Periodic changes in light flux from light sources and luminaires can cause removable or non-removable artifacts in image recording - e.g., differently bright streaks in photographs or flickering areas in video footage. The human eye perceives an image essentially continuously and Talbot's law applies, i.e. above the fusion frequency, one registers only the mean value of the light quantity.

However, cameras and camcorders work with time sampling, i.e. they capture light for a certain time interval given by the opening of the shutter.

The rolling shutter works on the principle of successive reading of the individual rows of the matrix of the exposed image sensor. The individual rows of the sensor are thus exposed with a certain delay. The image defects caused by this shutter are the skewing of a moving object (e.g. a vehicle), the bending of rotating parts (e.g. a propeller), the dropping out of a part of the object (a propeller, a flash in the sky), the appearance of differently bright or coloured bands in passing light - see Fig. 1, where due to the gradual exposure of individual rows the shutter has detected different phases of the light flow. A mechanical analogue of the rolling shutter is the slit shutter of a camera, which can lengthen or shorten images of moving objects.

Fig. 1. Photograph of a luminaire with streaks caused by flickering fluorescent lamps with a dimmer; taken with a Nokia 6303i phone with a scrolling electronic shutter, in reality the scene is without streaks

The global electronic shutter is implemented directly on the image sensor in the form of masked pixels located next to each exposed pixel. Some sensors have half of the matrix masked. The disadvantages may be higher cost and larger sensor size. The advantage is the synchronous exposure of all pixels. The disadvantages associated with delayed line exposure mentioned for rolling shutter are not present. The readout of the charges from the shaded part is the same as with the rolling shutter, but without the pressure on the transfer rate. Its mechanical equivalent is the central shutter of a camera or the rotary sector shutter of a film camera. Some high end digital camera models also have a mechanical rotary sector shutter. More advanced digital cameras also have a mechanical (central) shutter. The advantage, especially at high resolutions, is the ability to use a simpler sensor with a rolling shutter while avoiding its shortcomings. Some devices give the user the choice of using the mechanical shutter or leaving it open and using only the electronic shutter.

Origin of flicker

When the shutter is open, the light is converted into a charge that is integrated into the individual pixels of the sensor. If this time is an integral multiple of the illumination fluctuation period, the integration will result in a mean illumination value. Therefore, in countries with a 50 Hz mains frequency it is advisable to choose camera exposure times of 1/50 and 1/100 s, and in countries with a 60 Hz mains frequency times of 1/60 and 1/120 s [1]. However, these times may be too long for fast moving shots. If the exposure time is not an integer multiple of the illumination fluctuation period, flicker may appear in the footage, which is very distracting, especially if large bright areas are flashing. Think of the shutter function as multiplying the image signal by a rectangular waveform; it is possible to use the goniometric identity of product for the first harmonic components of the two waveforms:

sin (2πf1)sin (2πf2) = ½cos [2π(f1 - f2)] - ½cos [2π(f1 + f2)]

A low differential frequency element is essential for flicker interference.

More advanced cameras can detect the flicker and set the necessary exposure time. Most cameras have a flicker mitigation setting, referred to as Anti-Flicker or Fluorescent frequency, with options of 50 Hz/EU and 60 Hz/US, or off. According to Shannon's theorem, a signal of a certain frequency must be sampled at a frequency at least twice that of the sample in order to be reconstructed from the samples. Figure 2 shows a situation where a sinusoidal signal with a frequency of 100 Hz is sampled with a frequency of 330 Hz (blue) and with a frequency of 120 Hz (red). In the first case, the condition fs > 2f is satisfied and the original signal can be reconstructed from the samples. In the second case, the condition is not met, aliasing occurs and the reconstruction results in a waveform with a difference frequency of 20 Hz.

@obr2@
Fig. 2. Sampling and reconstruction of signals from samples and low frequency generation

Flashing record

If the footage has already been recorded with flicker, software filters with names like Anti-flicker, Flicker-free or DEflicker are available in professional video editing programs that can suppress flicker below a detectable limit. They are also used for time-lapse footage. Flicker in the TV signal can be suppressed in real time by using hardware filters that reconstruct the interfering signal and subtract it from the recorded signal at the cost of delaying the output by several TV frames. They are used, for example, in the transmission of sports matches illuminated only by mains frequency tubes. The design [2] comes from Japan, where the mains frequency is 50 Hz in the eastern part and 60 Hz in the western part, and analogue television was broadcast at 60 half-frames per second.

The temporal artefacts in the film are a record of the stroboscopic phenomenon. In English, it is possible to encounter the wagon wheel effect, which is known, for example, from westerns. Compared to the speed of a stagecoach, the wheels turn more slowly, to the opposite side, or seem at rest. When the stagecoach speeds up or slows down, a comical impression of a change of direction can be created. A similar phenomenon can also be observed in shots of moving tractors. In summer, when the intense light requires a short exposure time, the details of the wheels are clearly visible. See also Fig. 2 for the low frequencies.

When capturing rotating objects with a camera with a rolling shutter, e.g. a propeller, paradoxical situations can arise when the propeller moves very slowly in the video, has the wrong number of blades, is torn, twisted into a spiral or does not move at all (which is especially scary for helicopters).

Changeable traffic signs or traffic information boards are sometimes designed with LEDs operating in multiplex mode. For example, the individual elements are illuminated for one tenth of the time at ten times the average intensity. Above the splice frequency, one then perceives the overall image, but with the aid of a camera or a camera one observes a partial image from which the sign is not recognisable or the text of the information board is illegible, especially at high ambient light intensity. Special image sensors are therefore used in traffic sign recognition systems.

Barcode sensors can be interfered with by intense, deeply modulated light, up to frequencies of tens of kilohertz. Since they usually sense red light (LED or laser LED), they are most susceptible to interference in this region. Milder interference only increases the time to read the code, while stronger interference makes it impossible to read the codes at all.

LED displays or VFDs in multiplexed mode can also exhibit surprising effects on the record besides flickering - only part of the digits are captured on the digital clock image or the recording shows a slippage of individual characters or their elements, although the eye observes stable text. Flashing displays can distract from the intended focus of the scene.

Monitors with a conventional screen (CRT) usually flicker on the recording because the frame rate is usually between 60 and 85 Hz, higher than the half-frame rate of the camera.

TVs with a classic screen usually don't flicker on the record. Sometimes a lighter and darker horizontal bar slowly moves across the screen. The speed of this movement corresponds to the small difference in frame rate between the camera and the screen, with the lighter part apparently being exposed twice during a single camera frame.

In LCD displays and TVs, these effects occur to a much lesser extent, probably depending on the penetration of the mains frequency to the output of the inverter for the fluorescent or LED backlight of the display.

The tail lights of vehicles with LEDs operating in pulsed mode may appear to flicker on video footage, although the human eye observes a steady intensity of light.

From the cameraman's point of view

From the cameraman's point of view, in addition to the correct camera settings, it is also desirable to prepare for the shoot by checking that there is no flicker in the space at different lighting settings, including dimming, even when the camera is moving. The recording can be de-flicker-free in post-production as well.

From the perspective of lighting technology

From the point of view of lighting technology, it is desirable that the illumination does not flicker and create a stroboscopic effect, as required by EN 12464-1:2012 in Article 4.8 - both from the human point of view and from the point of view of the increasingly widespread image recording devices. This requirement can be met by supplying light sources with the lowest ripple current (preferably DC) or a higher frequency current (preferably in the order of kilohertz or more). If the lighting systems are dimmable, they should comply at all settings.

The next part of the article will be devoted to the power supply circuits of light emitting diodes in terms of mixing.

Literature:

[1] Avoid Video Flicker. Urban Video Inc. [online]. 2014 [cited 2018-05-04]. Available from: http://urbanvideo.ca/avoid-video-flicker

[2] OHTSUKA, Yoshimichi and Yuichi NINOMIYA. Television flicker eliminator for fluorescent lighting. Tokyo: NHK Science and Technical Research Laboratories, 1985. NHK Laboratories Note. Available from NTK.

Author. Antonín Fuksa, NASLI & Blue step
Published in Světlo 3/2018