Desktop CNC 'Thing'

[Misterg]

Thanks for the kind comments :)

As I hadn't really got the means to machine the mounting faces of the gantry feet properly flat, I intended to use epoxy as a shimming material and to adjust the gantry alignment before the epoxy set.

I fitted studs to the holes in the base and to try and stop the epoxy squeezing up the bolt holes in the gantry, I put an O ring around each stud.




The studs themselves and metal faces were also given a coating of furniture wax to act as a release agent (in case it needs to come apart). After a dry run with bits of packing, etc. to make sure it was possible to get it all set square, I masked off the area around the joints and gave the mounting faces a liberal buttering of epoxy:




There followed a frantic and stressfull couple of hours trying to get the gantry level to the table (side to side), and square to the Y rail travel and standing vertical (nod), all at the same time. I'd bought myself a 6" square to act as a reference, but I only had the one, so had to check each direction separately (with the square lightly clamped in place), make an adjustment, then go back and find out if it had affected any of the others (usually it had!). Do you ever wish you hadn't started something at 10:00 at night?

I'd put jacking screws in the flanges on the gantry feet to help with this process, and while they worked fine for levelling the gantry and correcting for nod, I'd completely overlooked the need to adjust the angle of the gantry to square the X and Y axes. The O rings I'd fitted on the studs were quite effective in keeping the studs centred in their holes - not where I wanted them to be! After a lot of sweat, frustration (and almost tears), I got the gantry and table reasonably squared (I ended up hammering a screwdriver down one of the stud holes to stop the blimmin' thing creeping back on me ).


I managed to get the 'nod' to ~0.01mm over ~100mm




And the axes square to a little bit better than that




And got the gantry levelled from side to side to within ~0.03mm over ~200mm as far as I could tell with the slightly warped table.





There was a healthy amount of squeeze-out from the joints that also had to be cleaned up before I could get to bed.




After a couple of days for the epoxy to go off, I fully tightened the nuts on the mounting studs and re-checked the perpendiculatity, etc.




With it together, I hastily cobbled together the basics to drive the steppers with GRBL on an Arduino. It's not pretty:




It would draw pictures though :)




After driving various bits off the end of their linear rails a few times (and having to play hunt the ball bearings) I decided to fit the homing/limit switches (I found some nice little normally closed inductive proximity switches that are only 6mm dia.)

The X axis homing/limit switch fits into the X/Z adapter and is triggered by 'flags' on the top of the gantry: (I was going to fit limit flags at each end of the travel, but ended up with just one that is used for homing.)




The Z axis switch fits into the hole on the other side of the X/Z adapter plate, but facing in the opposite direction to the X one. It's triggered by 'flags' on the side of the Z axis. (The Y axis one is under the table).


I also fitted a cable chain, etc. and did the final wiring on the machine side of things. There's a couple of tubes there intended for an oil mist style lubricator and a solenoid for the same on the side of the machine.



I'd originally (rather reluctantly) decided to make the working table out of MDF, but then a decent sized offcut of 15mm tooling plate popped up on ebay. After cutting it to size, I clamped it to the Y carriage and used the machine itself to drill holes through the work table and the carriage below it. The first time it cut anything in anger, and my first CAM program (the mouth was very dry before I pressed the button...)





Unfortunately, in the process, one of the clamps decapitated the Y axis homing switch!

[Misterg]

When I reassembled everything, I noticed that the X axis ballscrew seemed to be bent - the ballscrew nut mount was flexing noticeably as the screw turned.


I took it out again and got it centred up in the lathe and while it had some runout, I didn't think it was that bad (something like 1mm runout over ~400mm). Nevertheless, it seemed to be causing a problem, so I decided to try and straighten it.

I set the lathe leadscrew to the same pitch as the ballscrew so I could run the DTI down the ballscrew thread and plotted the total runout along the length.







Sure enough it had a couple of kinks in it (a small one at about 40mm and a bigger one at 100 mm - both in the same plane):






I managed to get them out by leaning on the end of the ballscrew with it held in the lathe chuck next to the bent bit, and got it running reasonably true. When I put everything back together, the ***** thing was still running out.


To cut a long story short, I eventually traced the problem to the way that the thrust bearing was mounted on the ballscrew.

The bearing arrangement is a standard FK type flanged block with a couple of angular contact bearings and a short spacer either side. The end of the ballscrew is machined down (by the manufacturer) to take this assembly and it all gets clamped into place by a locking nut. What was happening was that the spacer was a bit of a sloppy fit, and rather than bearing evenly on the (small) machined shoulder, it was bearing on a single point at the end of the ballscrew 'thread' - Tightening the nut was bending the ballscrew.

Once I'd figured this out, it was easy to make a new spacer that was a tight fit on the end of the ballscrew (loctited in place for good measure) so that the load would be transferred evenly to the ballscrew.





Runout with the original spacer (it ran true when the clamping nut was slack).




Runout with the new spacer:




With everything squared up, and the spindle trammed, I was still measuring a bow in the plate used for the Y axis carriage: +/- 0.07mm in the worst points.




The X axis had been set up straight to the glass plate and I went around in circles for a while trying to decide whether it was the glass plate or the Y carriage that was out, and whether the Y guide rail mounting was pulling the carriage out of true, but removing the Y axis carriage completely and checking it with a straight edge confirmed that it was much less straight than my glass plate.
I didn't want to pull my nice new table out of true, so I would have to trust the machine to skim the Y carriage flat (I know NOTHING about CAM or CNC programming, so feeds & speeds were, at best, educated guesses):

Some video snippets of the first, tentative cuts...

Taking the first 0.1mm off the Y carriage plate:




Another 0.1mm fully cleaned it up:




Flushed with success, I got it to drill and chamfer the matrix of fixturing holes as well:







I had to tap them all by hand though ...

[Misterg]

It's about now that I regretted my choice of spindle. When experimenting, trying to find cutting depths and feeds that worked, I became suspicious that something, somewhere wasn't right.

I went through the machine, looking for backlash, play or loose fasteners, but I couldn't find anything amiss - everything seemed tight. I then started clamping an indicator to various places to see how much they deflected and eventually found that firm hand pressure against the ER11 collet chuck would move it by ~0.1mm the spindle itself was only moving by ~0.01mm.

This led on to the discovery that the spindle bearings were a) weedy and b) wedged into the end cap with a rubber boot!



At this point, I should have cut my losses and found a different spindle. But I had a couple of spare angular contact bearings, and there seemed to be a lot of fresh air inside the front of the spindle housing...

I drew up the original design in CAD:



And indeed, as it seemed, there was enough space to fit in some more substantial bearings.

The only way I could think of to locate them axially was by sandwiching the original circlip (or something in its groove) between the pair of angular contact bearings that I had:



I had a suitable chunk of aluminium already, so the only thing I would be wasting was my time...

I needed to get the ER11 chuck off the old spindle to get to the bearing. My word, it was on tight! I had to heat it and jack it off the shaft with a bolt through a nut held inside the ER11 collet nut. It was that tight that I stripped the threads of a normal M6 bolt and had to use a HT cap screw one instead.

The new bearing housing and preload nut were turned up on the lathe:






The new bits and the old bits side by side



New bits assembled to the armature



All I had to do now was get that ***** ER11 chuck back on. I thought I might get away with being civilised - I heated the chuck and tried to press it onto the armature using my lathe tailstock. It didn't want to know at all. I eventually resorted to brute force, and drove it back on with a hammer.



In the process of replacing the chuck, I managed to smash up the brushes (I found out when I tried to power it up. There was a bang and a flash and all the lights went out...) Spares brushes are available from China with 3-4 weeks shipping. As I'm impatient, I ordered something close and sanded them to fit. While I was in there, I made a brass sleeve to replace the rubber bush on the rear bearing.

Perhaps unsurprisingly, the collet chuck didn't run awfully true afterwards (maybe +/- 0.05mm). I had marked the orientation of the chuck wrt the shaft when I took them apart, but my marks were inside the motor when I put them back together, so I couldn't see them! (I don't know if it would have made much difference.)

Anyway, it wasn't coming apart again, so any improvement would need to be made with the spindle assembled. I started thinking about how to set it up in the lathe to be able to re-cut the ER collet taper and realised that if I could support the spindle housing, I could true up the collet taper with it running in its own bearings. My fixed steady was *just* big enough to go around the new preload nut which held the housing steady so that I could re-cut the collet taper.



The best I could get was +/- 0.01mm runout when measured on the shank of a tool clamped in the collet. (I don't have any 'before' measurements to compare it to.)

I had measured the stiffness (compliance??) of the spindle before I took it to bits, and while there was an improvement afterwards, it was not as much as I had hoped - I was measuring ~800 N/mm before and ~2,500 N/mm after (measured at the top of the collet chuck - it would be lower at the tool tip). The stiffness of the spindle body measured next to the chuck is ~6000 N/mm (with the Z axis raised).

It is also very temperature sensitive - the bearings tighten up when cold, so I need to run the spindle for a few minutes to warm it up before trying to cut anything. (I probably shouldn't have used aluminium for the bearing housing)

I was hoping to be able to turn a sow's ear into a silk purse, but I think all I've achieved is to make a slightly better sow's ear.

I recommend against following in my footsteps, but I'm running with it for the time being.