This page is partly based in a post I wrote in 2017 and also includes some material from an article that was published in the UV4Plants Bulleting titled Using LEDs: drivers and dimming. Please, consult this earlier article for additional information about the different approaches used to adjust the light output of LEDs.
Mains AC frequency
Incandescent (tungsten) lamps, including tungsten-halogen lamps as well as many fluorescent lamps flicker at twice the frequency of the mains power. This flickering is not on/off but between brighter and less bright output. The AC power follows a sinusoidal wave shape and there is enough delay in the cooling of the tunsten fillament or in the decay of the fluorescence of the coating to maintain some light output at the crossover when voltage is at zero.
With a light sensor with a fast response and an ordinary oscilloscope the pulses can be observed and the frequency and duty cycle precisely measured.
Pulse Width Modulation
A commonly used approach to dimming of LEDs is to very quickly switching them on and off. This is a square or rectangular wave shape, making it more obvious at low frequency. However, if the frequency is high enough, eyes do not perceive the flicker. Above about 100 Hz humans see light as constant even when pulsed on and off. Some animals do perceive higher frequencies than humans.
In the post Small fill/video LED lights revisited I mention that of three video LED lights I tested two used PWM. However, the frequencies used are 49 kHz and 146 kHz. When PWM dimming is used for general illumination, the frequencies used vary between 100 Hz and 1 kHz.
Mechanical shutters
Different types of shutters are in use. In large format cameras and in some medium format cameras shutters are frequently located in the lens and simply open and close. In most medium, full-frame and smaller formats in cameras with interchangeable lenses the shutter is located very near the focal plane, that is just in front of the film or sensor. These shutters, when set to relatively low shutter speeds, actually open and close. However, high shutter speeds are obtained with a slit that travels across the focal plane, and the exposure time is adjusted by the changing width of the slit keeping constant the speed at which the curtain with the slit travels. The consequence is that at fast shutter speeds different parts of the image are captured at slightly different times. Obviously, if the illumination varies during the time when the slit moves, different parts of the image receive different amounts of light resulting in visble banding. If the shutter speed is such that a slit is not used, instead of banding, the flickering results in succesive images in a series being inconsistent, with some ligher and others darker. The travel time of the slit across the photograph depends on the camera. Something like 2.4 ms to 15 ms can be guessed based on flash synchronization speeds.
One way to avoid uneven exposure is to use slow-enough shutters speeds and select speeds that match a whole multiple of the duration of a cycle. So, if the frequency is 100 Hz, using 1/100 s, 1/50 s, 1/25 s, etc., can help avoid banding. At low shutter speeds, enough cycles are included that any possible values works just fine.
Another artefact introduced, but unrelated to illumination, is the distortion of the shape of fast moving objects. In this case the object moves during the travel of the slit, an thus appears deformed in the photograph.
CMOS image sensors
Most modern digital cameras have a mechanical shutter and an electronic shutter. When the mechanical shutter controls the exposure time, the sensor is active for a longer time than the exposure length and consequently the discussion on the previous section applies. When the electronic shutter is in use, the mechanical shutter remains open, and the exposure time is controlled electronically by the integration time during which photons are collected.
CMOS digital sensors, except in exceptional cases, work so that pixels are read one row at a time. We have the equivalent of a moving slit, but a very narrow one. The readout speed describes the time it takes to read the data from the every pixel once. We face again the problem not all the image is captured simultaneously, and this, as in the case of the slit, can introduce both banding and deformation of fast moving objects. Until recently, readout times for CMOS sensors were relatively long at around 50 to 70 ms. Newer sensors, have readout times as low as 5 ms. The just announced newest top-of-the-line Sony camera has a global electronic shutter, meaning that the whole image is captured simultaneously.
Measuring the readout speed
If we have as pulsing light source at a known frequency, counting the number of bright bands we see on the image and multiplying this number by the duration of one cycle gives the readout time. If we assess this number to a fraction of a fraction of a band we can get a very good estimate. In the old times a TV set with a scanning rate of 50 Hz or 60 Hz was used to test camera shutters. With a PWM-dimmed LED lamp the bands are very well defined. The values I measured for three cameras using a LED lamps with PWM dimming at 166 Hz are shown in
| Camera | Year | Resolution | Readout time |
|---|---|---|---|
| E-M1 | 2013 | 16.3 MPix | 63 ms (1/16 s) |
| E-M1 II | 2016 | 20.4 MPix | 15.6 ms (1/64 s) |
| OM-1 | 2022 | 20.4 MPix, stacked BSI | 7.8 ms (1/128 s) |
