There is no universal number because it depends on field of view, required tolerance, and sensor readout time, but as a practical starting point, any application where a part moves more than a few pixels’ width during a single frame’s exposure window deserves serious consideration of global shutter. Running a side-by-side test capture at actual production speed remains the most reliable way to make this decision rather than relying on a generic velocity figure.
A telecentric lens provides constant magnification over the entire depth of field, which eliminates perspective error and parallax. This is critical for accurate 3D profiling and when measuring dimensions precisely. For pure surface inspection where log diameter does not vary more than ±10 cm, a conventional fixed focal length lens with a large depth of field (e.g., f/8) can be adequate and is more compact. Telecentric lenses are also bulkier and more expensive. Cost-sensitive mills often use hybrid approaches: telecentric for the 3D sensor and conventional for the colour camera.
Lighting is equally critical. Red LED line lights (660 nm) are standard for surface inspection because they minimise scatter from knots and produce high contrast. For shallow-angle illumination to highlight grain orientation, blue or white LEDs with diffusers are used. High-quality machine vision software systems integrate the light source into the camera housing to prevent shadows from moving logs. Systems that rely on external lighting often suffer from non-uniform illumination as the log rotates or shifts laterally.
In a controlled laboratory setting, this might be a minor inconvenience. On a factory floor, where parts arrive with slight tilt, vibration, or height variation from fixture wear, parallax error compounds into real measurement uncertainty. Consider a connector pin inspection system checking for correct pin height across a 15mm field of view. If the entocentric lens has even a 2-degree angular field of view at the edges, a 0.5mm variation in pin height can translate into a measurable lateral position shift of several microns – enough to cause inconsistent pass/fail decisions on a gauge with 10-micron tolerance requirements.
No – calibration is still required to account for residual distortion, sensor pixel pitch, and any minor manufacturing tolerance in the lens itself. Telecentric optics reduce the size and variability of the errors calibration needs to correct, but they do not remove the calibration step from a properly validated inspection workflow.
Most industrial lenses with locking focus and iris rings hold calibration for years under normal conditions, but washdown environments and high-vibration lines warrant a visual and focus check during scheduled preventive maintenance, typically every three to six months. Any unexplained increase in false rejects or missed defects should trigger an immediate focus and alignment check rather than waiting for the next scheduled interval.
Sensor format compatibility is the second half of this equation. A lens designed for a 1/1.8-inch sensor will produce heavy vignetting and corner softness if paired with a 1-inch or larger sensor, even though it may physically thread onto the same C-mount. As camera manufacturers push toward larger, higher-resolution sensors to capture more of the packaging line in a single frame, lens image circles must be verified against the actual sensor diagonal, not just the mount type. This mismatch is one of the most common specification errors system integrators encounter when upgrading legacy machine vision cameras without revisiting the optical path.
Robotic guidance applications, particularly bin-picking and pick-and-place tasks involving metallic or plastic parts, also benefit substantially. When a robot vision system must locate the edges and orientation of a part reliably, glare that obscures those edges directly translates into failed grasps or misaligned placements. Polarization filtering in these systems is often paired with structured lighting to maximize contrast on features that would otherwise be washed out.
With a 4k line-scan camera operating at 50 kHz line rate, the maximum surface speed is about 2.5 m/s (assuming 0.05 mm per pixel across the log). At higher speeds, the image becomes compressed and defect detection accuracy drops. For speeds up to 4 m/s, a 8k camera at 80 kHz line rate is required, but this demands higher lighting intensity and more expensive lenses. High-quality machine vision systems can maintain performance at 3 m/s with a proper encoder synchronisation.
Is It Ever Acceptable to Use Rolling Shutter Cameras in Automation? Rolling shutter sensors are not obsolete, and dismissing them outright would ignore genuine cost and performance advantages in the right context. Applications involving completely stationary objects, such as final visual inspection of a part that has stopped under a fixed camera, gain nothing from global shutter and can achieve excellent results with a well-specified rolling shutter unit at a lower price point. Similarly, some low-speed sorting or presence-verification tasks tolerate minor skew because the algorithm is checking for gross features rather than fine dimensional tolerances.