Last CNC post was mostly the physical build. This post is system integration, which turned out to be much more tedious. You can see it running at:
First run on wood
First I had to gather the last of the parts. I bought a Dell Precision mid-tower on Ebay for $90 free up my. Laptop. I added 8 more GB ram and a spare solid state drive I had.
I tried a lot of different limit switches. Stay away from the hall effect sensors. The state of switching those proving isn't really discrete enough to give the smooth stepper board a definitive on/off. After about a month of trying to get those to work, I gave up and bought some high quality spring rocker switches. For now, I am just running them for homing with soft limits on the other side.
I trammed the solid column using brass shims to 1.5 thousandths from left to right across the x axis and when I push on the head it will only wobble about +-1/2 thou. I calibrated each axis using the step calculator wizard to well within a thousandth repeatably. I discovered that you need to run your test in the same direction till you nail it. Then your next incremental move in the revered direction. The error here is your backlash. I found XYZ had 1.25/3/3 thou respectively. The software compensation corrected this.
I know its always best to eliminate the source of backlash so only compensated after tearing the entire machine apart. I discovered that ~1 thou will be inherent in each axis because the thrust bearings are not preloaded more than the very well machine housing provides. I accept this slop because they are so well done I can count on the software to fix it.
Now z axis had about 3/4 thus more backlash reversing to move up than it switching down. It took me some thinking to realize it is because it's the only axis with a belt. It is being stretched dragging the head up. That combined with a head weighing 3x more than stock, created the discrepancy. So I psudeosplit the difference and added compensation of 3.5 thousandth because Id rather have the to be too shallow than too deep. I'll add the 0.25 in my g code.
The y axis has a problem I don't know how to fix. The ball screw is pinned to the slider base with a single set screw. Physically it doesn't move linearly but when you reverse directions the nuts swings about a quarter turn before being engaged. I've got to fix this out later. It's hard to get a set screw or something in there because it's in the very heart of the machine with at least 2 inches of iron surrounding in either direction. The software seems to work great for this as well because the source is predictable.
Going for my first metal chips in 6061 the next dragon reared its ugly head. I tried running a 6mm carbide 4 flute at 10ipm, max rpm (~2700) in a 50 thou depth of cut. At first, it sounded great but it slowly started stalling because the belt was slipping. It is because my spring motor tensioner isn't enough. Back to ths drawing back. I'm going to be replacing it with a 1/2 nylon stainless steel bolt with a lock nut for some springyness. I'm going to be adding a secondary drive pulley as well to bump up the 3.7:1 ratio I'm running. The bearings and grease can handle 8k so I'm going to shoot for 2:1 to 1.75:1.
