I could go on for a long time about calibration, a really long time, but I'll throw in some quickies here. It's really nice to have basic ideas on some of these numbers when you start. Like Andy said, ballscrews come in a number of common pitches and they're likely to be metric(the ones on my Fadal were 8mm and 10mm). Encoders are almost always powers of 2, so you're going to be expecting 1024/2048/4096/8192 counts per rev on most commodity encoders.
Order of operations (assuming you tried to build it square and flat in the first place):
-Rough scale X and Y until they both move the same distance for the same input (ie: 16" on X = 16" on Y...actually doesn't matter if it's really 16", though!)
-Rough level the table (stick a tool in spindle and adjust the table until it's the same height everywhere, or machine the table at constant Z height, or both)
-Tram the Z axis (stick a bent something in the spindle, make sure that something stays the same height over the table everywhere when you spin it)
-Rough scale Z (I like to jog and measure with a ruler, then fine tune by cutting a slot or hole of specific depth)
-Check square by drilling four holes in a big square on the table and measuring diagonals, adjust the machine until it's actually square or use software compensation
-Start over from step 1 until your square is square, on size, and a 2" deep hole is a 2" deep hole.
Tools that are just great to have for all these processes:
-A very solid 36" steel ruler. A very sold one meter ruler would be even better, but they don't sell 'em at Princess Auto for $5!
-A long level, for a big machine
-Some sort of tram gauge (a stiff bent wire works well enough...)
-1" Gauge Block
-Decent 6" steel ruler
-6" Vernier/Dial/Digital Caliper
-A dial indicator and a dial caliper (Shars)
-A 1-2-3 block
Tools that are SUPER to have for some of these:
-A tool setting gauge (Shars)
-A decent square (Carpenter's square < Machinist's square < Big Machnist's square < Granite machinist's square)
-A really good short ruler (I have a 6" ruler from Fowler that's graduated in half-millimeters on one side and decimal inches down to 0.01" lines on the other. And it's got a gold coating.)
-A linear scale with DRO (get a one axis lathe DRO from Shars, use it as a super ruler). You can now measure backlash and screw error!
-A really good level (Carpenter's level < Machinist's level < Precision Level < Two axis digital master level I'm so happy I had an excuse to get from McMaster-Carr)
-A real tram gauge (Shars)
If you're calibrating a linear scaling value, like steps or encoder counts per inch, then the longest rigid calibration scale you can use is generally best. A 4' long ruler with 1/16" lines on it is as good as a 1' ruler with 1/64" lines on it: they'll both measure an error to better than 1.5%. But the 4' ruler is better because the other errors inherent in measurement (deflection, user reading error, tool location, list goes on forever) are mitigated more over the longer distance. A nice solid 3/32-1/8" thick 3' steel ruler is one of the greatest rough calibration tools there is for scaling -and- flatness.
Get your axes scaled to eachother, that's step 1, then you can cut a square. It actually doesn't matter if your ruler is really 3' long, so long as both axes read the same when they move the length of the ruler. If you can make a square square, you just need to multiply both axes movement by the same number to make a square on size square. Rough level the table against a tool in the spindle (either jack it to level, or use a tool in the spindle to surface the whole thing at a constant Z height). With a roughly level table, you can tram the spindle with whatever's around (I have a two-indicator tram gauge now, but a piece of bent wire will get you dang close with a little patience).
After all this fun, you can then check out how badly things go out of whack due to mechanical errors. Set a 10lb weight on your gantry with the spindle in the middle of motion and see how much it moves down. Set a 20lb weight in the middle of your table and check the same. Do backlash tests. Do backlash tests at rapid speed. Cut a circle and see how circular it is (can be mechanical and/or control in this case).
OK, maybe that went on a bit longer than I was expecting. I'm really into calibration. 
|
