I took the Mantis to the TriLug meeting at Splatspace last night (Thursday, Dec 8 ) and it generated a lot of interest, both in the mill and the PCBs it creates.
I had only a tiny scrap of FR1 left, so I thought I could try milling a fake board on a piece of acrylic as a demo. LOL- with no spindle speed control, it immediately began to melt it and the bit grew a giant blob/bead of melted and cooled acrylic. It might be possible with a higher feed speed and something like a dremel steel bit.
I could control the speed of the motor but unlike a true spindle it will have less torque as it slows down.
I ended up toolpathing a tiny picture of a rose for a young fan, on that last little scrap of PCB board. That was fun. And finally I stuck in a cheap diamond-abrasive teardrop bit and tried a piece of glass.
It worked surprisingly well, which I honestly was not expecting. I had to lie about the diameter of the bit to get cad.py to fit paths into the lines of the image, but it worked.
My single-point diamond bit arrived today- I have to see what it can do. I also have more of those diamond abrasive bits, but they are really of poor quality and may not stand up to much use at all.
(Most of these experiments are for the Holiday Ornament Make-A-Thon next weekend.)
Also, today I finally got the chance to actually mate the Atmega 168s with the Fabio 1.1 boards. I haven’t soldered them yet; this was just the first time I actually aligned the board with the chip under the scope- confirmation that my math and the milling paths were all correct. It fits great. I hope to get the two I have cut, soldered up this weekend.
Mantis– the next thing to try is a bit of graphic etching, probably on acrylic with a conical Dremel bit. Still haven’t scored any new FR1/FR2- I should get that this weekend so I can start cutting some other boards. No progress on designing new boards in Eagle because of Real Life and Related Concerns.
Fabio 1.1– I got 2 boards cut and my Mouser order arrived with all necessary components for 5 of them. It works out to about $7 per board. The only definite thing I would like to try after I play with them for a while is to build a version of the Garduino garden controller, concentrating on irrigation only.
Cupcakes– My 2nd Makerbot cupcake developed Z-axis problems. Arrgh- the pain returns. I have parts on order and once my “good” bot returns from its stint at the Mobile FabLab, I’ll continue printing the plastic parts for an upgrade: a Lowrider XY axis, and a Z-rider Z axis (both excellent designs from Thingiverse.) I also need a better extruder but I am still researching the options there.
The Mantis is going to the TriLug meeting at Splatspace tomorrow night for the Open Source extravaganza, and both of my printers and the Mantis will be working at Splatspace’s Holiday Ornament Make-A-Thon on December 17.
After all this quiets down I will resume work on my Mendel, which I stopped when I started concentrating on the Mantis.
Tonight I ran the same fabio 1.1 board, only generating the Gcode with David Carr’s cad-opt.py, which was definitely more efficient, just as advertised. It was visibly spending more work time in each area of the board, radically reducing the amount of unnecessary travel time. It reduced the total time required by about 50%.
The board was identical except that the “i” in “Fabio” was missing. I do not know if this is a glitch in the Gcode, a minor milling error in the bot itself, or the chip on the corner of the end mill getting worse and causing shenanigans. All of the actual traces look fine. I’ll re-trace the .png and see what the paths look like.
Also: I reduced the Z depth from 0.01″ to 0.005″, with no dropouts or un-milled areas. Awesome!
I continue to be impressed with the accuracy of this device, and also with the fact that it hasn’t actually blown up yet. After building 2 Makerbots I am so used to the Shakedown Period Of Pain in which every single assembly breaks in turn that I am continually apprehensive. But so far…
a group of 3 of us at the Splat have been working on building David Drake’s PirateBox, an anonymous, non-internet connected wireless file sharing and chat space.
We’ve been slow, mostly getting together once a week at the open meetings. Yesterday I discovered that the model of Pogoplug I bought (used) to serve as the file server does NOT work with the new PirateBox software. Dammit.
I am annoyed because I did not do my due diligence of reading all of the documentation and forums before acquiring parts. Best practices, y’all. OTOH, this fact was concealed in a thread on the forum and not documented elsewhere. Oh well.
We will prevail and fully document what we do for the benefit of others. A Seagate Dockstar is on its way to us and in the meantime, we can make a slightly less-featured PB with the Pogoplug.
More lessons learned that I forgot to post-
1. I am now using a laptop brick I found at Splat that puts out 12v at 4A. That is sufficient to handle the spindle and all 3 axes, no problem.
2. The spindle motor shaft has a plastic sleeve on it to bring it up to 1/4″. That sleeve wants to move around a bit. A bit of Goop took care of it, but not enough to keep it from wanting to creep axially. Specifically, what was happening was that it would want to move upwards, taking the vinyl tubing that connects it to the 1/4″ spindle shaft with it. Then the tubing would slip off of the shaft and I’d lose spindle power.
3. The solution was to use a longer piece of the tubing that allows no room for it to slip. I still think the tubing is a good idea- takes care of any minor misalignment of the motor vs. the spindle shaft and allows for tweaking of it. The torque is so tiny that there’s no need for anything tougher.
4. For the little board that I did (and for others I foresee doing) I didn’t have to level the bed at all. Larger boards may require leveling the bed by milling it flat. I’ll just have to see. (The “real” way to determine that would be with a dial indicator attached to the Z axis to measure the flatness as I ran the X and Y axes from end to end.)
5. I put the electronics on the rear, above the Y motor. They are subject to getting wacked back there, and the Y motor doesn’t protect them as much as I thought it might. So, I need a little shield back there.
6. FR1/FR2 is not so easy to find these days. MIT apparently buys theirs in bulk- we (Splatspace and FabLabs Carolinas) need to find a decent source for ourselves.
7. I need some Eagle-fu. I think I will take some old Mims circuits and use them for learning to design in Eagle and export for milling. Some of them will be through-hole, which will involve some drilling as well.
On Tuesday at Splatspace I made an attempt at a “real” board milling. Once more I had Z axis issues, AND the tubing that is used to couple the motor to the spindle shaft came loose. BAH.
After a quick Thanksgiving trip I made some adjustments: loctited the Z axis coupler to the stepper shaft, and installed a new piece of tubing on the spindle shaft. Fired it up and… success! Brief video:
And a pic of the final product:
1. One flute on this 4-flute 1/64″ endmill is chipped, so things might get a little smoother. I’m shopping around for carbide mill prices.
2. No missing traces, no bridges! Woot!
3. Z axis paranoia led me to make it too deep: .01″ instead of .005. I will try .005 next time.
4. Lord, this thing is sloooow. I don’t want to bump the feed up any higher (although with a shallower cut that might be okay). It took an hour to do this board, mostly due to inefficiently hopping around from point to point. Carr supposedly has an optimized cad-opt.py; this may reduce the time. I will try that next.
5. I didn’t cut it out using the mill.
6. I really need a USB microscope; that pic is kind of lame.
I spent most of my “make” time this weekend printing various upgrade parts for the Makerbots. I printed one Wade’s extruder, finessed some temp and bridging settings, and then printed it again much better. Also started in on the 20 parts to print for a Lowrider XY assembly.
When I got a chance to try the Mantis again I added in the G20 code to specify inches and etched most of a Fabio 1.1 microcontroller board- a dry run using thin plywood as a stand-in for FR1 circuit board.
I took video but Youtube was throwing up on its shoes last night. I will try again to upload it this PM.
Good- The accuracy is great! I can clearly see the “Fabio” name etched in tiny letters, and the separation on the Atmega leads- the finest part- is good.
Bad- the Y axis leadscrew came off of the stepper shaft partway through- I will reattach and use Loctite. I got enough done to see that overall the accuracy is very good.
To fix: I really need a 12v supply so that I can run the spindle off of the same supply as the shield. The 19v brick I am using is great for the steppers but too high for the little spindle motor. I can probably score an appropriate laptop brick at the Splat, or at least an ATX. That would be more bulky and ugly, though.
Cad.py may be what I need to use for gcode when my start point is a png, but it does have problems. The main thing I saw is that it does not optimize the paths in any sensible way, so it mills one little bit, then travels all the way to the other side of the board and does one little bit there, then travels back, etc.
It would be nice if “travel” speed was much higher than “feed,” but I don’t know if cad.py can specify that. I don’t see it on the controls.
Last night at the Splat mtg I had some real progress.
I installed the spindle- it pulled less than 1A at 12v, which is what David Carr recommends on the Makeyourbot site. It needs to be glued in place better, which I can do tonight. I am a little concerned about running that toy motor off of 19v; I may have to look into some way of dropping that.
I used MIT’s cad.py to generate a gcode file from the grblshield png image. I’ve used cad.py for other things in the Mobile, but hadn’t used it for gcodes before. I used a scrap piece of wood as a stand-in for copper board.
About half the times I streamed the code, it seemed to work correctly, although the feed rate was waaay off (too high) which was causing it to drop steps. The other times, it worked in one corner and made a verrry tiny rectangle.
I finally figured out that what I was seeing with the tiny rectangles was the entire board, shrunk by a factor of 25.4. In other words, GRBL was randomly selecting either mm or inches as its units.
In looking over the code again I realized that cad.py is not adding any G70 (inches) or G71 (mm) code to the file. (EDIT: it’s probably actually G20 and G21.) And with no units specified, the feed rate is also wildly variable. Duh. (Cad.py also specifies a spindle speed, which is irrelevant for the Mantis, but that doesn’t seem to cause any trouble; it just ignores it.)
I will manually add the G70 code and specify a reasonable feed rate in inches and then try it again.