Resolution
How many pixels the sensor has across and down. More pixels across the same view put more pixels on each feature, at the cost of a larger image to send and process on every unit.
Get started
×Tell us about the line. A person reads it and replies.
Or talk it through first:
30-minute Google Meet call, Monday to Friday, 9 am to 5 pm Pacific.
Need a full study of the line first?
Resources
Estimate the focal length, the angle of view, and the size of one pixel on the part. Start from the sensor, the distance to the part, and how wide a view you need.
Common sensor sizes fill the width and height. Replace those numbers with the size on the camera sheet. Use a period for decimals, as in 2.5.
Working distance is the space from the front of the lens to the part. The width of the view is how much of the part fills the sensor from side to side. The height and the diagonal follow the shape of the sensor.
The focal length is shorter than the sensor width, so this estimate is less accurate.
This is a first estimate. It ignores distortion and the space inside the lens. One pixel on the part is the pixel size divided by the magnification. A real feature usually takes at least two pixels, and the lens has to resolve them.
Enter the lens focal length and the working distance. The sensor size comes from the first calculator above.
Use this to check whether a lens on the shelf covers the part, and how many pixels the camera needs for the smallest feature.
How much image data one camera sends, and which camera link carries it.
Usable link speeds are rough planning numbers: about 115 MB/s for GigE, 580 MB/s for 5GigE, 360 MB/s for USB3 and 1,150 MB/s for 10GigE. Leave headroom.
How many pixels the sensor has across and down. More pixels across the same view put more pixels on each feature, at the cost of a larger image to send and process on every unit.
A larger pixel collects more light in the same exposure, which helps at short strobe times. Sensitivity changes with the color of the light. The quantum efficiency curve on the camera sheet shows the share of light the sensor converts at each wavelength, which matters most for infrared and ultraviolet light.
A monochrome camera gives the sharpest detail and the most light per pixel. Most color cameras put a mosaic of red, green, and blue filters over the pixels and fill in the missing colors from neighbors, which softens fine detail. Pick color when color is the fault.
An 8-bit pixel has 256 gray levels, enough for most checks. High dynamic range (HDR) combines a short and a long exposure, so bright foil and dark print both keep their detail in one picture.
C-mount is the common thread on machine vision cameras and lenses. CS-mount uses the same thread with the sensor 5 mm closer to the flange. A C-mount lens fits a CS-mount camera with a 5 mm spacer ring; a CS-mount lens on a C-mount camera stays out of focus.
GigE Vision runs over Ethernet with long cables and many cameras on one switch. USB3 Vision keeps cabling short and simple. CoaXPress and Camera Link carry the fastest image streams through a frame grabber card. GenICam gives software one way to control cameras from different makers. Check that transfer time stays steady, because the reject timing depends on it.
The light matters as much as the camera. See Lighting and optics for inspection and How machine vision works.
FAQ
It is a first estimate from simple lens math. Lens distortion and lens design change the real numbers, so test with the actual lens.
At least two to three pixels across the smallest feature, and more for reliable measurement.
Each camera link carries a limited amount of data. If images arrive faster than the link carries them, frames are dropped.
Pick the one with comfortable headroom over the data rate, and allow for cable length on the line.