LED Flicker Explained: Pst LM, SVM and Dimmed Performance
Flicker complaints rarely arrive during commissioning. The fittings pass the walk-through at full output, the handover signs off, and the problem surfaces weeks later: a camera on a phone shows dark bands across a meeting-room display, warehouse staff report that a spinning conveyor roller “looks stopped”, or a hotel guest notices a shimmer only after the lounge scene is dimmed to 15%.
None of these is a lamp failure. Each is a temporal light artefact (TLA): a consequence of light output that is not constant over time. LEDs respond almost instantly to drive current, so driver ripple becomes light ripple; incandescent filaments smoothed it thermally.
The commercial difficulty is that most quotations describe this whole subject in two words — “flicker-free” — which is not a measured quantity and cannot be verified at incoming inspection. This article explains the metrics that do carry meaning, where regulatory limits actually apply, and how to write a requirement that still holds when the lights are dimmed.
Information in this article is current as of 30 September 2026. Standard and regulation editions are cited explicitly; verify against the current edition before writing a specification.
1. Flicker, Stroboscopic Effect and Phantom Array Are Different Problems
“Flicker” is used loosely for three distinct perceptual effects, and each one is captured by a different metric.
- Flicker — light-level variation that a static observer sees directly, looking at a static scene. It is dominated by low modulation frequencies, roughly below 80 Hz.
- Stroboscopic effect — a static observer looking at moving objects sees their motion distorted: a rotating part appears slowed, stopped or reversed. It matters for modulation frequencies from roughly 80 Hz up to about 2,000 Hz.
- Phantom array — when the observer’s own eyes move quickly across a light source, a trail of discrete images appears. Some studies report it at frequencies above 2 kHz.
The distinction matters because a driver can perform well on one effect and poorly on another. A clean low-frequency profile can coexist with strong 100 Hz or 120 Hz ripple — invisible on a wall, obvious on a lathe or a fan blade. A single “flicker” figure does not say which effect was assessed.
2. The Metrics, and What Each One Can Tell You
Five quantities appear in test reports: two simple waveform descriptors, two perception-weighted metrics, and the modulation frequency that gives them context.
| Metric | What it measures | Weighted for human perception? | Where it appears |
|---|---|---|---|
| Percent flicker (modulation depth) | (max − min) ÷ (max + min) of the light waveform, as a percentage | No | Legacy reports, California JA10 method, IEEE 1789 risk lines |
| Flicker index | Area above the average light level divided by total area under one cycle | No | Legacy reports |
| Pst LM | Short-term flicker severity from a light flickermeter; 1.0 means a 50% probability that an average observer detects flicker | Yes (flicker) | EU Regulation 2019/2020, IEC TR 61547-1 |
| SVM | Stroboscopic visibility measure; 1.0 is the visibility threshold for an average observer | Yes (stroboscopic effect) | EU Regulation 2019/2020, IEC TR 63158 |
| Frequency of the dominant component | Where the modulation energy sits, in Hz | No, but essential context | Needed to interpret percent flicker against IEEE 1789 |
Two practical conclusions follow from the table.
Percent flicker alone is incomplete. Compare 30% modulation at 100 Hz with 30% at 3 kHz: the same percent flicker, very different consequences. Without the frequency, the figure cannot be judged.
Pst LM and SVM are not interchangeable. Pst LM targets the low-frequency visible flicker; SVM targets the stroboscopic range. A product can show Pst LM well below 1.0 and still fail an SVM limit, because the dominant ripple sits at twice mains frequency — exactly the band SVM is designed to catch.
The measurement methods are standardised: Pst LM is measured with the objective light flickermeter described in IEC TR 61547-1, which adapts the voltage flickermeter of IEC 61000-4-15 to illuminance; SVM is measured per IEC TR 63158. Asking for results “per IEC TR 61547-1 and IEC TR 63158” is a quick way to see whether a supplier’s figures come from a defined method or from a generic meter.
3. Where the Limits Actually Apply
European Union: Regulation (EU) 2019/2020
The ecodesign regulation for light sources and separate control gears sets functional requirements, in its Annex II Table 4, for LED and OLED mains light sources (MLS) at full load:
- Pst LM ≤ 1.0
- SVM ≤ 0.4
The SVM limit was phased. Amending Regulation (EU) 2021/341 allowed SVM ≤ 0.9 until 1 September 2024; the 0.4 limit has applied since that date. The SVM requirement also carries exceptions: “HID with Φuse > 4 klm”, and light sources intended for outdoor, industrial or other applications “where lighting standards allow a CRI < 80”, when a clear indication is shown on the packaging and documentation. The HID wording sits in the same table row; the flicker limits themselves apply only to LED and OLED mains light sources.
Three details are regularly misread:
- The limits are written for mains light sources. An integrated luminaire connected directly to the mains, where the LED module cannot be removed without damage, is treated as a light source itself and falls under them. A non-mains LED module sold with a separate driver is a different case: the regulation’s flicker rows do not name non-mains sources or control gear. The flicker behaviour of that combination is therefore something the buyer must specify, not something the regulation guarantees. Confirm the classification with your test laboratory before relying on it.
- The limits apply at full load. The regulation does not set a flicker limit at dimmed levels. Section 4 explains why that gap matters.
- Compliance is declared, not certified. Market surveillance checks declared values after the product is on the market. A declaration of conformity is not a test report, and the test report should name the driver that was fitted.
United States: IEEE 1789-2015 and California
There is no federal flicker limit for general lighting in the United States. Two references dominate specifications:
- IEEE 1789-2015 is a recommended practice, not a regulation. Above 90 Hz it defines a low-risk line at percent flicker ≤ 0.08 × frequency and a no observable effect level (NOEL) line at percent flicker ≤ 0.0333 × frequency; below 90 Hz the lines tighten to 0.025 × frequency and 0.01 × frequency. At a 120 Hz ripple frequency, that is 9.6% for low risk and roughly 4% for NOEL. The low-risk line has no depth limit above 1,250 Hz and the NOEL line none above 3,000 Hz.
- California Title 24 Joint Appendix JA8 (2025 edition, in effect since 1 January 2026), with its JA10 test method, requires “reduced flicker operation”: percent flicker below 30% at frequencies below 200 Hz, measured at 100% and at 20% light output. JA8 applies to light sources qualifying as high-efficacy under the code, mainly in residential applications — it is not a general limit on every commercial luminaire sold in California, but it is a useful, published test template.
The California requirement is notable because it tests at a dimmed level. The IEEE 1789 lines are notable because they are stricter than the regulatory numbers most suppliers quote — a product that meets a 30% limit at 120 Hz sits roughly three times above the IEEE 1789 low-risk line.
4. The Dimming Problem: Where Compliant Products Start to Flicker
Flicker problems frequently surface at deep dimming levels — the operating points that full-load regulatory tests do not cover. The reason is how drivers reduce light.
Pulse-width modulation (PWM). The driver switches the LED current fully on and off, and varies the duty cycle to reduce average output. Modulation depth approaches 100% by construction. Whether that is acceptable depends almost entirely on the switching frequency: PWM at a few hundred hertz is a strong stroboscopic source; PWM at several kilohertz, above the IEEE 1789 depth-independent frequencies, removes most of the stroboscopic risk — though full-depth PWM in the low-kilohertz range can still produce a visible phantom array.
Constant-current reduction (CCR, or analogue dimming). The driver lowers the DC current. The absolute ripple current often stays roughly constant while the average current falls, so modulation depth — percent flicker — rises as dimming deepens, still at twice mains frequency, which is the band SVM is designed to catch.
Hybrid schemes. Many drivers use CCR down to a threshold and switch to PWM below it. The flicker profile can change sharply at that crossover.
Phase-cut control. Leading-edge and trailing-edge dimmers chop the mains waveform before it reaches the driver. Driver–dimmer compatibility becomes a flicker variable in its own right, and a product tested on a laboratory power supply may behave differently on a site dimmer with a different minimum load.
The consequence for procurement is direct. A full-load Pst LM and SVM declaration says nothing about a scene set at 10% in a restaurant or a daylight-harvesting zone trimmed to 25% in an office. If the installation will run dimmed for most of its operating hours, the dimmed profile is the relevant performance, and it has to be requested explicitly.
5. Application Sensitivity: Where Flicker Costs Money
Not every space needs the strictest limits. Specifying NOEL-level performance everywhere raises driver cost without benefit; specifying nothing leaves the riskiest spaces exposed. A tiered approach works better.
| Application | Main risk | Suggested target to request |
|---|---|---|
| Circulation, storage, back-of-house | Low; short occupancy | Regulatory minimum in the target market |
| Open-plan offices, classrooms | Long occupancy, computer work, headaches and eye strain reported by occupants | SVM ≤ 0.4 and Pst LM ≤ 1.0 at full load and at the lowest dimmed level used; IEEE 1789 low-risk at the dominant frequency |
| Workshops, warehouses, food processing | Stroboscopic effect on rotating machinery — a safety issue, not a comfort issue | IEEE 1789 NOEL at the dominant frequency, or no significant modulation in the stroboscopic band, across the dimming range |
| Hospitality and retail scenes run at low output | Visible shimmer at deep dimming; camera banding in content filmed by guests and shoppers | Dimmed-level test at 10–20% output with the actual control system |
| Sports halls, broadcast and filmed spaces | Camera banding, slow-motion replay artefacts | Project-specific: take the frame-rate and shutter range from the broadcast brief and request modulation data across it |
| Healthcare, special-needs education | Sensitive occupant groups | IEEE 1789 NOEL across the dimming range |
The workshop row deserves emphasis: a rotating chuck that appears stationary under strong 100 Hz modulation is a machine-guarding hazard, so stroboscopic performance belongs in the safety part of the specification.
6. Common Misconceptions
“Flicker-free” means no modulation. It means whatever the supplier decided it means. Some use it for “below the EU limits at full load”, some for “passes IEEE 1789 low-risk”, and some without any measurement at all. Treat the phrase as a question, not an answer.
A good driver guarantees a good luminaire. Flicker is a system property. The same driver on a different LED load, or at a different output point, produces a different waveform. The report must cover the luminaire as shipped.
Phones reveal all flicker. A phone camera shows banding from some modulation frequencies and hides others depending on shutter speed. It is a useful red flag at a showroom visit and not an acceptance test. The camera cannot produce a Pst LM or SVM value.
Passing at 230 V means passing everywhere. Pst LM, SVM and percent flicker all change between 230 V / 50 Hz and 120 V / 60 Hz, and network voltage fluctuations on site add flicker a stable-supply lab test never shows. A product sold into 120 V / 60 Hz and 230 V / 50 Hz markets needs results for both conditions.
7. Specification Checklist
Replace “flicker-free” with requirements that can be verified.
State the metrics and the methods. Require Pst LM per IEC TR 61547-1 and SVM per IEC TR 63158, plus percent flicker with the frequency of the dominant modulation component.
State the operating points. Require results at 100% output, at 20% output, and at the lowest level the control scheme will actually use. For tunable-white products, add results at the warmest, coolest and a mid-range CCT setting, where both channels are dimmed at the same time.
Name the control path. Specify the dimming protocol — 0-10V, DALI, phase-cut or wireless — and require the test to use it. For phase-cut products, list the dimmers that were verified for compatibility.
Tie the report to the shipped configuration. The report should identify the driver model and the LED module. A driver substitution during production is a change that invalidates the flicker evidence, and the purchase contract should require notification before it happens.
Set tiered targets by space. Use the table in Section 5 rather than one project-wide number. Put stroboscopic limits for machine areas in the safety clauses.
Verify on samples, not only on paper. Check pre-shipment samples with a handheld flicker meter at full and dimmed output. It catches the most common failure: a production batch fitted with a different driver than the tested sample.
8. FAQ: Common Questions on LED Flicker
Is SVM ≤ 0.4 good enough for machine areas? SVM is calibrated so that 1.0 is the threshold of visibility for an average observer; 0.4 therefore keeps stroboscopic visibility well below that threshold under the tested conditions. The limit applies at full load only. In machine areas, confirm the value at every dimmed level the controls will use, and treat any rotating-equipment area as a safety review item.
Can existing installations be improved without replacing fittings? Sometimes. Where the fitting uses a separate, accessible driver, replacing the driver with a lower-ripple or high-frequency PWM type can resolve the problem, provided the replacement is compatible with the LED module and the control system. Where the driver is integrated and sealed, the whole fitting usually has to be changed; driver accessibility is therefore worth including in the specification.
Conclusion
Flicker is one of the few lighting quality parameters that can be fully compliant on paper and still fail the space. The EU limits of Pst LM ≤ 1.0 and SVM ≤ 0.4 apply to mains light sources at full load; the IEEE 1789 recommended practice sets stricter frequency-dependent lines; California tests at 20% output. None of these alone describes how a luminaire behaves at the dimmed levels where many commercial spaces operate for much of the day.
The practical shift is small: stop accepting a single adjective, and request measured Pst LM, SVM and percent-flicker values at full and dimmed output, tied to the driver and control protocol that will actually ship.
If you are preparing a lighting package with dimmed scenes or rotating machinery, send us the fixture schedule and dimming plan — we will reply with the flicker test points and sample checks to request from each bidder before you commit to the order.
Sources
- Commission Regulation (EU) 2019/2020 of 1 October 2019 (EUR-Lex CELEX 32019R2020), ecodesign requirements for light sources and separate control gears, Annex II Table 4 (functional requirements: Pst LM ≤ 1.0 and SVM ≤ 0.4 at full load for LED and OLED mains light sources), as amended by Commission Regulation (EU) 2021/341 of 23 February 2021 (CELEX 32021R0341; SVM ≤ 0.9 until 1 September 2024)
- IEC TR 61547-1:2020, Equipment for general lighting purposes — EMC immunity requirements — Part 1: Objective light flickermeter and voltage fluctuation immunity test method
- IEC TR 63158:2018+AMD1:2021, Equipment for general lighting purposes — Objective test method for stroboscopic effects of lighting equipment
- IEC 61000-4-15, Testing and measurement techniques — Flickermeter — Functional and design specifications
- IEEE Std 1789-2015, IEEE Recommended Practices for Modulating Current in High-Brightness LEDs for Mitigating Health Risks to Viewers (low-risk line 0.08 × f up to 1,250 Hz; NOEL line 0.0333 × f up to 3,000 Hz)
- California Energy Commission, 2025 Building Energy Efficiency Standards (Title 24, Part 6), Reference Appendices JA8 and JA10, effective 1 January 2026 (reduced flicker operation: percent flicker below 30% at frequencies below 200 Hz, at 100% and 20% output)
- U.S. Department of Energy, SSL Technology Fact Sheet: Flicker, PNNL-SA-94791, March 2013 (percent flicker and flicker index definitions)
Next scheduled review: March 2027, or earlier if Regulation (EU) 2019/2020 is replaced under the Ecodesign for Sustainable Products Regulation or IEEE 1789 is revised.