3
« on: May 15, 2011, 04:32:37 PM »
It seems that after searching through the intertubes, nobody seems to have posted any instructions for getting the most accuracy out of a MaxNC/MachIII setup. I'm talking about properly setting up backlash compensation and properly “aligning” the spindle on this machine (MaxNC15) so it is perfectly orthogonal to the X-Y table. I guess I'm not surprised by the fact MaxNC has no such instructions, but surely others out there have run into accuracy issues over time. My journey began when I crashed my Z axis, and suddenly woke up to the wonderful world of alignment issues.
In the end I went from roughly .01” of change for every inch of Z I plunged. Maybe this is acceptable to some, but for me it is NOT. I do crazy things like cut solder stencils with .005” apertures. I want the advertised accuracy out of my machine! After I did the calibration I'm about to describe, I can actually get the accuracy and repeatability claimed by the manufacturer of .00025"!!
First off, I wanted to get rid of backlash so this didn't mess me up through what I'm about to describe. MachIII has some nice dialogs for modifying these parameters, and unlike the naysaysers out there, I'll make the wild claim they actually work WELL. I take a value much higher than the actual expected backlash... say .01", plug that in, then go into JOG (tab) mode / 1% and hit left-right very briefly and alternately on that axis. I also use my favorite indicator setup so I can see how much the spindle has moved on the axis. If you have too much backlash configured, the needle jumps. I then lower the backlash value it until the needle doesn't move at all when you hit left/right. Repeat on the other axes and you've gotten some additional error fixed thanks to MachIII. Out of the box, my MaxNC15 had .003" of backlash on X and Y but now ZERO thanks to this feature. Note that I'm talking about pulsing left then right very briefly so the s/w only has enough time to take up the programmed backlash and nothing more. You'll see the steppers moving but the table won't.
The following are sources of error in the MaxNC mill as they relate to the tool:
1) Wobble from the tool (runout)
2) Misalignment between the tool's rotational axis and the axis of travel (Z).
3) Lack of orthogonality between the tool's rotational axis and the X-Y table.
You can't just lump these all together and fix them at once. EACH one must be accounted for seperately or you'll go mad (like me). I'll admit that I'm a n00b here, but what I'm about to describe worked for me. Maybe you have a simpler solution. If so, let me know!
n my machine, I noticed the issues when trying to use an edge finder to locate the center of bar stock. I would touch the edges on X divide by 2, set zero. Same for Y. Theoretically I now have zero. However I was always off. This drove me mad until I realized the three issues above. Esentially the edge finder was tilted relative to the table, finding the center at a skewed and incorrect location.
Here's how I addressed each issue:
1) Runout: Not much you can do about this. Make sure your collet is tight. I have the original TAIG style spindle with the ER adapter and runout is about 3 mills. I measure this by putting a longish 3/8” shank bit (has to be smooth!) in the collet and rotating it by hand. I put a dial gauge anchored to the mill with its point on the bit, and rotated. My machine has .003” - .002” runout/ depending on how well I did inserting the tool.
2) Misalignment of rotational axis vs Z movement axis: This can be measured on X and Y independently. Using the dial gauge on an arm (affixed to either the table or the bench, doesn't matter) touch the front of the bit for Y changes, or the sides of the bit for X changes. When jogging the machine up and down, you'd ideally want to see ZERO change. Unfortunately you will see the #1 and #2 error sources, so you need to get rid of #1 (runout / tool not straight) and #2, actual misalignment. Rotate the tool and get the max or min deflection on the needle. Then rotate 90 degrees. This will temporarily get rid of the #1 error for the measurement you're about to do. Why? Because the tip of the bit is basically offset (aka wobble/runout) and draws a circle as you rotate. 90 degrees from the max or min rotates that error so you don't see it. Then when you jog up or down you don't get the runout error since the runout is now pointing perpendicular to the measuring device. Now we can fix error source #2 without source #1 messing with us. So for #2, I found that this source of error was 0 in the Y direction most likely to the way the machine is constructed. X however had 5 mils per inch of delta. I fixed this by loosening the 4 screws holding the spindle and applying some force as I re-tightened the screws. After doing this and re-measuring a few times, both X and Y were ZERO delta over a few inches. Perfect!
3) Lack of Z orthogonality: If all that weren't enough, I made the Z axis tilt relative to X and Y because I crashed once and popped the “Z fuse” so-to-speak. Turns out since the Z axis sits on the Y axis FLAT, and because rotation of the Z axis along the long axis just adds a fixed offset, the only misalignment that matters is front-back movement. I fixed this by putting a dial gauge in an Indicol arm in the spindle and touching the side of a perfectly machined block on the table . Make sure to touch the block pointing toward the back or the front of the machine so you get the tilt! Jogging up-down changed initially over .01” per inch. I loosened the screws and slightly moved the tilt of the Z axis slide (motor and all) forward or backward bit by bit while jogging up and down. Eventually I saw ZERO change. I tightened the screws down and retested. Perfect!
I hope this helps someone. I don't know why I can't find anyone else doing this with their MAXNC, nor do I understand why the instructions for aligning the machine weren't included. Maybe this is common knowledge?
I've attached an image of a solder stencil I cut.. the apertures are roughly .009" wide. All done on a maxNC15! The stencil was so accurate I was actually able to lay it over the circuit board, hear a click and it attached itself to the pads!
Guest
Show Posts
