Rolling shutter global reset with flash for moving objects

A rolling shutter sensor exposes its rows one after another, a few microseconds apart, so a part that moves during the scan comes out leaning, sheared or smeared. That single drawback keeps many otherwise affordable cameras out of conveyor work. A global reset mode, available on many rolling shutter sensors, removes the skew, and a microsecond flash paired with it removes the brightness gradient too. Here is the timing rule, a worked example at 30 fps with the part moving at 200 mm/s, the ambient-light number that decides whether you need a shroud, and the point at which a global shutter camera is simply the better buy.

What global reset changes, and what it does not

In ordinary rolling shutter operation, each row starts exposing one row period after the row above it, so a moving part is drawn at a different position by every row. The result is the classic shear or lean that disqualifies these cameras from continuous-motion work, despite a design that reaches smaller pixels at lower cost than a global shutter part.

Global reset mode — sold under names such as global reset release (GRR), global start or all-rows mode, and set among the shutter controls of the camera's feature list — resets every row at the same instant. All rows then integrate the scene while the part sits in one position, and the shear disappears.

The readout does not change. A rolling shutter sensor has no storage behind its pixels, so each row must be read the moment it stops integrating, and readout proceeds one row at a time. Row 1 ends after the set exposure; every row below it ends one row period later. The last row of a 3000-row frame therefore integrates about 33 ms longer than the first. Under continuous light the image brightens toward the bottom and the lowest rows smear more. Three formulas cover the whole behaviour:

N is the row count, t_row the row period, t_set the set exposure (row 1's active window), v the part speed and t_flash the flash duration.

The flash timing rule

Fire the flash after the common reset and end it before row 1 stops. Row 1 owns the shortest window in the frame, so a pulse that fits inside it fits inside every row's window: each row then records exactly the same flash energy, and with ambient blocked that energy is the whole image.

t_delay is the wait from exposure start to the flash edge. Keep the sum at or below about 80% of t_set: driver rise time and trigger jitter of a few microseconds eat the difference, and LED strobe rise time shows where those microseconds go. With a 50 µs set exposure, a 20 µs flash starting 15 µs in leaves a 30% margin.

If frames come back black while the strobe visibly fires, the pulse is landing outside the window; strobe light on but camera image is black walks that timing chain end to end.

Worked example: 12 MP at 30 fps, part at 200 mm/s

The LS-CMR-0166 reads 4000 × 3000 pixels at 30 fps over USB3, from a 1/1.7" rolling shutter sensor — a representative 12 MP conveyor camera.

The frame period is 33.3 ms, so a sensor that reads continuously steps about 11.1 µs per row (33.3 ms ÷ 3000 rows). The exact row period sits in the camera's timing controls; run your final numbers with the value your camera reports.

Set the exposure to 50 µs. Without a flash, the last row integrates 50 µs + 2999 × 11.1 µs ≈ 33.3 ms, which is 667 times the first row's dose — about 9.4 stops of gradient. With the shutter left in its normal rolling mode, the same part shears by 200 mm/s × 2999 × 11.1 µs ≈ 6.7 mm, roughly 266 pixels at a 100 mm field width (25 µm per pixel).

Now add the flash: 20 µs long, fired 15 µs after exposure start. The window check passes — 15 + 20 = 35 µs against a 50 µs budget. Motion blur is 200 mm/s × 20 µs = 4 µm, about 0.16 pixels: invisible. Skew is zero by construction, and all 3000 rows record the same 20 µs of flash. Those four numbers are the entire design.

How much ambient you can tolerate

A shroud changes the problem from hard to trivial: with ambient blocked, any pulse that meets the timing rule gives a uniform frame, whatever the room is doing.

Without a shroud, ambient light integrates for each row's full window while the flash contributes only t_flash. Take the worked example: the bottom row sees ambient for 33.3 ms against 20 µs of flash. For ambient to stay under 10% of the dose even in that brightest row:

The flash must arrive roughly 1.7 × 10⁴ times brighter than the ambient, irradiance for irradiance. Factory lighting misses by orders of magnitude, daylight misses further. A narrowband filter matched to the LED helps — a 10 nm passband across the visible range passes only a few percent of broadband light, one to two orders of magnitude — rarely enough on its own in daylight.

Mains flicker stacks a second problem on the same rows: at 50 Hz, a 33.3 ms window spans more than 1.5 flicker cycles, so ambient variation prints as bands on top of the gradient. Camera flicker under artificial lighting covers the diagnosis when you inherit such a station. Shrouding removes both problems at once, and it keeps a daylight-dominating flash out of operators' eyes — treat the shroud as part of the light, not an accessory.

Set-up: one trigger source, two devices

Trigger the camera and the light from one source. A part-present sensor feeds the camera's trigger input and the light controller at the same instant; the controller needs a trigger input with programmable delay and pulse width so the flash lands inside row 1's window. If the camera ignores the trigger, its acquisition mode is the usual suspect — GigE camera ignores external trigger is the fix.

On the camera, three settings do the work: shutter mode set to global reset, exposure set to t_set, and every auto mode off, so exposure and gain cannot drift between frames. Some cameras expose the common reset as a global-reset-release time — the wait between reset and readout; where that parameter exists, it is your t_delay.

On the light side, constant-current strobe control holds the pulse current flat so flash energy repeats frame to frame; the LS-CTL-0180 is a 4-channel, 120 W constant-current controller for 24 V lights. A constant-voltage box with a strobe feature works at these pulse lengths too: the LS-CTL-0070 is a 4-channel, 200 W digital control box with a strobe feature.

Scope the chain before trusting images: trigger edge in, light pulse out, delay and width as computed. The wider decision order — lighting and exposure before software — is laid out in stable 2D vision inspection systems.

Choosing the flash source

For a small field, a strobed spot light is the flash. The LS-SPT-0024 covers a 49.9 × 66.6 mm lit area through a 32° lens at 455 nm, and the LS-SPT-0162 puts a 24 mm diameter white spot on tighter targets.

Size the lit area to the field of view with 20-30% margin, pick the wavelength the part reflects best, and confirm the light's rated pulse duration and duty cycle at your frame rate. Strobing also extends LED life, because the diode rests between pulses — LED lifetime L70 gives the numbers behind that. More light-side options sit in the spot lights category and its shop finder.

When a global shutter camera is the better answer

Skip the whole technique when one of these rows matches your job:

The 0.75 mm/s figure comes from the worked example's geometry: one pixel of shear is 25 µm of travel in the 33.3 ms scan. Rescale with skew ÷ pixel size for your own field and speed.

Buying the global shutter camera deletes the shroud, the controller channel and every timing calculation above, at the cost of a more expensive sensor. In multi-camera systems the flash hardware repeats per station and can outgrow the camera price difference, so price both routes before committing. For a 2.5 MP global shutter option on GigE, the MV-CU020-90GC is the compact pick.

Blur on the global shutter camera is then capped only by exposure — exposure time vs shutter speed for moving-part inspection sets that ceiling — and the wider shutter comparison lives in global vs rolling shutter sensors: choosing for moving targets.

Checks before you order

• Read the row period from the camera's timing controls and redo the window arithmetic with the real value; the 11.1 µs above is frame time ÷ rows, an estimate.
• Keep t_delay + t_flash at or below 80% of t_set once the real row numbers are in.
• Confirm global reset in the camera's feature list before designing the station: it is a sensor capability, and the feature list is the proof.
• Check the light's rated pulse duration and duty cycle against your frame rate and flash energy.
• Measure ambient irradiance inside the planned field and apply the 16,700× test, or budget a shroud.
• Acceptance test on the finished station: image a uniform target and require the top-50-row mean to sit within 10% of the bottom-50-row mean — about 25 gray levels on an 8-bit image.

To build the strobed route, pick the sensor from the area scan cameras category, add a strobe-rated channel from the light controllers category, or work through the camera configurator step by step. For a second opinion on a station you already run, the contact page reaches our applications team.