Do global shutter cameras need external trigger? How to check

A global shutter camera and an externally triggered camera answer two different questions. The shutter type decides whether a moving part looks straight inside one frame. The trigger decides whether that frame is taken while the part sits where you want it. Confusing the two produces cameras that render smear-free images of parts already out of position, or trigger hardware on stations where nothing ever moves. This explainer separates the jobs, works the arithmetic that links speed, exposure and blur, and lists the checks — I/O pinout, GenICam trigger features, interface limits — that confirm a specific camera has a trigger input before you order it.

Global shutter and external trigger answer different questions

No, a global shutter camera does not need an external trigger, and an external trigger does not need a global shutter. Global shutter means every pixel starts and ends its exposure at the same instant, so a moving part is sampled as one slice of time. An external trigger is an input line: a pulse on it starts the exposure. The shutter controls distortion inside the frame; the trigger controls when the frame happens.

Put numbers on it. A rolling shutter camera running at 30 fps reads its rows over roughly the frame period, about 33 ms. A part moving at 150 mm/s travels 150 mm/s × 33 ms = 4.95 mm between the first row and the last, so a round bore images as an ellipse. On a global shutter sensor the skew is zero at any speed, triggered or not.

Now leave the trigger out. A global shutter camera free-running at 30 fps starts frames on its own clock, so the part can be anywhere in the field when exposure begins. Position uncertainty is part speed × frame period — here 150 mm/s × 33 ms ≈ 5 mm — and when part spacing varies you also get two frames of one part and none of the next. Straight but mistimed is as unusable as timed but crooked.

Readout time sits close to the frame period when a datasheet states no separate figure, so treat the skew value as an estimate. For how the sensor architecture trades speed against dynamic range, see the global shutter CMOS sensor comparison.

What a hardware trigger buys you

A pulse on the opto input starts the exposure, and that single fact buys four things on a machine:

• One frame per part. The photoelectric sensor or the PLC decides what a part is; the camera obeys. No duplicates, no misses when spacing varies.
• Fixed geometry. Exposure starts at the same point in travel every cycle, so each feature lands on the same pixels. Thresholds, regions of interest and focus hold their settings.
• Strobe sync. The light fires inside the exposure window instead of washing the whole cycle.
• Multi-device sync. One pulse, several cameras, one instant — the basis of hardware sync between two USB cameras and of IMU-fused motion work with global shutter camera IMU modules.

Budget the latency chain in millimeters of part travel, not in abstract microseconds. A photoelectric sensor responds in under about 1 ms; a compact PLC adds 1-10 ms of scan time; the camera's trigger-to-exposure delay on industrial area scan cameras is typically well under 0.1 ms. At 150 mm/s, a 6 ms chain means 150 × 0.006 = 0.9 mm of lag. If that fits inside your field-of-view margin, the trigger is doing its job; the frame rate you can push through external pulses is the other limit to size.

A software trigger is the middle ground: the command reaches a GigE or USB3 camera in about a millisecond, but its jitter grows with bus and host load. Use it on stationary parts. On a moving line it becomes a position error you cannot tune out.

How to verify a camera has a trigger input

Four checks, ordered by what they cost you:

1. The I/O drawing. The pinout sheet of an area scan camera lists the opto-isolated input with its voltage range, commonly 5-24 V, which a 24 V PLC output drives directly. If the drawing shows only outputs and a strobe line, there is no trigger input. For polarity and which side drives which, read camera opto-isolated GPIO circuit direction.
2. The feature tree. On a GigE Vision or USB3 Vision camera the GenICam feature list is the contract. It must expose TriggerMode and TriggerSource, and TriggerSource must offer a hardware line, not only a software command. Set single-frame acquisition, enable TriggerMode, point TriggerSource at the line and pulse it: one frame per pulse is a pass. Two classic failures — trigger source not detected and trigger fires once then stops — are configuration mistakes, not missing hardware.
3. The interface. MIPI CSI-2 modules do not implement GenICam natively; their trigger arrives on a carrier-board pin, so read the connector pinout before matching a camera. UVC-class USB cameras usually expose no trigger at all — how to decide on USB trigger support walks those cases.
4. The device class. Smart cameras and vision sensors invert the question: a vision sensor's job is one image per trigger, so the I/O sits on the base unit.

Two trigger-first devices from the vision-sensor side:

Fixed-focus sensors are ordered by working distance rather than by lens choice, which removes one variable from the trigger chain.

Choosing the camera for a triggered station

Pick resolution first: at least three pixels across the smallest feature, which fixes the sensor class before anything else. The camera finder filters by resolution and shutter type together. Then the interface. GigE keeps wiring simple and, with PoE, runs on one cable — check the port supports it. USB3 suits benches and short runs, with the USB trigger caveats above. CoaXPress carries the trigger on the same coax as the video, so a frame grabber system needs no separate trigger wiring, as on the MV-GX1008 octo CXP-6 frame grabber.

Three global shutter area scan cameras across the resolution range:

All 3 share: Sensor type CMOS, global shutter; Exposure mode Off/Once/Continuous exposure mode; Certification CE, RoHS, KC; Mono/color Color. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

The 2.5 MP GigE model covers small fields with the lightest cabling; the 25 MP 10GigE F-mount model inspects a full tray in one exposure. Resolution first, interface second, then confirm the I/O block through the checks above.

Worked example: bracket on a 150 mm/s conveyor

The part is a 90 mm stamped bracket on a belt running 150 mm/s; a photoeye fires when it crosses a fixed point. The smallest feature to resolve is a 0.2 mm edge radius.

Camera. A 12 MP global shutter camera with a 1.1" sensor carries the resolution with margin; the MV-CH120-11UC is one such model — 12 MP, USB3, global shutter, 1.1" sensor, C-mount. A sensor of this class carries a 4096 × 3000 grid of 3.45 µm pixels, so the active area is 4096 × 3.45 µm = 14.1 mm by 3000 × 3.45 µm = 10.4 mm.

Lens and field of view. Target 105 mm across to hold the bracket with margin. Magnification m = sensor width / FOV = 14.1 / 105 = 0.134. At a 320 mm working distance, f = WD × m / (1 + m) = 320 × 0.134 / 1.134 = 37.8 mm, so order a 35 mm C-mount lens. Fitted, m = f / (WD − f) = 35 / 285 = 0.123, FOV = 14.1 / 0.123 = 115 mm, and one pixel covers 3.45 µm / 0.123 = 28 µm at the part. The 0.2 mm radius spans 0.2 / 0.028 ≈ 7 pixels, better than double the three-pixel minimum.

Exposure and blur. At 150 mm/s, 100 µs of exposure smears the image by 150 × 0.0001 s = 0.015 mm = 15 µm, about half a pixel — acceptable. A 20 µs strobe pulse inside the window cuts the blur to 3 µm and supplies the light that a 100 µs exposure at f/4 would otherwise struggle to gather. Plan the strobe from the start on a line this fast.

Depth of field at f/4. With a one-pixel circle of confusion, total depth ≈ 2 × N × c × (1 + m) / m² = 2 × 4 × 0.00345 × 1.123 / 0.0151 = 2.0 mm. Bracket height varies ±1.5 mm, so it fits; closing to f/8 roughly doubles the estimate to 4 mm at the price of light.

Trigger budget. Photoeye under 1 ms, PLC scan 5 ms, camera delay small by comparison: about 6 ms total, during which the bracket moves 150 mm/s × 6 ms = 0.9 mm. Against 12.5 mm of spare field on each side — (115 − 90) / 2 — the bracket lands within a millimetre of the same pixels every cycle. That repeatability, not the shutter, is what the trigger sells.

Decision. Global shutter, hardware trigger, 35 mm lens, 100 µs exposure with a strobe. That is the standard configuration for any part moving at this speed.

Wiring the trigger line

The opto input wants a clean switched 5-24 V. A PLC transistor output drives it directly; a 5 V logic input run at 24 V needs the series resistor stated on the I/O drawing — the arithmetic is in Arduino trigger to camera opto-isolated input wiring. One opto output can drive two or three camera inputs; past that, move to a powered distribution board as in triggering multiple cameras from one output. For the machine side end to end, the PLC external trigger solutions guide covers wiring from the controller out.

Most industrial cameras also drive a strobe output that tracks the exposure window. Wire the light to it rather than timing the light separately from the PLC.

When the usual answer is wrong

• The part stops under the camera. Indexed stations need no trigger and no global shutter: exposure can run for milliseconds, and software picks the frame after the PLC reports position. Spending on trigger hardware here buys nothing.
• The part is slow. Jitter is speed × frame period. At 5 mm/s and 30 fps that is 5 × 0.033 = 0.17 mm per frame; if the field-of-view margin swallows it, free-run with software selection is a sound design.
• You already own a rolling shutter camera. Set the exposure longer than the readout and fire one short strobe inside it: every row integrates the same pulse, which is a global capture in everything but silicon. The trick costs a light, not a camera.
• The product never stops — webs, extrusions, continuous flow. Area scan is the wrong tool; line scan with an encoder is the standard answer, sized with the line scan line rate calculator.

Decision table

Checklist before you order

• The I/O drawing shows an opto-isolated input with a voltage range covering your controller output — commonly 5-24 V.
• The feature list exposes TriggerMode and TriggerSource with a hardware line as a source, plus single-frame acquisition.
• The interface carries the trigger end to end: native on GigE, USB3 and CoaXPress area scan cameras; a carrier-board pin on MIPI; usually absent on UVC.
• Latency budget: (sensor + controller + camera delay) × part speed fits inside the field-of-view margin.
• Blur budget: exposure or strobe width × part speed stays under one object-side pixel.
• Trigger pins exist on the connector variant you are buying, not only on an optional breakout board.
• On CXP systems the grabber forwards the trigger over the coax, so count ports, not wires.

Next step: run the camera configurator with the resolution and interface fixed above, or browse the cameras topic for the neighbouring decisions.