First, a little background. The part is a spacer on the Y-axis of the milling machine that holds 24mm diameter ball bearings on both ends. The two bearings are built from what looks from the outside like a washer on either end and a set of ball bearings in a frame (they can't fall out) between the washers. The washers have a groove on them that the balls roll in. Also, the bearing sets are not exactly the same size: the inner bearings have a slightly larger inner diameter hole than the outer bearings. I screen-captured the block with the front bearings in place and the back parts of that bearing set in place on the shaft from the video provided on the CNC conversion DVD I'm using as my source. Just to emphasize, what I'm machining is the red block. It goes on the leadscrew shaft below it.
The reason for the .360" deep, .945" diameter counterbore is to hold these bearings. It seems like the most important dimensions here are the depth of the counterbore and that the bore is in centered properly. The slop in the outer ring seems more cosmetic than anything, since the races can't shift around and the ball bearings are also captive. At one point in the video, Hoss says to make sure there's enough room for the washers to rotate in the bore to ensure they can rotate around the shaft. I'm not sure how much slop there should be, but the bearings are said to be "24mm", which is 0.9449 inch and 0.945 is only 1 tenth (of a thousandth) bigger than that. I'd imagine it could comfortably handle a few thousandths.
Yesterday, I decided to see how the hole would work out with my corrected G-code, which should center the hole better, but leave the gap (Wednesday's goof), on the left side. Everything was still in the same fixture from Wednesday, so I ran the corrected G-code. (I spent Thursday watching all the videos on how this works - again). That's when I realized I didn't have any way to measure that now oblong, weird hole, to make sure it was the right diameter. I put that on hold and turned a piece of 1 1/2" aluminum bar down to 0.945, about half an inch long, as a homemade gage plug. It didn't fit. I noticed that while the circle was being cut, it would hesitate at the cardinal points. If the 0 degree point was on the positive X-axis, it would pause - briefly but perceptibly - at 90, 180 and 270 degrees. The 90 and 270 degree points were too tight and I could see the edge was actually too far in at those points - it wasn't a smooth circle. I started trying to round them to shape with a half round file, and then realized I could put the end mill in position to smooth out those points and did that manually. The plug then fit in place. Right side:
Finally, I turned the block over to do the other side (it's dimensioned exactly the same), re-verified my 0,0,0 point, and ran the corrected code. The plug didn't fit! I was able to quickly determine that the milled circle was too small by around 10 thousandths radius - which is a pretty big error. I then spent a little while closing in on this size by bumping the starting point to the right and extending the radius of the cut in the circle cutting command, enlarging the circle.
In the "lessons learned" department, I noticed several things about my existing CNC system while doing this that I need to investigate. Even though it was in aluminum, a relatively soft metal, cutting this circle at 5 IPM (inches per minute) resulted in a rougher circle than feeding at 2.5 IPM. The top side was initially cut at 10. By the sound of the motors, nothing is getting bogged down at 10. Yes, it sounds like it's under heavier load at 10 than 5, but I didn't think it was enough load to make a difference. I think one explanation for the irregularity of the circle could be backlash, and it might be good to re-check and re-verify all the settings I have previously entered into Mach3, my machine controller. Another explanation is that the Sherline is a light duty machine, much lighter duty than the one I'm building will be. It might be that during some cuts, there was enough flex in the machine that the cutter wasn't where it should have been. In that "Left side" picture in the middle up above, there's a big, obvious, circle that's obviously deeper and possibly even wider than the track it was cutting. This is where I stopped the path, seeing it was very wrong, and the different depth there is from the cutter simply sitting there and spinning before I commanded the motor to go up and get the cutter out of there. That sounds to me like perhaps the Z-axis is settling down, letting it's compression out like taking the load off a spring, while everything else sat there.
The biggest lesson learned of all is that I probably should have put this in the four-jaw chuck and cut the center hole and counterbores on the lathe. I didn't do that because going between machines means losing the alignment, and I wasn't sure how to index from the middle of the hole. Right now it doesn't seem like a big deal. I'd just put it on the mill and set it up the way I did this time.
Hope this is useful to someone!
Have you ever considered taking up ham radio? Then all you have to do is sit around and drink beer waiting for the concrete to cure. :-)
ReplyDeleteKind of like fishing but more relaxed.
Ham... radio? I understand ham sandwiches and ham omelets but radios? ;-)
DeleteClayton Cramer has been struggling with a similar mill. They seem more like a toy, and not worth putting time into making them function usefully.
ReplyDeletehttp://claytonecramer.blogspot.com/
You could use an edge finder to determine your x-y coordinates, and calculate center from that.
I took a look at Clayton Cramer's blog but didn't see anything about a mill. Since the one I'm using for this is somewhat of a "FrankenMill" I put together myself, I wanted to see what he had.
DeleteI have 3 edge finders, a laser and two mechanical ones. Getting a piece of work exactly in the same place between two machines is kind of a deep topic, though. It depends on how "exactly the same" they need to be.
I'm betting it's spindle/cutter flex.
ReplyDeleteI did a TON of small/precision parts machining (manually) when I worked for Hughes, and I always had problems like you're seeing.
The only way I could consistently avoid it on the small milling machine we had was to take very light cuts, far lighter than the sound of the motor indicated.
Even though you might not consider what you're doing to be "hogging" cuts, there small machines are fairly flexible.
At least you're working towards an understanding of what happened and what's happening, and that will make you a better machinist.
The Sherline's strong point is that they're very accurate right out of the box. I know I've told the story of seeing a demonstration where a real master modeler takes a piece of steel bar about an inch long and half inch diameter, chucks it in a lathe just taken out of the box brand new, turns the steel down in a few cuts to .010" diameter, then drills a .005 hole down the center of the .010 piece. They're fantastic for micro modeling where precision is the priority.
DeleteThey're also the heart of several commercial systems used for carving wax for jewelry casting (or, rather, they were several years ago when I looked into it). That stuff requires high precision, too.
My machine is heavily modified, and about the only Sherline parts are the Z axis, motor and headstock. And the Z-axis is a non-standard part: it's longer than the standard Sherline Z axis. The X and Y axes are A2Z CNC; both axes are longer than the bigger Sherline mill, and both axes use 4TPI Kerk lead screws instead of the 20 TPI lead screws that Sherline sells in their mills. Although this whole project has been done very close to the dead center of the table, longer axes are going to flex more. To get the longer Y axis, I use a spacer under the headstock and motor. I'm sure that's more flexible.
After posting this, I measured the backlash on all 3 axes and reconfigured the compensation. They had all drifted in the years since I first set them up. Backlash is a bit more of a concern in a CNC system than when using it manually. I should try to measure flex.
Once the G0704 is running, this is likely to just get used for small pieces and high precision. In other words, it'll be used at what it's best at.
In such a situation (flexing axes due to lightweight equipment) it always made sense to me to set up the final cut so that it only takes off a few thousandths. You can be in error for the hogging out, but you want the least load on your system for the final pass.
ReplyDeleteAbsolutely right. That has always been the way I work. On this one, I'll claim stupidity on going too fast.
DeleteThe big advantage of CNC, aside from being able to do things that are darn near impossible without it, is that it doesn't get bored and forget where it is. It doesn't space out and not read the DROs. If you want to do a thousand passes taking off one mil at a time, it'll do it for you.
With some of my equipment, a mil at a time in steel is about all it'll do anyway!
DeleteObviously, I'm not "in production" ;-)
The more you push the envelope, the better your skills get.
ReplyDeleteWhen I worked at the shipyard, my partner and I got good a alignments. As a consequence we did a lot of alignments and got really good at it.
Keep the information flowing, this is interesting.
The more you push the envelope, the better your skills get.
ReplyDeleteWhen I worked at the shipyard, my partner and I got good a alignments. As a consequence we did a lot of alignments and got really good at it.
Keep the information flowing, this is interesting.
This design is a little "stiffer" than those small Chinese machines.
ReplyDeletehttp://opensourcemachinetools.org/multimachine/
So this part is essentially a spacer, between two thrust bearings, to control end play on the lead screw? So radial loads are minimal, as long as the shaft is centered and parallel with the ways?
ReplyDeleteSounds like some radial float room on the bearings would not be a problem at all.
That's what I'm thinking. With the bearing washers being held in place by the leadscrew, those bearings aren't going to move sideways into the area that I boogered up.
DeleteUltimately, I'm going to test it and look for problems.