I have some questions about how it is possible to dim an LED using PWM. I saw some videos and explanations of how the duty cycle affects the brightness and some methods to effectively dim a LED with different components, but I´m missing some detail here.

What can be summed up is that for a duty cycle of n% the brightness of the LED will be n% of what it would be with the 100% duty cycle and this is explained with the average values and total energy in time that is being provided.

If we had two cases, one simple signal-resistor-LED circuit with 50% duty cycle and one with 20% duty cycle (the only thing that varies is the duty cycle,) and we could inspect an instant of time in the positive cycle for each case we should get the same values of voltage and current since we aren't using energy storage devices. It is also said that the brightness of the LED depends on the amount of current provided, hence we should get the same brightness during the positive cycle even if one lasts longer than the other. If it is receiving the same current, where does the dimming happens? If the answer is in the way our vision works, does that mean a specialized tool would perceive the same brightness?

  • 2
    \$\begingroup\$ "brightness of the LED depends on the amount of current provided" Yes - average current. Our eyesight perceives brightness non-linearly. Half-brightness corresponding to 50% PWM on/off seems much more than half to our eye. For low on/off ratio like 10%on, 90%off - the LED seems brighter than you'd expect. I'm assuming PWM frequency is higher than roughly 50 pulses-per-second. Most of the eye-apparent brightness change occurs at low on/off ratios. \$\endgroup\$
    – glen_geek
    Mar 11, 2023 at 15:09
  • 2
    \$\begingroup\$ There's average brightness (averaged over time) and peak brightness. What our eyes perceive is average brightness. If your theoretical 'specialized tool' would measure actual peak brightness then yes, you would be correct that it would show the same value for any duty cycle. \$\endgroup\$
    – floppydisk
    Mar 11, 2023 at 15:40
  • \$\begingroup\$ Because light isn't linear and our eyes can adjust between broad daylight (100,000 lux), shade (20,000 lux), overcast (1000 lux), or moonlight (1 lux). So increasing duty cycle from 50% to 100% isn't much. \$\endgroup\$ Mar 12, 2023 at 0:12
  • \$\begingroup\$ @glen_geek : Can you source references to anything which might detail persistence of vision is much stronger when the peak is much greater than the average? \$\endgroup\$
    – kando
    Jun 2, 2023 at 1:41
  • 1
    \$\begingroup\$ @kando For PWMing LEDs, people.csail.mit.edu/jstraub/blog/How-to-dim-a-LED-for-humans suggests that half-brightness occurs at about 8% PWM in the 2nd-to-last graph. \$\endgroup\$
    – glen_geek
    Jun 2, 2023 at 5:04

2 Answers 2


You are mostly correct: for reasonable PWM frequencies, a monochrome LED is at full brightness for part of the cycle and off in the other part. The human eye cannot perceive the flicker beyond a certain frequency, so we perceive the LED as a dim source.

During certain rapid eye movements (e.g. saccades) humans can perceive a little bit of this flicker. In my personal experience, this shows up as bands or dots of light along the movement path of the light.

A fast photodiode can also detect the flicker, and fast photodiodes/phototransistors/etc are used for high speed optical communication well above the human eye's flicker fusion threshold.

Fiber optic internet makes use of this principle to achieve truly remarkable speeds, with speeds of hundreds of gigabits per second in single mode fibers, and remarkably more when multiple modes or wavelengths are used. However, the fastest communication methods use laser diodes and specialized biasing/modulation schemes rather than PWM with an LED.

The exception here is white LEDs. They contain a phosphor (1), or a solid material that absorbs optical energy and emits it at a different frequency. This material has a bit of persistence due to the chemical energy it briefly stores, meaning that it may continue to glow even during the period when that LED is off, if the PWM frequency is fast enough.

(1) despite the name, they need not contain phosphorus

  • \$\begingroup\$ I'll calmly look further into the concept of "Flicker fusion threshold", but if I get it right dimming by PWM is only posible by our vision's nature, so if we were to share information based on brightness instead of light flicking (lets propose a simple e.g as max brightness is 1 and medium is 0) would any sensor perceive this brightness change?, if not, which parameter is the key in those measurings? \$\endgroup\$
    – Seidenki
    Mar 11, 2023 at 15:07
  • \$\begingroup\$ @CristianRugeles the property you are looking for is bandwidth. If the sensor bandwidth is too low, or if you intentionally use a low pass filter downstream of the sensor, you will get an averaged intensity value when the PWM is fast enough. A fast sensor without a filter can still decode your proposed mechanism: you just need to look at the length of a bright pulse rather than just the brightness value. \$\endgroup\$
    – nanofarad
    Mar 11, 2023 at 15:14
  • \$\begingroup\$ @CristianRugeles - If the sensor is fast enough, then he is able to make this visible for you. We have very fast optical sensors to make data transmission and this sensors have to be so fast that he can recognize a blinking with several Megahertz. You only need the right sensor for the job. \$\endgroup\$
    – MikroPower
    Mar 11, 2023 at 15:19
  • \$\begingroup\$ Thanks to both of you, would you mind posting any source that could be of help (books or videos, any kind works), I'm having trouble to find appropiate ones in my language and looking for them in english is not as easy as I'd expect \$\endgroup\$
    – Seidenki
    Mar 11, 2023 at 15:27
  • 1
    \$\begingroup\$ @Seidenki I don't know of any resources off the top of my head, because I studied many years ago and no longer have all my texts. I'll think about it and will update if anything comes to mind. For the basics, Art of Electronics is a good book, and it'll give you a foundation to understand more advanced documents like research papers, datasheets, manufacturer app notes, etc. \$\endgroup\$
    – nanofarad
    Mar 11, 2023 at 15:33

To dim an LED with PWM you do this:

  1. Reduce the current of the LED with a resistor to the maximum.
  2. Then you can use PWM to make it blink very fast, so that your eye can not recognize the flickering of the light.

With a speed of 1Hz you can clearly see it if the LED is to 50% on (500ms) and 50% off (500ms), then you can see the blinking.

If the speed is faster, then the eye will catch photons for a while during several on-off phases. (But possibly in the brain is a filter mechanism too.)

The PWM of car-backlights is often too slow, so that people can see it.

The photoreceptors of the retina are sluggish, they capture photons for a certain period of time and after a certain period of time they evaluate the amount that has reached the receptor. In the eye, this is a process that is triggered by mechanical changes in molecules and is then passed on electrically via the optic nerves. This is similar to what happens on a normal camera sensor when photons hit it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.