Kayvon wrote:How are you measuring your triangles? It's really odd that everything else is perfect, but the triangles aren't.
Yeah, I wonder if he's ignoring that the bit radius means it can't get all the way into the corners. Such as in this diagram, where the vector is in red, and the purple shows where the bit actually cuts. Might this be what's going on?
That's a great idea to periodically check that the machine is still put together after a lot of hard work
![Idea :idea:](./images/smilies/icon_idea.gif)
.
If you're going to measure triangles, you might get better information checking for smoothness along the hypotenuse. Another test would be a very slightly angled slot, say 0.010" displaced in the Y-axis, over 10" in the X-axis. And then another one in the other direction. Make the slot a little deeper than you might. Plunge a tiny divot with the bit just off the two ends, then cut the full slot, with a slow plunge. Two things to look for: if you can see the 0.001" increments in the small-excursion axis. make sure that these increments are evenly spaced. Also, if it's like a normal shark, the center of the at-depth slot will in fact NOT be in a direct line between the two divots, due to bit, carriage and gantry deflection. Keeping a watch on that can catch if there might be loosening bolts in the structure of the gantry and Z-axis carriage.
And, of course, check that your pocket depths are accurate; I didn't notice you testing any of them, but maybe you did.
If you have the ability, check the roundness of the circles. If you have something you know is round, make the hole exactly that size, and see if the circumference is actually round. There will "always" be "some" backlash in the system, even if it's under 0.001". If you can detect any flatness on the X-parallel or Y-parallel portions, this will indicate the amount of backlash in the lead-screw following anti-backlash T-nuts. One might argue that if you can see a flattening, that's too much backlash. But, your projects will decide how much inaccuracy is too much.
One final note: (kindly ignore if you're already using a "finishing pass") you saw the mention of deflection above, right? If you're really, truly needing thousandths-accurate holes, then you're going to want to cut it in two passes (at least): a primary pocketing path that leave ~0.005-0.015" inches of material margin, and then a finishing pass to take that last little thickness. Those are very useful in real work as well, but for testing machine dimensional accuracy, you're going to get the best results when the last pass, that establishes the final edge, has almost no cutting stress on the system. Trust me, I work in metals with the Shark, and deflection in the range of 0.020" and more is not uncommon when the cut is especially hard.
Regards, and thanks for the idea!
Thom