under 7%. LEDs reopened the conversation because some drivers are worse than the fluorescent they replaced, and cameras finally exposed it.
The 2015 standard plots modulation depth against modulation frequency and shades two regions: the no-observable-effect region and the low-risk region. It names no third zone, and it says outright that operating outside them "does not necessarily imply high risk". The word NIOSH appears nowhere in it. Above 1250 Hz, 100% modulation becomes low risk; it only becomes no-observable-effect above 3000 Hz. And 1250 Hz is not a rule in its own right, it is simply where the line 0.08 × f reaches 100%, which is why it attaches to PWM dimming.
A rolling-shutter sensor writes the light's waveform down the frame, so a phone can reveal modulation a viewer cannot see. It is a one-way test: a phone can reveal temporal light modulation but cannot certify its absence, because the anti-banding control defaults to automatic and its documented job is to choose an exposure that hides the bands. A clean image is not a compliance test. To screen a fixture, lock the phone to manual exposure at a speed that does not divide the modulation period, such as 1/250 on the Saudi grid, and look for bands.
PWM dimming produces 100% percent flicker by construction; reducing the drive current instead holds it near zero. When California tested 25 LED A-lamps against its own filtered limit, 13 passed at both full output and 25% output, and 12 failed at 100%, at 25%, or at both. Dimming can expose modulation that a full-output test alone does not show. On a separate set of PAR38 lamps, simply installing a phase-cut dimmer and leaving it at full output raised measured flicker index by as much as 0.15, and by a ratio of at least three to one the products that moved got worse rather than better. Specify the requirement at 100%, 50%, 20% and minimum output.
Only one major scheme has a real gate: WELL v2 Feature L08. LEED v4.1 has no flicker requirement at all, in either the BD+C or the ID+C guides. BREEAM's only flicker criterion is a legacy high-frequency-ballast one for fluorescent lamps. Mostadam has no flicker credit, and its indoor lighting credit awards its point to fluorescent with high-frequency ballasts. The SASO lighting energy-efficiency standards reviewed, 2870, 2902 and 2927, set no Pst LM, SVM or modulation-depth limit. In Saudi projects, flicker performance therefore often rests on the project specification: write a limit and it can be tested and enforced.
"Flicker-free" is not a standardized compliance claim, so on its own it cannot be tested for. Ask any supplier, including us, for the pair that can be: percent amplitude modulation and its fundamental frequency, for the complete luminaire with its actual driver and control device, measured at 100%, 50%, 20% and minimum output at the site's nominal supply, 230 V and 60 Hz.
Your eye averages the pulse away. A camera sensor samples it in milliseconds, so the strobe your eye misses lands straight on the frame.
The light pulses with the mains, hundreds of times a second. Here it is slowed right down so you can watch it dim and brighten.
Many rolling-shutter sensors expose one row at a time, top to bottom, rather than the whole frame at once.
Rows caught on a dim pulse record dark, so the frame fills with the rolling dark bands you see on camera.
Most CMOS sensors, in phones and cinema cameras alike, expose the image one row of pixels at a time, top to bottom, rather than all at once. While the sensor scans down, a flickering light keeps pulsing bright then dim. Rows captured on a bright pulse come out light; rows captured on a dim pulse come out dark. The result is a stack of horizontal dark bands rolling through the frame.
In a single photo the bands freeze in place and ruin a product shot or a portrait under the wrong panel. In video they drift frame to frame and the clip breathes between bright and dark. High frame rates force short exposures, and a short exposure is precisely the case where almost none of the modulation is averaged away, so slow motion under a flickering source strobes hard. Frame rate and shutter speed fix different problems: frame-rate divisibility settles frame-to-frame pulsing, only exposure-time divisibility removes the bands.
The exposure has to contain a whole number of modulation periods. Integrating the light over a window of length T attenuates its modulation by |sinc(f × T)|, which is exactly zero when T is a whole multiple of 1/f, and because every harmonic is nulled at the same moment the rule works on any periodic waveform. Saudi mains runs at 60 Hz, so the light pulses at 120 Hz and the safe shutter speeds are 1/120, 1/60, 1/40, 1/30 and 1/24. 1/240 is not one of them: it is half a period, and it leaves about 64% of the modulation in the frame for a sinusoidal source and up to all of it for a PWM driver. 1/100 leaves about 16%, 1/250 about 66%, and 1/1000 about 98%.
Shutter tricks only patch one camera at one setting, so the fix belongs in the fixture. Raising the modulation frequency buys headroom rather than immunity: what survives an exposure is still |sinc(f × T)|, so a 1.25 kHz driver reads clean at ordinary shutter speeds, where the exposure spans many pulses, and still writes about 96% of its depth into a 1/8000 s frame. The depth is what makes it invisible: a driver held under the no-observable-effect line has little left to record at any shutter speed. Specify the depth and the frequency together at the fixture, and stop chasing either one in the camera.
Practical lighting know-how from NLC Academy. No spam, unsubscribe anytime.
By subscribing you agree to receive NLC emails.
The standards, regulations and test reports behind every number on this page. Where a document defines a measurement method but sets no limit, that is stated.
A 15-slide explainer, ready to post straight to LinkedIn as an image carousel. Same content, faster reading.
