Shutter Types, Rolling Shutter and Flicker

A dancer leaves the floor under LED (light-emitting diode) stage light, and you catch her at 1/1000 s in silent mode. On a large screen the frame carries soft horizontal stripes, and her raised leg leans to one side. Nothing broke. A light pulsing faster than the eye can follow met a shutter that records the top of the picture several milliseconds before the bottom.

The Z6III starts and stops an exposure in three ways, chosen in Custom Setting d6, and offers four menu items aimed at flicker. Page numbers in parentheses point to Nikon's Z6III Reference Guide, edition 09, which matches firmware 2.00.

01.What a Shutter Does

While a pixel is allowed to collect, incoming light frees electrical charge inside it, and the camera later measures how much built up. Write the brightness of the light as $B$ and the collecting time as $T$. The recorded exposure is

$$H = B \, T$$

so doubling the time doubles the recorded light. The shutter sets $T$ by deciding when collection starts and when it ends.

Nikon describes the mechanical shutter as a vertical-travel focal-plane shutter (p. 1000). It sits just in front of the sensor and moves across the short side of the frame, 23.9 mm here (p. 997). A front curtain moves away to uncover the sensor and a rear curtain follows to cover it again, so each row of pixels collects from the moment it is uncovered to the moment it is covered. At slow speeds the front curtain finishes before the rear one sets off. At fast speeds the rear curtain sets off before the front one has finished, so the frame is exposed through a travelling slit. Every row still collects for the same $T$, but the rows start at different moments, and that one fact is the root of both rolling-shutter distortion and flicker banding.

Nikon does not publish how long the curtains take to cross, but the flash sync speed caps it. A flash fires in one brief burst, so the whole frame must be uncovered at one instant, and the Z6III syncs at 1/200 s and slower (p. 1000). At 1/200 s the rear curtain sets off 5 ms (milliseconds) after the front one, and the front curtain must have finished by then, so its crossing time $\tau$ obeys

$$\tau \le \frac{1}{200}\ \text{s} = 5\ \text{ms}$$

That bound is my inference, not a Nikon figure, and the real crossing may be quicker. Faster flash uses 1/200 s (Auto FP), Nikon's high-speed sync, under Custom Setting e1 (p. 690).

02.Custom Setting d6 Shutter Type

d6 chooses the shutter for still photographs from Auto, Mechanical shutter, Electronic front-curtain shutter and Electronic shutter (p. 668). The default is Auto (p. 624).

Mechanical Shutter

With Mechanical shutter both curtains move for every frame (p. 668). They make a sound, and their starting and stopping shakes the camera slightly, which Nikon calls "shutter shock" (p. 846). Shake that arrives while the pixels are still collecting smears fine detail. Speeds run from 1/8000 to 30 s in modes S and M, shutter-priority and manual (pp. 140, 142), with Bulb and Time beyond that in mode M (p. 146). Nikon notes that this option is unavailable with some lenses, without naming them (p. 668).

The curtains still sweep, so rolling effects shrink rather than vanish. A subject crossing the frame at speed $v$ moves $v\tau$ between the first edge being exposed and the last. For a dancer's arm at an illustrative 3 m/s, the 5 ms ceiling gives at most $3 \times 0.005 = 0.015$ m, or 1.5 cm of lean.

Electronic Front-Curtain Shutter

Electronic front-curtain shutter, EFCS for short, hands the start of the exposure to the sensor. The sensor is already uncovered for the live view, so the camera starts each row by emptying it of charge, a step called a reset, in a sweep that follows the path a front curtain would take. The rear curtain then closes mechanically. Nothing moves as the exposure begins, which is why Auto uses EFCS at slow speeds, to reduce blur from camera shake (p. 668).

Its limit is speed. EFCS stops at 1/2000 s (p. 668), a faster setting in modes S and M is held there (pp. 140, 142), and Auto FP flash tops out there too (p. 690). It does raise Continuous H to about 10.5 frames per second (fps), against 8.1 fps with Mechanical shutter or Auto (p. 154).

Nikon's troubleshooting pages warn that fast shutter speeds or fast lenses, those with a wide maximum aperture, can make bokeh, the way out-of-focus areas are rendered, irregular in shape. The remedy given is a slower speed or a higher f-number (p. 915). The page names no shutter type.

Electronic Shutter

Electronic shutter moves nothing (p. 846). Each row is reset to begin collecting and read out to finish, one row after another down the frame. It reaches 1/16000 s (p. 140) and makes no mechanical noise or vibration (p. 846).

The cost is readout time. Two independent measurements gathered in What the Z6III Is put the stills readout near 14.5 ms: Thom Hogan's read-out table gives 14.4 ms and DPReview's review about 14.6 ms in 14-bit mode. That is nearly three times the 5 ms mechanical ceiling, so the illustrative arm now leans $3 \times 0.0145 \approx 0.044$ m, about 4.4 cm, and a panned camera skews the whole frame. Nikon lists motion distortion, banding under fluorescent, mercury-vapour or sodium lamps, and bands from strobes among silent-mode effects (p. 846) and general shooting cautions (p. 105).

Custom Setting e1 Flash sync speed becomes unavailable (p. 668), flash syncs no faster than 1/60 s, and Auto FP is off the table (pp. 417, 1000). In exchange you get the fastest continuous frame rates (p. 153): Continuous H reaches about 15.3 fps with RAW files and 16 fps when only processed JPEG or HEIF files are recorded (pp. 154, 1000). Every other d6 setting caps a burst at 200 shots even when d2 Maximum shots per burst is set to no limit, though not at 1 s or slower in S or M (p. 663). Release Modes, Frame Rates and the Buffer covers both.

Auto

Auto switches between the mechanical shutter and EFCS according to shutter speed (p. 668). It never chooses the fully electronic shutter, and because EFCS stops at 1/2000 s, any faster speed must use the mechanical one. Nikon does not publish the changeover speed, so I cannot say which shutter fires at 1/500 s.

When the Camera Overrides d6

Silent mode in the setup menu forces the electronic shutter whatever d6 says, stops an attached flash from firing, and disables Photo flicker reduction and Long exposure NR, the long-exposure noise reduction (p. 846). Interval-timer shooting, time-lapse video and focus shift each have an Electronic shutter options > Electronic shutter setting that, when ON, overrides d6 (pp. 522, 532, 542), and it is ON by default for all three (p. 448). Pixel shift always uses the electronic shutter (p. 554). Silent Shooting, Timers and Time-Lapse covers these features.

03.Flicker From First Principles

Why Lamps Pulse at Twice the Mains Frequency

Mains electricity is alternating current (AC): the voltage swings positive and negative 50 times a second in some regions and 60 in others. One hertz (Hz) is one cycle per second. Write the voltage as $V(t) = V_0 \sin(2\pi f_m t)$, with $V_0$ the peak voltage and $f_m$ the mains frequency.

A lamp's output ignores the sign of the voltage. For a simple resistive load, power goes with the voltage squared, and the identity $\sin^2\theta = \frac{1}{2}(1 - \cos 2\theta)$ gives

$$P(t) \propto V_0^2 \sin^2(2\pi f_m t) = \frac{V_0^2}{2}\left[1 - \cos(2\pi \cdot 2 f_m \, t)\right]$$

The power swings at $2 f_m$: 100 times a second on 50 Hz mains and 120 on 60 Hz, the two rates photo flicker reduction looks for (p. 483). A hot filament glows through each trough, so its flicker is shallow. Fluorescent and mercury-vapour lamps, the types Nikon names (pp. 483, 600), can dim much further. Call the depth of the swing $m$, from 0 for steady light to 1 for light that falls to nothing.

What One Row Records

Model the lamp's brightness as

$$B(t) = \bar{B}\left[1 + m\cos(2\pi f t)\right]$$

where $\bar{B}$ is the average and $f$ the flicker frequency. A row that collects for time $T$, centred on the moment $t_c$, records the integral of $B$ over that window, which works out to

$$H_{\text{row}} = \bar{B}\,T\left[1 + m\,D\cos(2\pi f t_c)\right], \qquad D = \frac{\sin(\pi f T)}{\pi f T}$$

$\bar{B}T$ is what a steady lamp would give. The bracket depends on where in the cycle the window sits, through $\cos(2\pi f t_c)$, and on how much of the cycle the window averages away, through $D$. When $fT$ is small, $D$ is near 1 and the row inherits the lamp's full swing. When $fT$ is a whole number, the window holds complete cycles, $D = 0$, and every row records the same light.

Take a 100 Hz lamp with $m = 0.4$, an illustrative depth rather than a measurement. At 1/1000 s, $fT = 100 \times 0.001 = 0.1$ and $D = \sin(0.1\pi)/(0.1\pi) = 0.309/0.314 = 0.984$. The swing is $0.4 \times 0.984 = 0.393$, so rows range from 0.607 to 1.393 times the average. A stop is a factor of 2, so

$$\frac{1.393}{0.607} = 2.30, \qquad \log_2 2.30 \approx 1.20\ \text{stops}$$

At 1/200 s, $fT = 0.5$ and $D = 1/1.571 = 0.637$, the swing is $0.255$, and the spread is $\log_2(1.255/0.745) = \log_2 1.68 \approx 0.75$ stop. At 1/100 s, $fT = 1$ and $D = 0$. The speeds that freeze a dancer let flicker through, and at 100 Hz, taking the labels at face value, only 1/100 s, 1/50 s, 1/25 s and other whole multiples of the 10 ms period cancel it.

From Rows to Bands

With the electronic shutter, the window centre $t_c$ slides later by the whole readout time $t_r$ between the first row and the last, so down the frame $\cos(2\pi f t_c)$ runs through $n = f\,t_r$ cycles. At 100 Hz, $n = 100 \times 0.0145 = 1.45$, and at 120 Hz, $n = 120 \times 0.0145 = 1.74$. A frame at 1/1000 s shows one and a half to nearly two bright-and-dark cycles stacked down the picture, about 1.2 stops apart at the illustrative depth.

With the mechanical shutter the sweep takes $\tau \le 5$ ms, so $n \le 0.5$ at 100 Hz and $n \le 0.6$ at 120 Hz. Less than one cycle fits, so the frame shows a gradient rather than stripes. In a burst each frame also starts at a different point in the cycle, so exposure and colour wander from frame to frame, as Nikon warns (p. 483).

High-Frequency LED Lighting

Many LED fixtures and screens are dimmed by pulse-width modulation (PWM): the driver switches the LED fully on and off at a fixed rate and sets brightness by the fraction of each period it stays on. The fixture's designer picks that rate, not the mains, and it can sit far above 120 Hz. Nikon calls this high-frequency LED lighting and includes LED displays (p. 485).

An on-off wave is a sum of sine waves at its own rate and whole-number multiples of it, and all of them cancel under the same condition, $fT$ a whole number. Take an illustrative fixture at 1,500 Hz with $m = 0.4$ for its main component. The readout spans $1500 \times 0.0145 = 21.75$ cycles, so an electronic-shutter frame carries about 22 thin stripes. At 1/2000 s, $fT = 0.75$ and $D = 0.707/2.356 = 0.300$, a swing of 0.12, or about 0.35 stop. At 1/1500 s, $fT = 1$ and the stripes vanish.

But 1/1500 s is not on the ordinary scale. The neighbouring marks, 1/1600 s and 1/1250 s, give $|D|$ of 0.066 and 0.156 if the labels are taken at face value, and the labels are rounded: 1/500 s is precisely 1/512 s (p. 485). Finding the null for an unknown fixture needs finer steps and a way to watch the result.

04.The Flicker Reduction Items

Photo Flicker Reduction

The options are ON and OFF, default OFF (p. 447). It targets lamps such as fluorescent or mercury-vapour types and detects 100 and 120 Hz only (p. 483). When it finds flicker at release, a green indicator appears beside the FLICKER icon (p. 483).

Nikon says only that it may slightly delay the shutter response (p. 483). My inference is that it times the release to a consistent point in the lamp's cycle, which fits the delay and explains why the electronic shutter defeats it. A 14.5 ms readout spans 1.45 cycles at 100 Hz, so every frame holds peaks and troughs. A curtain sweep of at most 5 ms spans at most half a cycle, and centred on a peak its rows run from the average to 1.39 times it at 1/1000 s and the illustrative depth, $\log_2 1.39 \approx 0.48$ stop against 1.2.

In bursts, continuous high-speed (extended) mode favours frame rate and may lose the effect in long bursts, while the other continuous modes favour flicker reduction, blank the display every frame and may slow or stutter (p. 483). It does not take effect during HDR (high dynamic range) overlay, pixel shift or high-speed frame capture +, in silent mode, or with Electronic shutter in d6, a list Nikon opens with "including" (p. 484). It can miss flicker in bright scenes, against dark backgrounds or under decorative lighting (p. 483). If it fails, Nikon suggests restarting the camera while it points at the subject or the light (p. 483). Among its bird-detection cautions, Nikon recommends OFF when nothing flickers (p. 129).

High-Frequency Flicker Reduction for Photos

The options are On (shutter speed fine-tuning) and Off, default Off (p. 447). In modes S and M it lets you move the shutter speed in smaller steps than the normal scale while watching the preview for bands (p. 485). Nikon describes no detection, so you are the detector. Its range for each shutter is in the table above (p. 485). The display shows the denominator with one decimal, so 1/512 s reads 512.0, and turning the item off returns to the nearest normal speed (p. 485).

Nikon's method is to pick a speed for the subject, turn the item on, and fine-tune while zoomed in on the display (p. 485). The display can differ from the photograph, so take test shots (p. 486). Exposure maintenance is disabled meanwhile, even if b7 Keep exp. when f/changes in the Custom Settings menu is set to Shutter speed (p. 486).

Video Flicker Reduction

This item reduces flicker and banding in the display and in footage under fluorescent or mercury-vapour lighting (p. 600). Frequency offers Auto, 50 Hz and 60 Hz, default Auto, and Digital flicker reduction offers ON and OFF, default OFF (p. 579). It is separate from the photo item and does not follow it, though a change to Frequency also applies to the photo-mode shooting display (p. 600). Nikon does not say it changes photographs.

If Auto fails, Nikon advises picking 50 Hz or 60 Hz for the local supply and testing, and for a very bright subject a smaller aperture (p. 600). Its fallback is mode S or M at 1/100, 1/50 or 1/25 s on 50 Hz, or 1/125, 1/60 or 1/30 s on 60 Hz (p. 600). Taking the labels at face value, those give $fT$ of 1, 2 and 4 at 100 Hz. At 120 Hz, 1/60 s and 1/30 s give 2 and 4, while 1/125 s gives $120/125 = 0.96$ and $D = 0.042$, a swing of about 0.017 at the illustrative depth, against 0.39 at 1/1000 s. Slow-motion video cannot use the item (p. 204).

Digital flicker reduction processes the images in mode S or M, and Nikon suggests it for speeds quicker than 1/100 s or 1/125 s (p. 601). It is enabled only when all of these hold (p. 601):

• The shooting mode is S or M.
• Video file type is neither N-RAW 12-bit (NEV) nor ProRes RAW HQ 12-bit (MOV).
• The frame size and rate is $3840 \times 2160$ at 30p, 25p or 24p without Hi-Res Zoom, $1920 \times 1080$ at 60p without Hi-Res Zoom or the DX crop image area, $1920 \times 1080$ at 50p without Hi-Res Zoom, or $1920 \times 1080$ at 30p, 25p or 24p.
• Skin softening is Off.
• Extended shutter speeds (S/M), Custom Setting g11, is OFF.

Nikon lists what it calls lower limits for shutter speed: 1/160 s at 60p on a 50 Hz supply or 1/125 s on 60 Hz, 1/100 s at 50p, 1/60 s at 30p, and 1/50 s at 25p and 24p (p. 601). I read these as the slowest speeds it still works at, so 1/100 s is advice, not a floor. Intense flicker may still defeat it, movement can leave afterimages, and an icon shows when it is disabled (p. 601).

High-Frequency Flicker Reduction for Video

The options are On (shutter speed fine-tuning) and Off, default Off (p. 579), fine-tuning 1/8000 to 1/30 s in modes S and M (p. 602). It is a separate item that Nikon explains by pointing back to the photo version (p. 602), and I found no statement that one follows the other. Choosing Shutter angle for Custom Setting g12 Shutter mode disables it (p. 770).

05.Stage Dance Under LED Light

A stage can carry both kinds of flicker: house lights at mains rate and LED fixtures dimmed by PWM at a rate nobody prints in the programme.

I leave d6 at its default, Auto, whenever the venue allows the sound. It pairs EFCS at slow speeds with the mechanical shutter at fast ones and never picks the electronic shutter (p. 668), so the exposure crosses the frame in at most 5 ms, not about 14.5 ms. A leaping dancer leans about a third as far at most, mains flicker grades instead of striping, and Photo flicker reduction stays available. Nothing in the Reference Guide favours Mechanical shutter over it, and both shoot Continuous H at about 8.1 fps (p. 154). Speeds that freeze a leap with the Tamron 70-180mm f/2.8 Di III VC VXD G2 sit well above 1/100 s, the fastest null at 100 Hz, so the sweep is the part I can control.

I work in mode M and, before the show, check the zoomed preview on a lit backdrop. If stripes show, I turn on High-frequency flicker reduction, nudge the speed until they fade, and take test frames (pp. 485-486). I leave Photo flicker reduction at OFF unless test frames drift from one to the next, because it can delay the release and make continuous rates erratic (p. 483), which can cost the top of a jump.

When silence is required, I set d6 to Electronic shutter and Shutter sound to OFF, as in First-Day Setup (p. 844), or turn on Silent mode, which also mutes the beep (p. 845). Either way I accept the costs: roughly three times the lean or more, readout banding, and no Photo flicker reduction (pp. 484, 846). Fine-tuning still works to 1/8000 s (p. 485) and becomes the main defence. For video I set Frequency to the local mains, as Nikon's troubleshooting advises (p. 914), 60 Hz in the United States, rather than leave it to Auto, and 1/60 s from Nikon's 60 Hz list suits 24p and 30p (p. 600).

06.Recommended Settings

07.Beginner Mistakes

Blaming the lights for bands the shutter choice created. A 14.5 ms readout turns 100 Hz flicker into about one and a half bands per frame, while the mechanical sweep gives a gentle gradient.

Expecting photo flicker reduction to fix LED stripes. It detects 100 and 120 Hz only (p. 483) and has no effect with the electronic shutter (p. 484).

Using flash in silent mode or with the electronic shutter. In silent mode the flash does not fire (p. 846), and with Electronic shutter it is held to 1/60 s or slower (p. 1000).

Forgetting that time-lapse ignores d6. Interval-timer shooting, time-lapse video and focus shift use the electronic shutter by default (p. 448).

Trusting the preview alone. The display can show flicker differently from the photograph (pp. 483, 486), so a test frame is the proof.

08.Related Chapters

• What a Partially Stacked Sensor Changes: why the readout takes about 14.5 ms.
• Release Modes, Frame Rates and the Buffer: frame rates for each shutter type.
• Silent Shooting, Timers and Time-Lapse: silent mode, the interval timer and time-lapse.
• Exposure Modes: modes S and M, where flicker fine-tuning works.
• Vibration Reduction: the other defence against camera shake.
• Glossary: short definitions of the terms used here.