Deciding whether a rolling shutter is good enough for your moving part

In short

Motion alone does not rule out a rolling shutter, and a global shutter does not remove blur. Shift each reference point by how far the part moves between the exposure times of its rows in your exact mode, then recompute the dimension. Under steady light, that change must fit the shutter's share of the tolerance, and blur needs its own budget. With a shared light pulse, the pulse must be short enough, fall while all the rows you need are exposing, and be much brighter than ambient light. Confirm the choice on a calibrated rigid target at worst-case speed.

The difference is when each row collects light

A global shutter exposes all sensor rows at once. A rolling shutter starts them row by row.

All three cameras in this article use CMOS sensors (datasheet). So check their shutter type, not the CMOS label.

Rolling distortion comes from the row delay, the time offset between rows, not from exposure smear. The readout time across the rows you read depends on the model and its configuration. Exposure time and frame rate do not give it. Basler cameras report it as SensorReadoutTime.

In a conventional rolling mode, rows can have equal exposure times yet start at different moments. A moving part then comes out sheared, with straight edges leaning. What counts is when each row collects light, not how long the finished image takes to transfer.

From row timing to the error in your measurement

Speed times the row delay tells you how far the part moves between two rows. That is a displacement, not the error of every size you measure.

For a rigid part moving at constant speed across the image width, both edges on one row shift equally. Subtracting their coordinates cancels the shift, so shear alone leaves that width unchanged. Points on different rows shift by different amounts. So distances between them, angles, positions and contours need their own calculation. This is constant-speed geometry, not a camera test.

First decide what share of the tolerance the shutter may use. Then find the reference points, the edges or features the measurement is taken between, and their rows. For your operating mode, take the time between the moments those rows collect light. Multiply it by the image velocity, the part's speed on the sensor in pixels per second. Move each point by its own displacement, recompute the measurement and compare the change with the shutter's share.

For a linear size, the change cannot exceed the displacement across the whole region of interest (ROI), the sensor area you read out. If that displacement sits comfortably inside the shutter's share, row timing passes. If not, the task has not failed yet: recompute through the actual points. Passing does not prove the whole measurement accurate. Angles and contour shapes have no such bound.

What it looks like in the image, and what looks similar

With sideways motion and rows starting from top to bottom, lower rows see the part later, further along. A vertical edge therefore leans. Rotation or vibration can bend a straight spoke and make a sequence of frames wobble, because rows catch different positions or orientations. The nickname for this, the jello effect, is not a datasheet term.

Shear is easiest to see when blur is small. Put a straight edge in the scene that should stand vertical across many rows. If it stays sharp but tilts, that fits shear from the row delay. A soft edge can still lean, and a leaning edge can still carry blur.

Horizontal bands are a clue, not a diagnosis. Unstable lighting driven by pulse-width modulation (PWM) can cause them. To test this, repeat the shot under light you know is stable, with exposure, gain and trigger timing unchanged. If the bands vanish, or move with the light rather than the part, the lighting is the likely cause.

When a rolling shutter is fine for a moving part

A still part under a fixed camera is the easy case. It shows no motion distortion and needs only light that stays steady while the rows collect it.

Under steady, continuous light, a moving part needs two budgets: blur within one row's exposure, and the geometric change from the row delay. The rolling shutter passes when both fit at worst-case speed. It does not fail just because the camera has no documented flash window.

A shared light pulse is the other route, recommended for fast motion on a rolling shutter (Basler guidance). It works only when all four hold:

• Every row you need is light-sensitive at the same time. On models that document it, Basler's Flash Window signal marks that interval.
• The pulse is short enough for the blur budget.
• The pulse delay and its jitter, the small random scatter in that delay, keep the pulse inside the interval.
• Ambient light is too weak to leave a second, smeared copy of the part outside the pulse.

The exposure and strobe article walks through that setup.

Global Reset Release starts all rows together but ends them, or reads them out, one after another. Without a flash or another way to end all rows at once, brightness can change down the frame. A moving part can still distort. MV-CT0MU0060/R offers Rolling and Global Reset exposure modes (datasheet). Treat that as a timing option to check against your speed, not as the answer.

A global shutter removes the row delay but still needs a blur budget.

Worked example: a rolling-shutter camera on a conveyor, blur only

Parts ride a conveyor past MV-CA060-11GM. The table lists the inputs and our calculator's results.

Sampling is the number of pixels on each millimetre of the part. The calculated exposure ceiling lies inside the camera's exposure range, but that comparison alone does not prove the scene works.

These results describe blur within one row's exposure, not shear. For the displacement between two rows, multiply the 30 700 px/s image velocity by the time between the moments they collect light. Take that timing for your exact mode from the model documentation, or send us the model and we will confirm it.

Calculate for your case

The prepared link opens the calculator with the camera, both field dimensions, the working distance, the speed and the motion axis set. The 1 px allowance does not travel in the address, and the page has no field for it. To reproduce the example exactly, use the optics route of the tools API. That request returns values.maxExposureUs of about 32.6 µs (calculated), from the camera's 3072 px over 100 mm.

The page itself shows 20 µs. It also takes two assumed inputs from the address: a 0.1 mm feature and 5 px across it. That requires 5000 px across 100 mm, or 50 px/mm, and gives 20 µs at 1000 mm/s with a 1 px budget (calculated). So 20 µs is what the task demands, and 32.6 µs is what the selected camera delivers. Neither number models the shutter.

Equipment that fits this example

MV-CA060-11GM is the rolling-shutter camera of the worked example. MV-CT0MU0060/R has the same 3072 × 2048 px at 2.4 µm and adds Rolling and Global Reset exposure modes. MV-CA050-12UM is the global-shutter option: 2448 × 2048 px at 3.45 µm, with no row delay. All these figures come from the datasheets. On all three, check the mount, interface, cable and lens image circle separately.

Common mistakes

• The calculator's exposure ceiling is taken as proof that the geometry is safe. Edges come back crisp, yet the measured dimension drifts with line speed. Before you accept the result, work out how the row delay shifts your reference points.
• The frame period, one divided by the frame rate, is used as the row timing, or full-frame timing is carried over to a cropped ROI. A correction built on either leaves a leftover lean that changes with the crop, because readout time can depend on the ROI height. Take the timing from the model documentation for your ROI.
• A short pulse is trusted just because it is short. If it misses the window when the rows are light-sensitive together, part of the frame goes dark or shows a brightness step. Check the pulse, its delay and its jitter against the documented common-row window.
• Global Reset is treated as a global shutter. Its rows can end at different times, leaving a brightness gradient and distorted moving shapes. Check geometry and blur under continuous light, or validate the shared pulse as described above.

When this rule breaks

The simple lean assumes one constant speed along one axis for the whole frame. Acceleration, vibration and rotation break that, and no single skew factor can then correct the error. The same-row cancellation also holds only in the constant-speed model. It says nothing about lens distortion, calibration or the algorithm that finds the edge.

To correct the geometry in software, you need four inputs: when each row involved collects light, the ROI, the readout direction and the motion vector. Confirm that timing for the exact ROI and mode you will run.

Next step

Decide it on the machine. Put a calibrated rigid target in the scene and move it at worst-case speed and direction, inside the ROI you will read. Compare the measured dimensions with its known values.

For a number before that test, use the camera documentation. Find the effective exposure start and end of the rows carrying your reference points, and the gap between them. These values must match your exact settings: shutter mode, exposure, ROI height, resolution, frame rate and readout direction. Also find out how the ROI changes that timing. For the pulse route, you also need the documented common-row or flash-window timing. Send us the camera model, the ROI, the part speed and direction, and the dimension with its tolerance. We will confirm what the model documentation states.

Related

• Setting exposure for a moving part, and deciding whether to pulse the light
• Deciding how many pixels a feature needs before you choose a camera

How this material was prepared

Prepared by MVisionPro from the stated sources and the MVisionPro calculation engine. A physical test is claimed only when the material says so explicitly. Read the editorial method.

Editorial status: verified. Content updated 2026-09-23.