I need to get some idea of what's causing the vertical misalignment (I'm calling the driven axis vertical, as in a portrait page). Last night I did some tests. I printed a horizontal line near the top of the board and measured the distance to the edge. Measurements were performed with a vernier caliper to a measuring accuracy of about +-0.02mm. I performed the test with and without a second guide fence, and with the edge sensor mounted to look at the trailing edge of the board, to eliminate variation at the point where the leading edge mashes into the imaging rollers. The numbers are distances in mm from the board edge to the image. The standard deviations are given in bold after the rule.
I screwed up a measurement in the second experiment (shown in italics). If one excludes the dodgy result (standard deviations non-italic), it is clear that the second fence minimises skew. The left and right SD's show that the fence also helps with vertical alignment.
Then I moved the edge sensor to look at the trailing board edge:
There is still around 0.1mm variation in vertical image position. With the sensor mounted at the trailing edge the board is already moving through the imaging rollers when the imaging begins. Therefore there are few possible sources of the misalignment:
I have pretty much eliminated the first option by printing sets of parallel horizontal lines. The gaps between the lines stayed constant down the board, so there could have been no vertical slip.
I plan to test the second option by replacing the optical sensor with a mechanical microswitch and switch debouncer connected directly to the microcontroller input pin. This should be suitably invariable as the point where a microswitch clicks is going to stay pretty constant - at least between one print run and the next.
The third option is rather worse, and could only be rectified by loading the board into the printer at a fixed time each run.
I will repeat the second and third tests to establish whether looking at the trailing edge really is better than the leading edge. Excluding the possibly erroneous result, the deviation in measurements is the same (0.08mm).
Submitted by jeff on Fri, 03/09/2007 - 15:21. categories [ ]