1. Introduction — Dimming Is Not Just Lower Current

Dimming an LED is apparently simple — reduce the current and it gets dimmer. But a useful, reliable dimming implementation must preserve a wide dimming range, avoid a color shift, keep the light flicker-free to the eye and to cameras, maintain constant chromaticity, and not wreck the driver’s efficiency. Two fundamentally different techniques exist: analog (continuous current) dimming and PWM (pulsed current) dimming, plus hybrid combinations used by smart bulbs, drivers and camera-sensitive applications. This guide explains the physics of LED dimming, the two methods and their trade-offs, the flicker and contrast-ratio constraints, how dim-to-off works, the rail/brightness collapse from reduced current, and how to integrate dimming into constant-current LED drivers. It complements the driver topologies and LED driver design guides on this site.

2. Why LEDs Behave Differently When Dimmed

An LED’s luminous flux tracks current with a slightly sub-linear curve, and — critically — its forward voltage and emission spectrum shift with current and junction temperature. Lowering the current typically shifts the peak wavelength (a small color shift, noticeable on white LEDs as a change in warm/cool tint) and, because the current-to-flux gain falls at low current, creates a “dead zone” near zero where tiny current variations are invisible or where the light is dominated by leakage. Compounding this, reduced current does not linearly reduce perceived brightness to the eye (perception follows an approximate power law, so 50% photometric flux is NOT perceived as 50% brightness). A good dimming system therefore sets the output by a gamma-like curve that maps the user’s 0–100% to the driver’s 0–100% with perceptual linearity.

3. Analog (Continuous-Current) Dimming

Analog dimming reduces the DC average current continuously. In a linear driver this is achieved by adjusting the current-set resistor or the control voltage; in a switching (buck/boost) constant-current driver, by lowering the current reference into the current-sense comparator. Roughly I_ref = V_ref / R_sense. The advantages: smooth, fully continuous, no flicker, no switching-induced artifacts, high contrast ratio if the reference can go low, and trivially simple to implement (a DAC or a filtered PWM into the reference pin). The drawbacks: at low current the color shift is worst, efficiency can drop because the driver’s fixed losses dominate, and the practical minimum current is bounded by the reference voltage accuracy and comparator offsets — a driver that regulates to ±2% at full current will show large percentage error at 1% of nominal (the “can’t get truly dark” problem).

4. PWM Dimming

PWM dimming switches the LED string on and off at a fixed frequency while holding the on-state current at the full nominal value (the point of best chromaticity and efficiency) and controlling brightness by the duty cycle. Because LEDs track current nearly instantly (no slow thermal lag), the on-state light is exactly full-flux and the average is:

Φ_avg = Φ_100% · D,   with D = t_on / (t_on + t_off)

PWM yields excellent dim-to-zero (the LED simply stops switching), negligible color shift (the on-state current never changes), and a wide dynamic range set by the minimum achievable pulse width: a 32 kHz PWM with a 100 ns minimum-on pulse theoretically spans 1:3200, though the driver’s switching speed and the LED’s own turn-on/turn-off delay reduce that. The price is flicker: if the PWM frequency is inside the perceptible range (typically below ~90 Hz for direct perception, or into the stroboscopic range for cameras and for moving objects — the so-called “phantom array” and high-speed-camera rolling-shutter bands), the user sees or the video sees artifacts. In practice modern dimming systems use PWM either at a mandatory >3.125 kHz (IEEE 1789 recommended to avoid flicker) or even higher, or they use analog-style modulation at the mains-free frequencies.

5. Flicker — the Hidden Quality Metric

Flicker is measured by the flicker index (cumulative) and the percent flicker (modulation depth): % flicker = (max − min)/(max + min) × 100%. A 100% duty-cycle analog dimming has ~0% flicker; PWM at 100% duty also ~0%; but PWM below 100% at a low frequency can be 100% deep flicker. Human perception fades above roughly 90–200 Hz for most observers but the 30–90 kHz range is required to be safe for cameras and for the “temporal light modulation” (TLM) sensitivity of newer standards. For eye-comfort the modern recommendation for general lighting is PWM frequency ≥ 1.25 kHz (IEEE 1789 low risk to 3.125 kHz), and significantly higher on printed/scanning systems. The switching frequency of the driving converter — whether it is coherent with the PWM dimming (burst dimming) or independent — determines whether the driver’s own ripple modulates the light. In burst-mode PWM, the driver is turned off entirely during the off-phase: maximum contrast but maximum current stress on the LED and its power dissipation must be checked against the pulse cycles by the thermal system.

6. Comparison of the Two Methods

Property	Analog (current)	PWM
Color / chromaticity	Shifts at low current	Stable (current fixed)
Dim-to-dark depth	Limited by ref accuracy	Excellent (min pulse width)
Flicker	None	Depends on f_PWM & depth
Driver efficiency at low dim	Drops (fixed losses)	High per-pulse
EMI / audible noise	Lower	PWM coupling into coils
Control complexity	Simple (DAC/ref)	Timer/PWM + sync

In practice the two are combined: encode the user’s requested level in a PWM waveform, low-pass filter it to an analog DC reference (analog dimming with a PWM interface and no flicker), or use it to modulate the driver’s reference directly (PWM dimming). Smart bulbs typically do low-frequency PWM with a high-frequency boost converter for efficiency and depth, while high-CRI retrofit drivers favor analog or high-frequency PWM.

7. Worked Example — 100 W Outdoor LED Driver with PWM Dimming

Target: a 100 W constant-current driver (buck topology, 1 A) with PWM dim-to-1% down to 0.1% without flicker to a high-speed camera.

• Choose f_PWM = 25 kHz (inside IEEE 1789 low-risk for photobiological purposes, camera-safe): t_on at 1% = 0.01/25 kHz = 400 ns; the driver must respond within < 100 ns to reach min pulse — check the LED module's turn-on delay.
• The driver current-sense comparator toggles the buck switch at its switching frequency (say 100 kHz); PWM dimming in this case uses a separate enable that gates the buck at 25 kHz (burst mode) — the off-phase is long so the output capacitor charge decays; the LED current ramp must settle to full current within the on-phase or the effective flux is lower than D predicts.
• Compute the LED’s average and peak power: average = 1 A × 37 V × D; at D = 0.01, average power = 0.37 W while the LED sees 37 W pulses — the LED die must survive the thermal cycle; the power calculator gives the exact average and junction rise via the thermal resistance (see the LED thermal guide).
• For the dim-to-1% analog option, keep the reference accurate: a 12-bit DAC driving a 0.1% current-sense network holds ±1% of set point down to the DAC’s lowest code — but the comparator offset at 10 mA needs a trim, otherwise light “stops” at ~0.3%.
• Verify with a photodiode + scope: % flicker at 25 kHz with 1% depth shows a clean square; raise to 50 kHz if the camera banding appears.

For the resistor-based current setting, the LED resistor calculator sizes the current-set resistor and the string resistor drop quickly.

8. Common Mistakes

• Dimming with low-frequency PWM: 500 Hz–2 kHz PWM is visible to the eye/camera and fails modern flicker standards.
• Letting the driver switch frequency alias the dimming: incoherent switching ripples the light despite “clean” PWM.
• Reducing current to analog-dim a white LED to dark: color shifts warm/cool and the minimum is present as a fixed drop — it never really dims to zero.
• Ignoring LED turn-on delay: the min on-pulse shorter than the LED’s switching delay produces no measurable light at small duty cycles.
• Forgetting the thermal duty-cycle stress: PWM at deep dim heats the junction in pulses; the average-vs-peak power trade-off must enter the thermal design.
• Using the reverse-gamma missing: responding linearly to the 0–100% command gives a perceptually compressed low end.

9. FAQ

Q: Which dimming is “best”? A: For constant chromaticity and deep dim range, PWM; for flicker-free smoothness with a fixed color point and simple control, analog. Most high-end designs combine both.

Q: Is PWM dimming flicker-safe for cameras? A: Only above the camera’s line/frame rate interaction — a 25 kHz PWM is invisible to the eye and to most 30–60 fps cameras; very fast/shutterless cameras need higher frequencies or analog.

Q: Can I dim by reducing the mains triac? A: Triac/forward-phase dimming chops the AC, which modern constant-current drivers translate into a dim command — but the chopping adds flicker and noise; the driver must filter and map the phase to its own dimming reference.

Q: Why does my LED glow at “0%”? A: The driver’s minimum current reference, comparator offsets, or the LED string leakage/ghost current — solve with a real zero-current shutdown path or a series switch.

10. Conclusion

Dimming is a quality dimension as important as the LED driver’s efficiency. Choose PWM for depth and color stability, analog for smoothness and simplicity, and set the frequency/depth budget against the human eye and the camera. Confirm the current-set and power math with the LED resistor and power calculators, keep the driver’s switching coherent with the dimming, and the fixture will dim beautifully rather than glaringly.