Knowledge · Fundamentals

Triggering, timing and shutter choice without guesswork.

Missing images, blurred edges or distorted parts can make an otherwise usable inspection fail. Coordinate the trigger, sensor exposure and lighting before setting image-processing thresholds. Then check that the result reaches the machine before it must act.

Timing in Computer Vision

Core choice

Global vs rolling shutter

A global shutter gives the pixels a common exposure interval. A rolling shutter exposes different rows at different times, so a moving object can appear skewed even when each row is sharp. Choose from the permitted geometric distortion, motion and exposure requirements. Confirm the exact sensor timing rather than inferring it from the camera interface.

Choose global when

A common exposure interval avoids row-to-row geometric skew on moving parts. It does not remove motion blur within that interval. Set exposure from the allowed displacement and verify the resulting signal and contrast.

Rolling can work when

The object is stationary or the row-to-row displacement remains within tolerance. A strobe only avoids rolling-shutter distortion when its timing illuminates all required rows during a compatible exposure interval; verify this with the sensor mode and actual light pulse.

Include lighting

A short pulse can define effective exposure when it dominates ambient illumination and overlaps the required sensor exposure window. Check the exact sensor mode: a pulse outside that window or covering only some rolling-shutter rows can produce missing or uneven image information.

Timing

Trigger path you must specify

Use the sequence trigger → accepted trigger → sensor exposure (light pulse inside the compatible window) → readout/transfer → processing → valid result → machine action as a timing record. Some cameras overlap exposure, readout and transfer; do not assume every duration simply adds. Record start/end events and the actual end-to-end deadline with the selected mode.

  1. 01

    Trigger source

    Identify the timing reference and trigger source: sensor, encoder, controller message or software request. Measure delay and jitter, the variation in that delay, at the camera. Verify ready/busy behaviour and what happens when another request arrives before the camera can accept it.

  2. 02

    Exposure vs strobe

    Illustrative calculation: image scale is 0.05 mm/pixel, allowed blur is 2 pixels and speed is 1,000 mm/s. Allowed displacement is 2 × 0.05 = 0.10 mm, so maximum exposure is 0.10 ÷ 1,000 = 0.0001 s = 100 µs. This limit concerns blur, not trigger delay or total result time. Verify that the light supplies enough signal at that exposure.

  3. 03

    Debounce & filtering

    Inspect the actual trigger waveform for repeated edges or noise and validate any debounce/filter setting. Filtering can suppress a genuine short pulse or add delay. Test the shortest valid signal and closest product spacing, then check the number of accepted triggers against images and results.

  4. 04

    Validate at takt

    Test speed extremes, closely spaced products, stop/restart and missed triggers with required logging and communication active. Record trigger, exposure-active, result-valid and actuator events where available. Compare the longest observed result time with the machine deadline and test the specified response to late or missing data. A high average frame rate does not prove this deadline is met.

Vincent van Montfoort

Timing problem on the line?

Bring speed, image scale, blur limit, trigger waveform and the required result deadline. Include camera mode, light/controller settings and a failed image or event trace. For the cost comparison, record fault-related interventions and missed inspections, then estimate diagnostic, hardware and validation work in hours × rate plus parts. Validate the timing change before claiming reduced downtime.