A hand calculation from web width, line speed and smallest defect to line-scan pixels and lines per second, plus speed changes and IntScan Pro parallel…
How should line speed, width and minimum defect size be organized to review paper line-scan inspection?
Three numbers nearly decide the configuration of a line-scan inspection machine. Web width sets the pixels across the width and the number of cameras, line speed sets the lines to capture per second, and the smallest defect you want to find sets the size of one pixel. Write the three values in mm and m/min, giving both the minimum and the maximum speed.
Work it out by hand once and you can read why a quote lists the number of cameras it does. You need only three numbers: web width, line speed and the smallest defect you want to find.
First, the principle in brief. The sensor of a line-scan camera has its pixels in just one row. As the paper passes, it captures one line, and after the paper has moved a little it captures the next, stitching them into one image. So the resolution across the width is set by the sensor's pixel count, and the resolution in the running direction by the lines captured per second (line rate) and the line speed.
Illustrative example: to see a 2,000 mm width at 0.25 mm per pixel, you need 8,000 pixels across the width. At a line speed of 600 m/min, 10,000 mm pass every second, so matching 0.25 mm in the running direction as well means capturing 40,000 lines per second.
How do the numbers change if you want to see down to 0.1 mm on the same line? 20,000 pixels across the width and 100,000 lines per second. You have only cut the pixel size by more than half, yet the pixels across the width and the lines per second each become two and a half times as many, and the pixels to process per second grow by their product, more than six times. So rather than writing the "smallest defect" smaller just for margin, it is better to write the size you truly have to catch. Actual pixel counts and maximum imaging speeds differ by camera model, so the camera specification and count are decided in consultation.
If one camera does not have enough pixels across the width, cameras are placed side by side with their seams overlapping. IntScan Pro inspects one width with 1 to 5 cameras in parallel.
Even if the calculation is right, a defect that covers only one or two pixels is easy to miss, because too few pixels stand out from the background. How many it takes to find a defect reliably depends on the brightness difference between defect and background and on the defect's shape, so it is confirmed by imaging samples. Black foreign matter on white paper is clear with just a few pixels, but faint contamination of the same size may not show up well even with more pixels, and the lighting may need another look first.
If doing this calculation yourself is a chore, choose paper in the quick check of the Optics Lab and enter the web width, the speed and the smallest defect. If speed is the limiting condition, it tells you so along with the lines to capture per second; if the conditions are hard for an ordinary line-scan camera to keep up with, it says that too. It is an estimate from a simple uncalibrated model, so whether it is really possible is confirmed with samples.
Writing only the normal running speed is only half the answer. Also tell us how often acceleration and deceleration, roll changes and stops occur. If lines are captured at fixed time intervals, the image stretches lengthwise when the line slows and shrinks when it speeds up. A round particle gets imaged like an elongated smear.
So a common method is to measure the distance the paper has moved with pulses from an encoder on a roll shaft, and capture one line per fixed distance. IntScan Pro also links to the trigger and encoder and images following the movement of the running material. Let us know whether there is a roll that can take an encoder, and whether the signal from an encoder you already use can be shared.
At a position like a rewinder, which stops at each roll change and then ramps up again, this section is especially long. The paper inspection listed among the applications of IntCloud Optic is also rewinder inspection, and it is still at the PoC stage.
Lighting varies with the paper grade, color and whether it is coated. Where there is a lot of paper dust and other dust, it can build up on camera and lighting windows and blur the image, so we also look at whether the position is easy to clean. Write down the space and height for the camera and lighting too.
Decide in advance, too, what happens after a defect is found (marking, recording the position, alarm). The summary of the paper manufacturing defect inspection use case likewise describes detecting foreign matter, holes, wrinkles and contamination in real time with line scan to stop defects early.
For how to read test results, see how to check missed defects, false rejects and throughput.
Quality items that surface imaging cannot reveal, such as basis weight, thickness and moisture, need other measurement methods. The calculations above are only examples of unit conversion, not the specification of any particular machine. A correct calculation does not mean a defect is visible, so the final judgment is always made from images of your samples.