Dimensions off on diagonals

My steps/mm are dialed in on X and Y, moves are accurate to within .05mm. Also I have elimated backlash from X and Y as well, no more than about .03mm as measured with a dial gauge. 20mm squares come out to 19.98-20.01. However I noticed that whenever I did a cylinder, there would be a bulge to it, as much as .5mm. Like the cubes, the dimensions would be correct straight up X and Y but too large on the diagonal. I turned a cube 45 degrees and printed it. The width from front right to back left was long by .3mm, and the width from font left to back right was short by .2 or so. It seems that something is wrong when both axis move together. Is there something that would cause this despite the steps and backlash already being calibrated?

You will loose dimension to material shrinkage, and a cylinder will cool dimensionally different than a cube due to varying tension between layer angles and the curve on the inner and outer perimeters.

I did have my machine calibrated with a dial indicator down to .01mm in the X and Y, and could mount a pen to the X carriage and draw dimensionally perfect (+/-.05mm) circles and squares. But when i started printing, material shrinkage changed the dimensions of the print.

Also had the different measurements between X and Y you are getting. To compensate for shrinkage, I increased my steps/mm from 63.xxxxxx to 64 on X and 63.5 on Y and reconfigured backlash. Mechanically, with the same belt and gear set for X and Y, steps/mm should be the same, but skrinkage is dimensionally different.

I had to compromise between a cube and a cylinder dimension, printed in the same run next to each other my cubes are slightly over dimension, and circles are slightly under. Nature of the beast.

I would much more prefer to be mechanically dimesionally acurate, but none of the current slicing software compensate for material shrinkage.

I don't think it would be the difference between shrinkage of a cylinder and a cube, since my problem is diagonal dimensions being too long. Also the two cubes were the same shape, but printed at different orientations on the bed, One was straight on X and Y, with perfect dimensions. The one printed at a 45 degree turn had dimensions that were too long on one diagonal.

That may be backlash compensation being too much. I seem to get variances in backlash relative to print speed.

Edited 2 time(s). Last edit at 11/19/2012 11:58AM by Dirty Steve.

I've found when checking my steps on X that the moves are good to within .02mm at the right side of the printer. If I turn the dial gauge around and measure against the other side, the moves tend to be .1mm too long. The moves are spot on for Y at the front of the printer, but due to the design of the mount and the extruder, I can't test it against the back. I suspect that there may be inconsistency in Y as well.

What could cause the same # of steps to cause a different length in moves in opposite directions?