Cinematography: safe frame rates for PWM LEDs

Following on from previous posts on SFX electronics:

Here’s some data and discussion on reducing artefacts like flickering if using low-frequency PWM LED lighting on-camera.

Brief summary

How to avoid flickering and artefacts from PWM LED lighting?

Some options, best to worst:

  • Don’t use PWM lighting (ideal world)
  • Use high frequency PWM lighting (>= 2kHz lower end; ideally 20kHz for industry high-frequency) – no artefacts, love it
  • Be clever when using lower frequency PWM lighting to minimise artefacts

The rest of this article is about the last option.

Data

Here’s some info on avoiding (in theory) artefacts like flicker, given some LED tech choices.

Base acquisition rates of 24, 25, and 50 fps are common, so for each of those I present an ideal shutter angle and cranked frame rates1 that work best for removing or reducing flicker/artefacts.

24 fps base

Assuming we’re using shutter angle of 172.8°:

LED PWMExample device/classBase rate OK?Ideal crank FPSMaybe crank FPS
400 Hzold WS2812/WS2812B❌--
580 Hz*APA102 global brightness❌--
1 kHzsome SK6812 variants✅30-
1.2 kHzSK6812✅-30
2 kHzWS2813/14, newer WS2812✅30, 40, 48, 6050
2.5 kHzWS2801 class✅30, 40, 48, 50, 6072
4.7 kHzSK9822✅4830, 40, 50, 60, 72, 96, 100, 120, 144
8 kHzGS8208 / CS8812 class✅30, 40, 48, 60, 96, 120, 24050, 72, 100, 144, 200
10 kHzWS2816B class✅30, 40, 48, 50, 60, 96, 100, 120, 200, 24072, 144
19.2 kHzAPA102 RGB PWM✅48, 72, 96, 14430, 40, 50, 60, 100, 120, 200, 240
20 kHzgeneric high-frequency PWM✅30, 40, 48, 50, 60, 96, 100, 120, 200, 24072, 144
26 kHzHD107S class✅30, 40, 48, 60, 96, 120, 24050, 72, 100, 144, 200
27 kHzHD108 class✅30, 40, 48, 60, 72, 96, 120, 144, 24050, 100, 200

Comments: At 24 fps there are no good PWM LED options below 1 kHz PWM.

Ideal crank FPS means at least 15 nominal PWM cycles2 per exposure3 and an integer number of PWM cycles per exposure.

Maybe crank FPS means at least 15 nominal PWM cycles per exposure, but not an integer number of cycles.

* APA102’s ~580 Hz figure refers to global-brightness modulation rather than its normal RGB PWM.

25 fps base

Assuming we’re using shutter angle of 180°:

LED PWMExample device/classBase rate OK?Ideal crank FPSMaybe crank FPS
400 Hzold WS2812/WS2812B❌--
580 Hz*APA102 global brightness❌--
1 kHzsome SK6812 variants✅-30
1.2 kHzSK6812✅30, 40-
2 kHzWS2813/14, newer WS2812✅40, 5030, 48, 60
2.5 kHzWS2801 class✅5030, 40, 48, 60, 72
4.7 kHzSK9822✅5030, 40, 48, 60, 72, 96, 100, 120, 144
8 kHzGS8208 / CS8812 class✅40, 50, 100, 20030, 48, 60, 72, 96, 120, 144, 240
10 kHzWS2816B class✅40, 50, 100, 20030, 48, 60, 72, 96, 120, 144, 240
19.2 kHzAPA102 RGB PWM✅30, 40, 48, 50, 60, 96, 100, 120, 200, 24072, 144
20 kHzgeneric high-frequency PWM✅40, 50, 100, 20030, 48, 60, 72, 96, 120, 144, 240
26 kHzHD107S class✅40, 50, 100, 20030, 48, 60, 72, 96, 120, 144, 240
27 kHzHD108 class✅30, 50, 60, 10040, 48, 72, 96, 120, 144, 200, 240

Comments: At 25 fps there are no good PWM LED options below 1 kHz PWM.

Ideal crank FPS means at least 15 nominal PWM cycles2 per exposure3 and an integer number of PWM cycles per exposure.

Maybe crank FPS means at least 15 nominal PWM cycles per exposure, but not an integer number of cycles.

* APA102’s ~580 Hz figure refers to global-brightness modulation rather than its normal RGB PWM.

30 fps base

Assuming we’re using shutter angle of 216°:

LED PWMExample device/classBase rate OK?Ideal crank FPSMaybe crank FPS
400 Hzold WS2812/WS2812B❌--
580 Hz*APA102 global brightness❌--
1 kHzsome SK6812 variants✅40-
1.2 kHzSK6812✅40, 48-
2 kHzWS2813/14, newer WS2812✅40, 48, 50, 6072
2.5 kHzWS2801 class✅50, 60, 10040, 48, 72, 96
4.7 kHzSK9822✅6040, 48, 50, 72, 96, 100, 120, 144
8 kHzGS8208 / CS8812 class✅40, 48, 50, 60, 96, 100, 120, 200, 24072, 144
10 kHzWS2816B class✅40, 48, 50, 60, 100, 120, 200, 24072, 96, 144
19.2 kHzAPA102 RGB PWM✅40, 48, 60, 72, 96, 120, 144, 24050, 100, 200
20 kHzgeneric high-frequency PWM✅40, 48, 50, 60, 96, 100, 120, 200, 24072, 144
26 kHzHD107S class✅40, 48, 50, 60, 100, 120, 200, 24072, 96, 144
27 kHzHD108 class✅40, 50, 60, 72, 100, 120, 20048, 96, 144, 240

Comments: At 30 fps there are no good PWM LED options below 1 kHz PWM.

Ideal crank FPS means at least 15 nominal PWM cycles2 per exposure3 and an integer number of PWM cycles per exposure.

Maybe crank FPS means at least 15 nominal PWM cycles per exposure, but not an integer number of cycles.

* APA102’s ~580 Hz figure refers to global-brightness modulation rather than its normal RGB PWM.

Gory details

Here’s how the data above was derived:

We start with a base frame rate for shooting (say 24, 25, or 30).

From that we find out a shutter angle that gives a whole number of PWM cycles per exposure3 (which results in no flicker).

From there we work out the theoretical maximum crank fps that puts us at roughly 15 PWM cycles per exposure (rule of thumb).

And then we work out frame rates between base rate and the maximum that also give a whole number of PWM cycles in an exposure (we want to keep slow-motion footage on the table).

Generally speaking, what things contribute to flicker?

All of the following contribute to greater chance of getting flicker/artefacts:

  • lower shutter angle (because it results in lower exposure)
  • higher frame rate (because it results in lower exposure)
  • lower LED PWM duty cycle (i.e. lower brightness) for any one of the LEDs (because the average light is focused into a smaller amount of time)

But these only apply if we’re not using one of the ideal frame rates + shutter angle given in the data earlier.

If you’re using an ideal frame rate + shutter angle, in theory you’re golden; even reducing the brightness (duty cycle) down to very low values won’t cause artefacts.

A word on clamp/blowout

LEDs themselves usually won’t be what camera will expose for, and given their intensity, the core portion of LED light on the image can be blown out (clamped).

This can actually hide some artefacts because oscillation might be all above the clamp level and so hidden. But this shouldn’t be relied on in any way. In particular if there’s any spatial falloff of LED light (e.g. through cosmetics layers, reflections, flares from anamorphic lens) you might still see artefacts. Long side-flares from an anamorphic lens can look really nice but they might make any artefacts more obvious too.

Camera testing of low-frequency PWM lighting

This is just general good sense, but: test all conditions you will, or may want to, use. This might include:

  • frame rates (including any crank for slow-mo)
  • shutter angles
  • incidental light conditions (day? night?)
  • presence of other lighting equipment

As a final tweak consider ClearScan/ECS/Synchro Scan (etc.) tweaking to fine-tune out any flicker if possible. Even if the maths is all correct (see earlier data), low frequency PWM LEDs might not nail their stated frequency: your 400Hz PWM might actually be 397Hz, and so on.

What about readout time?

Readout is the time taken to acquire an image on a rolling-shutter camera (which is most of them).

Readout time does not itself contribute to LED flicker per se.

It does affect how any PWM artefacts are stretched down over the image: the number, spacing and movement of visible bands.

If you do have artefacts, lower readout times may be preferable: this will result in more gradual gradient. Higher readout times will look more like obvious banding.

Readout values for some common digital cinema cameras:

CameraReadout (ms)
Sony VENICE 23.0
Apple iPhone 15 Pro4.7
ARRI ALEXA Mini LF7.4
ARRI ALEXA 357.6
RED V-RAPTOR 8K VV8.0
Sony FX68.7
Blackmagic URSA Mini Pro 12K15.7
Sony BURANO18.9
Blackmagic Pocket Cinema 6K19.8

  1. cranking usually used to create slow-motion footage ↩︎

  2. at least 15 PWM cycles per exposure seems to be a widely accepted rule-of-thumb for avoiding artefacts ↩︎ ↩︎ ↩︎

  3. exposure means sensor exposure time per frame: \( T_E = \frac{1}{f_F}\frac{\theta}{360^\circ} \), where \( f_F \) is fps and \( \theta \) is shutter angle. Example: 24 fps with 180° shutter angle means exposure time of 1/48 sec. ↩︎ ↩︎ ↩︎ ↩︎