Two cameras, same resolution, same price bracket - one produces clean images of a moving part, the other bends it into a parallelogram. The difference is shutter architecture, and it is the first thing to check after you have settled resolution. Get it wrong and no amount of lighting or software will fix the artifacts.
How the two shutters differ
A global shutter exposes every pixel simultaneously: the sensor opens one electronic "curtain" across the whole array, captures, and closes. Motion stops dead, geometry stays true.
A rolling shutter exposes row by row, sweeping down the sensor in sequence. Each row is captured a few microseconds after the one above it. For a static scene this is invisible. For an object moving during the sweep - typically tens of microseconds to a few milliseconds for the full readout - every row photographs the object in a slightly different position. Vertical edges lean, rectangles become trapezoids, spinning parts wobble. The effect is called rolling shutter artifact, and it is a geometry error, not noise: no filter undoes it.
Which common sensors are which
Among widely used Sony machine vision sensors: the IMX264 (5 MP), IMX296 (1.6 MP) and IMX430 (2 MP) families are global shutter; the IMX183 (20 MP), IMX226 (12 MP) and IMX178 (6 MP) families are rolling shutter. That maps directly onto real cameras - the MV-CS050-10GM, MV-CS016-10GC and MV-CS020-10GM are global shutter workhorses, while the high-resolution MV-CS200-10GM, MV-CU120-10GM and MV-CU060-10GM are rolling shutter designs. Notice the pattern: global shutter lives in the small-pixel, moderate-resolution range; rolling shutter dominates where sensor makers pushed resolution or sensitivity per dollar.
So when is rolling shutter fine?
Whenever the scene does not move meaningfully during the sweep. Static inspection stations - parts presented on indexing tables, parts under a fixture, anything photographed with a strobe that freezes motion in microseconds - can use rolling shutter sensors without penalty, and they cost less per megapixel. Trigger-and-freeze with a short flash converts a moving line into a static scene; if your process allows strobing, rolling shutter is back on the table.
The failures come from continuous motion captured with ambient or continuous light: conveyors without strobes, web inspection, robot guidance mid-motion, vibration-heavy environments with long exposures. Also beware partial motion - a part that is static except for a wobbling fixture will show it.
The decision procedure
Estimate object displacement during exposure plus readout. With global shutter, only exposure time matters: displacement = speed x exposure. At 300 mm/s and 100 us, that is 0.03 mm - usually acceptable. With rolling shutter, add the sweep: displacement = speed x (exposure + readout time), and readout on a 12 MP sensor can approach a millisecond, which at 300 mm/s smears 0.3 mm. Same speed, ten times the error, purely from architecture.
So: continuous motion without strobes at more than a few tens of millimeters per second - global shutter. Static or strobed scenes, and any application where you need very high resolution at low cost - rolling shutter earns its place.
One nuance: shutter efficiency
Modern global shutter sensors differ in how completely the "curtain" blocks light during readout. Poor shutter efficiency leaks light into the readout and shows as smear in bright scenes; Sony's Pregius-class sensors (IMX264, IMX430) are well regarded here. It is a datasheet line most buyers skip - check it if your parts are bright and fast.
Our camera listings mark the shutter type for every model alongside sensor and frame rate, so the global-versus-rolling decision is a filter click, not a datasheet hunt.