Top surface profile.

I used a 1/4″ ball end mill with a .030″ step over and they came out pretty smooth. Just hit them with a little sandpaper and they will be ready for finish. Sweet. The program took about 15 minutes to run – much less time than doing it by hand and much more accurate.

I also played around with engraving, but I still need to figure out how to map the image onto the surface to generate the tool path.

I also needed to run off a couple of inserts to hold .50 BMG rounds in an ammo can. These are made up of 1-1/2″ glued up MDF stock. They have a relief cut into the back and a bunch of 1-1/16″ deep, 13/16″ diameter holes in them.

Milling the relief pocket.

In the past I used the router table and the drill press to make these, but I thought the ShapeOko would possibly be better.

Well, the jury is out on this one still. Milling the relief in the back worked fine, but was a little slower than using the router table. I could probably get more aggressive with the depth per pass and make it a wash.

Boring the holes was slow and a little frustrating. I started with a 1/4″ two flute down-spiral carbide router bit and got about 30 holes done before I decided that was not going to do it. It was at least twice as slow as using the drill press.

Then I realized that I had a 1/2″ diameter two flute straight carbide bit with a 1/4″ shank. That made the holes take half as long since they could be done in one pass.

Boring the holes.

But… MDF is hard on bits. The bits are pretty dull now and you could tell the difference from the first hole to the last with the 1/2″ bit. I have ordered some higher quality bits to try and will see if I can sharpen the straight bit again.

But the MDF also caused sharpness issues with the Forstner bits I was using in the drill press, so this is not really a mill related issue.

The other annoying thing was having to pause to move the clamps out of the way. I had to re-position them due to the fact that the dust shield ran into them because the bits are not long enough.

But I have a plan in the back of my head on how to deal with that issue…

Running total costs.

All Shapeoko CNC Mill Posts.

Grip Fixture

I also learned a few things:

1. No matter how tight the dowels are in the holes, put some double sided tape under the blank. I had the first one lift a little near the end.

2. The straight router bit does a better job than the spiral down cut bit. At least in oak it did. Actually the finish was about the same in the end.

3. I can speed up the feeds and increase the depth of each pass. I was pretty conservative for the first go-round.

Here is a close up – they look pretty good.

Closeup

Now I need to figure out the software tool chain to profile the tops. Then I can play around with engraving and machine checkering.

Oh boy, fun, fun!

Running total costs.

All Shapeoko CNC Mill Posts.

This evening I fixed it for real. Repeatable and within the limits I set for myself.

Now I’m happy.

Here is what I did:

I cut off a 16″ long scrap of MDF and clamped it to the table. Then I used my hole array generator script to make a series of .234″ diameter holes on 3″ spacings along the long (X) axis. Why .234″? Well I had some dowels cut that were .234″ diameter – that’s the hole size in a 1911 grip. I figured since I was using an 1/8″ end mill I could make whatever size hole I wanted.

Anyway, I bored a row of holes and then measured the center-to-center distances and wrote them down.

A series of holes.

The first run was off, as I expected. The total should have been 15″, but it ended up at 14.957″ with a pretty consistent hole-to-hole spacing. So I calculated the correction factor and entered the new steps/mm (43.700) value into GRBL.

Then I ran another row of holes. This time the hole-to-hole spacing was 3.000 +/-.005 and the overall measurement was 14.998″. I’ll take that!

Just for grins, I ran it a third time and got the same results. Wha-hoo!

Next I turned the board 90 degrees and wrote another hole pattern. I could only do four holes on 2″ spacing but that should do it. The Y travel is only about 7.5″ total.

The Y axis steps/mm were set to 43.717, so it was already close to the X value. When I ran the holes, they were all within .001″ center-to-center with an overall measurement of 6.000. Wow.

I ran it again, and this time the overall was 6.009″. So I changed the steps/mm to 43.700 to match the X axis and ran it a third time.

The hole-to-hole spacing came out at +/-.001″ and the overall was 6.002″. I’ll take that.

So now I’m happy again. I should probably double check the Z axis, but I’m not sure the best way to measure that. My cutters aren’t all that long. I think it’s probably close enough anyway.

Running total costs.

All Shapeoko CNC Mill Posts.

Yesterday I thought I had made some improvement in the accuracy department with my ShapeOko.

This evening I ran a job that demonstrated that I was all wet.

As you may or may not know, I make 1911 pistol grips. Each grip has two holes in it, spaced 3.074″ inches apart.

I have a couple of jigs and such for drilling the holes and shaping the grips. I thought I had written up a blog post about my process, but apparently not. I’ll add that to the to-do list.

In any case, one of the things I want to do is use the ShapeOko to mill out the grips.

Tonight I started making a fixture to hold the blanks. The first step was to bore two holes in the fixture that were 3.074″ apart. Simple enough. Or so I thought.

I used my Perl script to generate the G Code and ran it. It looked great. Then I grabbed one of my grip templates with the holes drilled in it and compared the holes. Oh hell.

I thought the template might be off a little bit, but this was really bad. So I measured the hole spacing.

3.114″ center to center. Off by .040″. Whisky-Tango-Foxtrot?

Apparently the correction factor I came up with was not quite right. Perhaps measuring the error over only .900″ is not enough. Maybe the cogs on the belts are not as evenly spaced as they should be.

In any case, I re-figured the correction factor using the 3″ dimension, entered it into the GRBL shield and did the holes again. They came out within .002″. Which is pretty damned close.

I did the same on the Y axis over a 2.25 inch span and got it to come within .006″.

So, I guess I need to clamp down a piece of MDF and bore some holes in it, measure them and figure out what’s really happening.

And here I had thought I was being so smart yesterday.

Well, I’m closer now than yesterday, so that’s something I guess.

Running total costs.

All Shapeoko CNC Mill Posts.

The blanks are about 3″ x 6″ and I’m boring 50 holes in them. When I set them up I put a pair of parallels under them in the vice and tighten the screw.

The problem I’m having is that the first holes go all the way through, like I want, but when I get down to the other end of the block the holes are a little short. The depth that I’m boring them to is .790″, so what gives?

This evening I did some investigating and discovered a few interesting things.

The first thing I did was take out my dial drop indicator to check whether when I told the Z axis to move 1.000″ it actually moved 1.000″. It turns out it did not.

I set up the indicator vertically and placed the point on the top of the router. My indicator has a travel of one inch (as most do) so I set it at about .050″ in. Then I told the Z axis to move .900″.

The indicator read .888″. Ah-ha. I repeated the measurement several times to be sure. I was pleased to see that there is very little backlash in the Z axis. When I sent it back to zero, it read zero. When I told it to move .900″, it moved .888″ every time.

So now I have a correction factor for the driver. The setting that is mathematically computed using the steps per revolution and the screw pitch is 188.947 steps/mm. I multiplied that by my correction factor (.900/.888 = 1.0135) to come up with a new value of 191.500 steps/mm. I entered this into the GRBL shield and viola! When I told the Z axis to move .900″ it moved .900″. Huzzah!

Then I thought “better check the other axis while I’m at it.” So I did.

I took a picture of the X axis setup.

Measuring the travel.

When I told the X axis to move .900″ it went .892″. So I followed the same method to generate a correction factor and fixed that.

Here is a very exciting video of the X axis traveling .892″

Then I did the same with the Y axis. Which moved .890″ when told to move .900″.

There is also not very much backlash in the X and Y axis either, but a little more than the Z. Still, hardly enough to worry too much about. I’m not making rocket engines. Yet.

So that solved part of my problem.

Next I decided to check and see if the Y axis was parallel to the table surface. I was pretty sure it was, but it never hurts to check again. So I hooked up my dial test indicator and ran it across the table. It was within .005″, so I called it good enough.

Then I put a blank into the vice and ran the test indicator across that. Ah-ha again. The indicator showed that the blank was about .025″ lower on the far end! That, plus the mis-callibration of the Z axis, would cause the hole boring problems.

So I grabbed some .010″ brass shim stock and put .020″ of shim under the low side of the vice and Bob’s yer uncle! Now the blank is within .005″ from end to end. I’ll take that.

Running total costs.

All Shapeoko CNC Mill Posts.

So I decided to do the dist-upgrade. What could go wrong?

Well, I’m still finding out. I’m not even sure what version I was running. 11.something I think. It was upgraded to 12.04.2 LTS.

What broke:

mysql – it appears to have removed the server application and not installed a new one. I had to install mysql-server (and dependencies) to get it running again.

sqlgrey – something changed so that it only listened on the ipv6 localhost address. I had to edit the config to force it to 127.0.0.1.

postfix – another ipv6 issue, so I disabled ipv6 in /etc/sysctl.conf

mailman – lost it’s template files. I believe they moved, maybe I originally installed from source, I don’t remember. I had to hunt down the template files (which were on the disk in /usr/share/mailman) and change the link to them.

carbon/graphite/whisper – not sure what happened here. I had to reinstall whisper and carbon. Graphite web pages were dorked up because python-whisper appears to have been removed and not re-installed.

That’s all I’ve found so far. Some fun.

And ironically, before I upgraded my Ubuntu server, I patched and rebooted my Windows 2003 server. In the past this has been a risky action, but today – flawless. Go figger.

I finally got a dust shoe finished up last night and I couldn’t be happier. I managed to run several unattended jobs over the past two evenings!

By unattended I mean I didn’t have to stand over the machine with the shop-vac hose in my hand. I setup the job, turned on the router and the vacuum and pressed go. Then I could wander around the basement and do other things while it ran. Bliss!

You may recall I had started to make a dust shoe a little while ago. I wasn’t happy with the plate though. I had offset the hole and it made me lose almost an inch on the Y axis. That’s the axis that is already short, and I didn’t want to give up the travel.

So I made a new plate, and bought some real vacuum hose. The new plate comes out the front and I lose a little X travel, but I have a bunch to spare there.

So here are some pictures.

Parts is parts.

From upper left clockwise we have the detachable shield, which is made of acrylic with craft foam glued around the edge using contact cement.

Next is the lower router mount with the new plate on it. You can see the larger hole for the vacuum tube.

At the bottom is the upper router mount with a new piece of acrylic screwed to it. This centers the vacuum tube and helps support it.

Assembled shoe.

All mounted up.

Here is the whole assembly on the mill. The shield is held on with velcro so I can easily pull it off to change tools and set zero.

Hose wrangling.

I fought with the hose for a while before I came up with this. I mounted a large wooden dowel to the ceiling and made a carriage to slide along it. (I bored the holes in the carriage with the ShapeOko!) This seems to work pretty well. There is enough flex in the hose and rotation in the carriage to track the Y axis, and it slides pretty well to track the X axis.

Yes, the hose runs into the lamp at the end of the travel. I still need to move the lamp. But it all works pretty well and collects at least 90% of the dust now.

This is actually not a ShapeOko post for a change. Well, it’s related…

One of the things I sell in my Etsy store is loading blocks for reloading ammunition. These are finished planks of maple with various sized holes drilled in them.

I have a pretty good system for drilling the holes using spacers that I made out of MDF (medium density fiberboard.) Below is an example of the normal blocks I make.

Loading Block

It has 50 holes in it, on 5/8″ centers. I start at one end and do each row, stacking spacers as I go. It works out pretty well and takes about 15 minutes to drill.

One of the things I want to use the ShapeOko for is to make the holes by boring them with a router bit, but I’m not quite there yet.

Then I got a custom order for a square block with 100 holes in it. No problem, I’ll just do the same kind of thing with the spacers.

Square Block and Spacers

I had figured out where the start point for the first hole should be and was drilling the first row. When I got to the end of the row I noticed that there was more space between the edge of the hole and the edge of the block then there should be.

Kind of a lot more actually. Enough to make my inner perfectionist sit up and take notice. Enough that I can’t sell this block. Hrm. What happened?

The blocks are 6-7/8″ square. The holes start 5/8″ in from the edge and are spaced 5/8″ apart. They should end up centered. But they were not.

Why not?

It turns out that my 5/8″ spacer blocks, instead of being .625″ wide, are actually .608″ wide. That’s only .017″ off. About 1/64″. Doesn’t seem like that much, does it?

It’s not if you aren’t trying to center a row of 10 holes in something. On my normal loading blocks it’s no big deal, because they are not centered the long way.

But over 9 spacer blocks that .017″ adds up to .153″ – more than 1/8″. And that’s a fairly obvious difference from one side of the block to the other.

Luckily I had extra blanks made up, so I fudged the difference and centered the next three that I made, but it was annoying to have to scrap one. Well I didn’t really scrap it, I’ll use it myself but I can’t sell it.

And now I’ll have to make new spacer blocks. And I’ll have to measure all the rest of the various sized spacer blocks that I have.

It was only .017″.

And boring the holes with the ShapeOko will avoid this problem altogether. Maybe I’ll get that sorted out before I have to make more loading blocks.

Over the past few days I’ve done more stuff to my ShapeOko. Here is all the stuff that I can remember, and some pictures.

Warning, very long.

1. First off (no pictures) I grabbed some 2x4s, cut them into three 4 foot lengths, jointed one edge and then ran them through the planer so they were all the same width.

I screwed these to the bottom of the MDF base for my ShapeOko and now it’s really stiff. Not very much deflection, so that’s a good thing.

2. Since the base is pretty stiff now, I chased some more accuracy (no pictures). I took my indicator and base and checked the height of the rails at both ends. They were all within .005″. Imagine my surprise. I also realized that I did the 10″ travel check incorrectly. So I set the steps per revolution back to the default and setup and did it again. This time I interpreted the results correctly and it appears that when I tell it to go 10″ it goes 10″ +/- .001″. So no real change needed.

I then clamped my dial indicator to the Z axis (see below for details) and ran it across the table the short way. It was off about .018″ in 6.5″, so I adjusted the rails to fix that up. Of course this assumes that the table is flat, which I’m sure it’s not completely, but it’s good enough for now.

3. I thought I had a bad Z axis motor, so I ordered another. I swapped the motor and discovered that it behaved exactly the same as the one that I thought was bad! So I guess it’s just that the threaded rod and nut are not very well machined. Sigh. Oh well, now I have a spare stepper motor in case I manage to burn one out.

4. I bought the Acme Screw Z-Axis upgrade from Improbable Construct. He made a forum post a few days ago saying they would be $10 off and that he probably wouldn’t make anymore for a while. Since I was not happy with my Z axis anyway, I dropped the $65 and ordered it on Thursday. It arrived on Saturday and I mounted it right up! It works great and now I can set the rapid value to 1400 instead of 800 and it works.

Acme Screw Z Axis Upgrade

5. I made an air deflector/dust shoe mount for the router. This was a hassle and a half, though it shouldn’t have been. The part is a square with two holes in it. You wouldn’t think it would be that difficult to mill, but I had no end of troubles. I drew up the part and generated three programs in CamBam – one for each circle and one for the profile. Unfortunately when you send an M30 command to the GRBL Shield it appears to treat it as a soft reset. So it lost my zero and when I tried to send the second circle it took off across the table. Which is always my favorite thing. Time to hit the reset button on the GRBL Shield!

So I fired up vi and joined all three programs together. But apparently the GRBL Shield was in some weird condition when I sent it – and it plunged the tool into the material and took off at a 45 degree angle! Man I have got to get that E-Stop button wired up! It was getting late on Saturday night at this point, so I just turned it all off and went to bed.

Today I ran the program up in the air to make sure it was all fine. It was so I started milling a part. It machined the two holes and was about 1/3 of the way done with the profile when the X axis locked up solid. WTF? Since this is open-loop control, the GRBL Shield has no way to know that the X axis didn’t complete it’s move and it kept sending motor pulses. The Y axis then started it’s move and cut a groove right up the middle of the part before I could get to the reset button!

I tried sending some X axis commands and it was not moving, so I powered down to see what the deal was. One end of the X axis was moving fine, the other end was not. Locked up solid. How is this possible?

I turns out that when cutting Acrylic the swarf that gets tossed and blown all over is very hard and non-compressible.  A tiny piece of swarf got under one of the V wheels and kept it from turning. I find it highly ironic that this happened while I was cutting out a part that should solve this problem.

Apparently the third time is the charm. I fired up the mill again and this time I watched the V wheels like a hawk, tooth brush in hand to clean any swarf out of them. I managed to get a part this time! Huzzah!

I mounted the plate to the lower motor mount with some 1/2″ spacers. The gap is where the air from the fan in the router will get deflected. Here are some photos of the plate mounting.

Plate – bottom view.

Note, one screw is missing. I only had three of the proper sized screws around – I’ll have to pick up another tomorrow.

Side view.

Here you can also see the 1/4″ aluminum rod I pressed into a hole on the mount. This allows me to mount my dial indicator for squaring up material.

Mounted to the mill.

All mounted up with a piece of 1-1/4″ plumbing fixture for dust collection. I still need to get some brush material to mount around the bottom of the plate.

Dial Indicator Mount.

Since the spindle lock is spring loaded, and does not really lock the spindle tight I wanted to have a more rigid mount for my dial indicator to enable me to square up my vice jaws or blocks of material. This works great.

6. I also made some hold down clamps. I actually did most of the work on these using my table saw. I ripped a strip of 3/8″ plywood, cut it into lengths and mitered the ends. But I used the mill to cut the slots and just wrote up a quick program in vi to do it. Worked out slicker than snot too!

Hold Down Clamp.

7. I took some hose that I had and ran it up and over for the dust collection. I don’t have a picture, but I took some video. I’m going to have to get some different hose apparently, check out the awesome noise it makes!

So that’s not really going to work for me. I’ll have to get some real vacuum hose or something.

All in all though I’m pretty happy with the mill.

Running total costs.

All Shapeoko CNC Mill Posts.

I posted the results of the CDS test on the ShapeOko forum and another member suggested that the results indicated that my X and Y axis were not square.

I attempted to square them up by squaring the Y axis rails to the rear plate, and then making sure that the back edges of the X carriage plates were the same distance from the rear plate. Then I ran the CDS test again. Same results.

Today I investigated further.

The straightest edge that I have is an aluminum circular saw cutting guide. I’ve been meaning to pick up a real straightedge for some time, but have not done so yet. I did purchase a set of engineering squares the other day, so I used the blade on the 6″ square to check the straightness of the aluminum. It looked acceptable for the time being, so I used it.

I clamped the straight edge across the rails and used the 6″ square to square it up to the front rail. Then I discovered that a 1/4″ rod is a nice slip fit in the slots of the Makerslide, thus making it easy to mount my dial indicator!

Straight edge.

Taversing the indicator along the straightedge showed that the axis were unsquare by about .020″ over 5.5″ That was not acceptable.

I ended up unbolting the X rails from the end plate and using a file to widen the bolt holes. Then I squared it up. Now it’s within .002″ over 5.5″ and I’ll take that.

The next issue to address is the Y travel. It was mentioned on the forums that the belts stretch under tension and that the .08″ pitch might open up a little bit. So I decided to test it out.

I have three 1-2-3 blocks and a gauge block set, so I made up a 10″ test length. Well, almost 10″.

First I measured all my blocks. The 2 block precision set measured 2.9995″ each, the cheaper block measured 2.9990″, the 3/4″ gauge block was .7490″ and the 1/4″ gauge block was .2490″. So the total length was really 9.9960″. But that’s close enough to do some figuring.

I clamped them all together with a square on the end for the indicator to touch off on and bolted them to the table.

Gauge blocks.

I set the indicator to zero on the end, lifted up 1/2″ and sent it 10″ down the table. The indicator read .007″ short. I backed up 11″, dropped down 1/2″ and moved forward 1″ into the end. The indicator was off by .0015″. Backlash I suppose.

Repeating the test showed .007″ again. That number, plus the .004″ short that the gauge block stack is, told me that the X axis needed tweaking by .011″. Since I was moving 10″, that means I can take the steps/mm and multiply it by 1.0011 to get the new value. I set up, ran the test again and the indicator showed .004″ the other way. Success.

However, I really need to build a torsion box table. Just leaning on the table while the indicator was against the end of the 1-2-3 block would move it .002″.

I also don’t know if the X and Y axis rails are parallel to the surface of the table (probably not) nor whether the Z axis is perpendicular to the table in either direction. So I have some work to do still. But I’m getting closer now.

It’s close enough to use for the things I want to do with it at the moment, but I’ll still keep tweaking it.

Oh, and my Z axis motor is shot. When I mounted the coupling to it I had to shorten the shaft. I used a Dremel cut-off wheel and I believe that I overheated it and ruined the bearings. And maybe something else inside. The Z axis motor makes a terrible noise and gets very hot to the touch, even when it’s not used very much.

I ordered a new motor today.

Running total costs.

All Shapeoko CNC Mill Posts.