Disclaimer

This blog is for entertainment purposes only, and is not meant to teach you how to build anything. The author is not responsible for any accident, injury, or loss that occurs as a result of reading this blog. Read this blog at your own risk.

Sunday, January 26, 2014

CNC mill conversion - part 12

More Z axis components, more issues solved, and more tools.

While I wait on the nose gear hardware for the Long EZ, I thought I’d post an update of the status of the mill conversion.

After working on backlash reduction by replacing the balls in the ball-nuts, I ran a couple of quick tests to see how repeatable the X and Y commanded positions were, and shot a couple of videos of the process.



Y axis repeatability test 




X axis CNC repeatability test 


Next, I started working on the mill’s Z axis left bracket.

This bracket is different from the right one because my mill has a big flange that is in the way, and as a result the bracket has the shape of an upside down “L”.


Flange on the right side of the mill

Modified holes layout


I started by squaring and surfacing a 3” x 3” (8 cm) blank, then milling the 0.025” (0.063 mm) recess for the top plate. However, something weird happened while surfacing the plate, because the finish had a strange looking cyclical pattern to it. Interesting really, but not what I was looking for, and while acceptable on this part, it might not be on others.


Top plate recess, and improper surface finish.


I decided to keep it that way while I would investigate further into its origin. 

Meanwhile, I wrote a little G code to cycle trough 4 set of coordinates, and center-drilled, then drilled all the necessary holes to later mount the bracket to the mill column.


Drilling holes in the soon-to-be bracket


I used the band-saw to rough the outer contour, and an end mill to finish it.


Rough cutting the bracket

Finished brackets


On a different subject, you might not know that last year I wrote a couple of articles for a machining magazine. Although they haven’t been published yet, and might never be for all I know, the process led me to strike a few interesting internet friendships along the way. One fellow writer that has actually been published many times is Jerry Prior aka “Ozzie”. 

Ozzie was kind enough to let me use one of the modified screen he developed for Mach3 (mill controller software) for use with a probe.


Ozzie's Mach3 probing screen interface


I decided that a probing capability will eventually help me save a lot of setup time, so I bought one on eBay, and with Ozzie’s helpful screen I started experimenting with it. I am still far from operational in this respect, but I can almost taste the possibilities.

Here’s a quick and dirty example of something that only takes a few seconds with a probe, but that would probably take me 5 minutes otherwise (I am a bit of slow, I know).




insert probing video


I will probably discuss probing in more details in the future, after thoroughly exploring its capabilities and limitations.

Returning to the strange cutting pattern, I tried all kinds of mill adjustments with no luck, and started to get frustrated by my inability to master this important aspect of machining. The strangest thing was that the pattern was highly localized to the right side of the X axis.


Surface finish worsening toward the right

Another example of bad surface finish


The pattern actually measured a depth of 0.001” (0.0025 mm), so it wasn’t just a “beauty mark”, instead it was somehow an actual digging action by the cutting tool into the material.

I eventually left on another trip leaving this issue unresolved, but I couldn't stop thinking about it, and I eventually realized that the only thing that could create such a cyclical pattern had somehow to be the ball-screw. So, as soon as I got back home, I stripped the mill down once more for a side to side comparison.


Things that make you go... mmm!




Sights and sounds of a troublesome issue


I think what was probably happening was that the ball-screw was binding somehow, lifting the mill table up and then down 0.0005” cyclicly with every turn. Things would get worse toward the right end of the table, as the binding increased where the ball-screw is held.

After another frustrating day of taking things apart, adjusting, putting things back together, and retesting, the solution ended up being to loosen everything up at the same time, GIBs, ball-nut, and shaft bearings-to-table nuts, then tighten them all down a little at a time while running the X axis back and forth, making sure the binding wasn’t reintroduced.

The surface finish is actually better now than when the mill was manual only. 


Before adjustment (top), and after (bottom).


Here’s a surfacing video of the same fly-cutter that left the wavy pattern on the left bracket, now creating a near-mirror finish...




Not perfect, but nearly so.


All and all, another great learning opportunity, and only a few more white hairs on my head because of it.

The next component to receive attention was the Z axis stepper motor mount... 


Fabricating the stepper motor adjustable mounting plate

Mounting plate ready for service


... and the motor stand-offs...


Z motor stand-offs before drilling and tapping

Working on one Z stepper motor stand-off


Here they are all together...


Stepper motor mounted on its adjustable support

The small pulley will drive the ball-nut through a belt

Flush screws at the bottom were needed in order to fit/slide over the plate above which it is mounted.


The Z axis spacer was simply a rectangular piece with 2 through holes drilled in it.


Spacer blank

Finished spacer


The Z axis inner block was much more interesting, as there are tapped hole on the top, and the sides, and they all must end up matching those drilled in all previously made parts. To ensure this, I had to change a few hole locations that did not line up correctly in CAD using the original plans dimensions.


Inner block with modified drawings

Hand tapping the 5/16-18 holes

Starting a ¼-20 tap straight, before taking over manually.

Spacer and inner block finished

Checking for hole alignments


The ball-screw mounting plate was just too big to be held in my vise, so I had to resort to a different, and more time consuming way to grab on to it, and align it for machining.


This top plate was too big for my mill, and had to be machined in sections.

Straightening the long edge

Squaring the short side


Holes had to be drilled for interfacing with the motor mount, the Z axis spacer, the inner block, and the two side brackets.


Starting a ¼-20 tap on the top plate

Starting the hole for the 2.3" pocket


The most exciting thing to date was machining the recess hole for the 2.3” ball bearing. This could have been done a number of different ways, like using a boring bar, but I wanted to leverage the 2D CNC capabilities the mill now possess, to make this difficult cut for the third axis.




CNC cutting of the pocket (15 min)


The results of this first CNC project were remarkable (to me at least)...


CNC milled pocket

Inner hole CNC machined wider

The pocket's surface is actually much smoother than it appears

Pretty nice wall finish with no tool chatter


Putting it all together for the first time, it was amazing how well the bearing fit in the pocket, and a real relief that all of the holes lined up so well.


Ball bearing pressed in the pocked by hand

Bottom view of the assembly






Tuesday, January 21, 2014

CNC mill conversion - part 11

Backlash reduction and ball-screw inconsistency

In the last CNC conversion post, I measured the backlash to be 0.0042” in X, and 0.0049” in Y. While these values seem minuscule, they are fairly substantial in the CNC world. 


Backlash measurements using the original 0.1250" ball bearings


One pretty common way to reduce this problem is to repack the ball-nuts with oversize ball-bearings. The original were 0.1250”, so I replaced them with 0.1270





Removing the ball bearings from the ball-nut




100 oversize ball bearings for $12 (times 2)

Y ball-screw being taken apart






Ball-nut repacking video (courtesy of Hossmachine.info)



After spending a couple of days messing with both axes, I did get some improvements out of them.  Unfortunately, they didn’t end up as close to zero as I was hoping, but they did initially go down to 0.0013” and 0.0027”. 


X and Y reduced backlash, after repacking both ball-nuts with bigger ball bearings.


Thinking that I could squeeze a little more out of the X axis, I purchased some 0.1280” ball bearings. 


Going for the "diminishing returns"


I was only able to fit 30 of the 0.1280” ball bearings in the Y ball-nut, before the the ball-screw locked up and stopped turning, so I ended up removing them, and using only 13 of the 0.1280”, and 53 of the 0.1270” ball bearings.


Packing the ball-nut

Ball-nut is filled up


Because there are 11 balls in the return tube at any one time, and I did not want to create a situation where I would have the majority of the big balls disappear in there, and have the ball-nut get loose cyclicly, I spaced the bigger balls at equal increments (1 big for every 5 small ones).


Packing the return tube

Return tube back in position (strap clamp missing)


I really should have bought an intermediate size, but there were none to be had that I could find between the 0.1270” and the 0.1280”.


No intermediate offerings available between the last two sizes


With this latest setup the ball-nut is pretty tight, much tighter than before, but it is still able to rotate consistently, although not freely anymore. Ideally, I’d like to reduce the amount of 0.1280” balls to maybe 7 or 8, in order to reduce the stress on the stepper motor, but I’m not about to take it apart again, at least not yet. The Y backlash is now down to 0.0020”.


Final backlash values


These are values that Mach3's backlash compensation should be able to handle easily, so I’ll call these two axes done.

My new issue is now with the consistency of the Roton ball-screw thread. In other words, the channel in which the ball bearings travel is not wound at exactly the same rate of 5 revolutions per inch throughout the length of the screw. 

This creates a new problem where the CNC controller might command a travel of... say 2.5000”, and while the screw will turn the appropriate amounts of turns, because of the inconsistently wound screw the tool might end up cutting at 2.5025”, for example.


Zero deviation represents the perfect ball-screw


This problem is entirely fixable by purchasing a more expensive screw (as always).

There is however a feature in Mach3 called “ball-screw mapping”, where one can record the actual travel achieved by the commanded movement, and train the CNC controller to undershoot/overshoot by the difference recorded.

I haven’t had much luck with it so far, but I’ll give it another chance to perform once the mill conversion is finished.