When a vision station underperforms, the instinct is to make the light brighter. In practice, brightness is rarely the limiting factor. The direction from which the light arrives, and the angle at which it leaves the surface toward the lens, decide whether the defect appears in the image at all. Choosing the wrong geometry and then turning up the current is the most common way to waste a week of commissioning.
Start from the defect, not the part
Every illumination choice is really a question about the defect's geometry. Ask three things: is the defect a change in reflectance (a stain, a print error), a change in surface slope (a scratch, a dent, a burr), or an absence of material (a hole, a gap, a missing pin)? Then ask whether the surrounding surface is matte or specular. Those two answers select the geometry almost unambiguously.
Ring light: the default, and often the wrong one
A ring light surrounds the lens and throws light onto the part from all sides at a shallow-to-moderate angle. Diffuse ring lights are the general-purpose choice for matte surfaces with reflectance defects: a dark stain on a matte plastic part will appear as a dark stain under almost any illumination, and the ring light is cheap, small and easy to mount.
Where ring lights fail is on specular surfaces. A machined aluminium face under a ring light returns a bright annulus with a dark hole in the middle, because the ring is reflected directly into the lens at one specific viewing angle. Scratches running in one direction may vanish entirely in one orientation and appear clearly in another.
Bar lights: control the direction
Bar lights let you choose azimuth independently from elevation. Two bars on opposite sides at 45 degrees produce fairly even illumination on curved parts. A single bar from one side produces strong directional shadowing, which is exactly what you want for detecting raised features, embossing and tool marks: the shadow is the signal, not an artifact.
The classic configuration is a low-angle bar grazing along the surface. At an angle of a few degrees, a flat surface reflects the light away from the lens and appears dark, while any scratch, edge or particle scatters light toward the lens and appears bright. This is dark-field illumination, and it detects the smallest surface defects of any geometry - at the cost of showing you every speck of dust as well as every real defect.
Dome light: uniformity and specular suppression
A dome light is a hemispherical diffuser with the camera looking through an aperture at the top. Light arrives at the part from every direction in the hemisphere, so a specular surface has nowhere to put a mirror reflection: the bright spot is spread out into a uniform field. Domed illumination is the standard answer for shiny, curved and complex parts - turned metal, chrome-plated fittings, polished bearings, solder joints.
The trade is contrast. Because the light is omnidirectional, slope changes produce almost no shading variation, and a dome light can make a part look flat and featureless. If the defect is a shallow dent, a dome is exactly the wrong choice. If the defect is a colour or reflectance change on a specular part, it is exactly the right one.
Coaxial light: for flat, reflective, planar work
Coaxial illumination sends light straight down the lens axis via a half-mirror. Only surfaces perpendicular to the axis reflect it back into the lens, so a flat wafer, a polished metal face or a printed circuit board lights evenly while tilted or recessed features stay dark. Coaxial lights are also the correct partners for telecentric lenses, which accept only on-axis rays and therefore waste off-axis illumination.
Their weakness follows directly from their strength: because only the axis-aligned component is used, coaxial light is very inefficient, and it is close to useless on a rough or curved surface where almost nothing is perpendicular to the axis.
Putting it together on a real station
A compact mono station - an MV-CU013-A0GM with a moderately wide lens - can usually try ring, bar and dome configurations in an afternoon because the illumination is the cheapest part of the system. Do that experiment before finalising the mechanical design: the right geometry often lets you relax exposure, use a cheaper lens and use a smaller sensor, while the wrong geometry can make a 20 MP camera look worse than a 1.3 MP one.
The rule that survives contact with real projects: choose geometry from the defect's physics, choose intensity last, and never select a light source by lumens alone.