First, thank you for any help on this issue. I have had on three separate projects the final cut out go off track. I had this issue twice and then did the fifty-hour maintenance. The next two projects, one was a 30% longer project, the other exactly the same project, came out perfectly. The image of the last project, had the board pressure tested (80 lbs), the board was longer and wider than required and it was tapped on the bottom edges with blue painter's tape.
Any suggestion on how to keep this from happening again?
Dan has a theory that when cutting near the brass roller vibration is induced into the encoder and the reading becomes unreliable, so the software ignores it and switches to the x-drive reading which is not as precise. I have had problems like yours when cutting a board with no jig. When I use a sled (3/4" base board with 1" rails and 4" end pieces) the cutting bit is always at least 1" from the brass roller, and the mass of the sled seems to dampen vibration. I get reliable cutouts using a sled.
There are tooth marks on the tape the entire length, and they don't look like there were any variations each time the board passed. Thank you for the idea on the sled, I have the plans for a sled, I guess it is time to make one.
Is this multiple passes on one cut through? I don't mean to make light of Dick's suggestion, but I wouldn't think the x motor reading would be off by that amount unless there was bad tracking involved as well.
Are you using out feed rollers? A board that long could bow, losing contact with the roller just enough to go haywire, the teeth aren't that tall.
Also check the level of the machine out feed arms. They should be a hair below your board level. If one is up, and you have droop on the opposite side, that could cause it to lose contact.
I think there could be something else going on.. but, I could be out to lunch!
That being said, all of my cutouts are done on a sled, started because I didn't like the pile of sawdust in the gap to clean out...now just habit- I use 1/2 in mdf.
Here is an experiment where I start a multi-pass cutout, stop the machine, move the board off of the brass roller, and then continue the cutout. The result is similar to the result posted. To me, this indicates that the brass roller data was being ignored by the controller during some of the x direction cutting. As Dick has mentioned, I believe it has to do with vibration. A straight cut up the edge of the board near the brass roller seems to frequently result in this type of behavior.