New Machine Build- Starter Machine

[O'Seanski]

So, anyway, my wife was a nanny. Long story short, her employer flaked out with postpartum and decided not to go back to work after all - i.e., my wife is out of a job. The hard part about it is that this was one of the few nanny jobs that would allow her to bring our newborn along to work. We talked about it, and decided it would be best if she was a stay-at-home mom anyways.

So, now I am the sole provider for my family. As it stands, we are short about four-hundred dollars a month to break even, about a thousand to live comfortably. I tried to get another job, but nobody will hire me for the pay we need because I don't have the paper to prove I know stuff. (a degree) So, the only option is (gulp) self-employment, at least part-time.

I have two things going for me. First, I have excellent spacial acuity. Every time I have taken an I.Q. test, everything looks pretty mundane. In fact, my general knowledge is a bit lousy. my spacial acuity, however is in the 99th percentile. I have always been able to take things apart in my head to figure out how they work, or identify a 3D object unwrapped into a 2D object. I totally blew the lid off the red and white blocks thing!

Second, my wife's uncle is a sign painter. I asked him if he would be interested if I could add a third dimension to his work. He asked a bunch of questions and seemed genuinely interested. He's been at it for fifteen or twenty years, and between the two of us, it is almost certain to be very lucrative.

The software for CNC, at least the software that generates toolpaths reliable, is just too expensive right now. I have been a mediocre Perl programmer for about three years, processing text documents. (I do know my regular expressions, though.) The solution I found was to make height fields (or, rather, "depth fields") in Blender 3D, then use Perl to trace a virtual stylus across the image. The original script generated some rudimentary G-code, which I tested with CNC Simulator. I can do raster and plunge, and I'll write edge-detection when I get around to 2.5D-type projects. I'll use inpout32.dll and Win32::API to send data through the parallel port to my Arduino.

Somehow, I had it in my head that software would be the hard part. Hacking software and hacking hardware seem similar, in that there is a recursive design/create/test loop. The hardware takes MUCH longer, though, and, I've found, requires a LOT more patience.

I've been looking at other projects for about a year, and it all seems pretty straightforward. The big hurdle is that it takes money to make money, so I've had to do this as much on-the-cheap as possible. The money that would have bought me Christmas presents is what is getting this thing built.

The table itself only has three legs because the bozo who "leveled" the garage floor before I bought the house did a crappy job. My machine will use ordinary DC motors, which I found at All Electronics. I tracked down some data on the model number. They supposedly push 12,000 rpm at 6 volts, and, if I converted and calculated correctly, they have a torque of something like 500 oz in. Unfortunately, they might also have an enormous amount of inertia. They might over-shoot a lot, but I'm only looking for .01 accuracy for now. I saw them mentioned on an RC site, reported to be a good replacement for radio-controlled racecar motors. The motors will be controlled by the Arduino, with quadrature feedback. I made the encoder templates with the "donut chart" feature in OpenOffice Calc, and printed off onto mailing labels. They are then affixed to polyethylene from a DVD case I had laying around. I'll have to write the Arduino code on the machine in the garage that will control my CNC.

I looked and didn't know where to find bearings I could afford, but I did find a mess of chair casters, which I tore down and made into linear bearings. The edges of my rails are hardened with EMT electrical conduit, which was only about three bucks for a ten-foot length.

Overall, I've taken a finish-carpentry approach to building this thing. There are really only a few points that need to be precisely aligned, and the rest is just built up around them. It's probably going to be an over-designed, heavy monster, and I might have to take it slow to get good work out of it. But it's only meant to hold up long enough to pull back on the flight yoke, clear the treetops, and get the parts cut for some stellar Joe-esque outfit. The software is going to be pretty clumsy for awhile, too.

I had post this before, for all of ten minutes, before pulling it again, but Eurisko managed to reply in that time, and I was encouraged to repost it. Heck, maybe somebody can get some encouragement, if something this crude actually works. Truth is, it's scary putting your stuff out there. Some of you guys really rock, and your work reflects a very long period of hard work and talent. It's intimidating.

Our original D-day way Jan 9, but my employers were very generous with bonuses this year, so we got a reprieve. We might be able to last through March. I might have that long to jump-start this thing. It would be kick-ass if I end up pulling this off!

[BillTodd]

I wish you the best of luck


I don't think you're in a position (financially) to experiment with software, DC motors and ardinios. If you want to build a cheap device fine, but buy the right tools for the job, not the cheapest (it'll cost you dear in the long run) i.e use steppers or servo motors with proper tested drives and power supplies (e.g. geckos, or if you really want to build your own, check the open source forum for Linistepper, UHU etc. )

If you're aiming to sell the product made on your machine, rather than the software that you intend to write, you'll be simply wasting your time writing software. Buy CAM software that does what you want.

Run the machine with a cheap second hand PC running Ubuntu Linux and EMC2 (EMC takes g-code and outputs to motor drivers - it works brilliantly and costs nothing).

[YZF]

Good luck, here are some thoughts:
My guess is those DC motors you looked at won't work for this application, if you post some details I can double check your calculations. Even if the motors are capable you would still need drivers and a closed loop servo controller.

I would think again about stepper motors, you should be able to get some off ebay or strip from some used electronics. If buying an off-the-shelf driver is too expensive there are a lot of free designs on the web you can use to build your own.

You can use something like TurboCNC or emc to control the machine for free. I've used an evaluation version of MeshCAM for generating some gcode for testing my CNC. There are a few other free/evaluation options with various capabilities (not too great but still). Once you actually make a few signs and see that you can make money it would probably be worth your while investing in more serious software.

From a mechanical perspective one challenge will be making the machine stiff and accurate enough to do quality work. I would review some of the other designs that people have used to see what works. Your linear axis need to be rigid in all directions except your travel direction.

I hope I haven't discouraged you too much I think that if you keep working on it and get advice from the people here you should be able to get a machine up and running.

[O'Seanski]

Man! This hardware hacking stuff is freaking hard! Now I know why hardware is called "hardware" and software is called "software"! I have learned three things.

First, rigidity and weight are interdependent. In order to have something rigid, it has to occupy a significant volume in every pertinent direction. But, that adds weight, which can pull things out of alignment, and make them less rigid. One also needs to be mindful of where the weight is resting.

Second, the bearings on any machine really need to cover a wide spread in relation to the distance away from the forces involved. The distance between the weight of my structure and my bearings was more than a foot, but the contact points on the rails were only about eight inches apart, if that much. Like with a lever, there is a fulcrum, a weight, and a force. The bearings, top and bottom, are like two fulcrums, both sharing the force countering the weigh of the gantry. The force of the weight of my gantry was actually multiplied against my bearings, because of the ratio.

Third (and this is something more that I am reminded of, rather than learning), people do things a certain way for a reason. Only the cluelessly inexperienced are arrogant enough to look at something that somebody else built and say "Well that's a dumb way to do that! I can do that much better!", then set out in a completely different direction. However, we learn by doing, and I learned why that smarter fellow did it that way after all of HIS dumb exploits.

I got the moving gantry up, but it was just too heavy. The linear bearings I built weren't long enough, and got all bound up with the weight of the gantry all lopsided on them. I was so bummed about my failure, then so jazzed about the lightbulb over my head that I forgot to get a picture of the crappy gantry before I tore it back down. I then went down another dead-end before I figured out what you see here. I'm still using casters instead of bearings, but I scrounged up enough bearings that I can use them on the other axises instead of casters.

The original rockcliff configuration with the moving table and stationary gantry seems to place most of the weigh on the table and less on the rails and motors. It seems like a good starter machine. I rebuilt things and got something that moves very well in the right direction, yet is also very rigid in every other direction. When it's all finished, I'm going to take a lesson from the late great David Gingery, and route the top of my table flat and true, using the business end of the machine itself.

I absolutely would LOVE to be able to get a proper set-up built, you guys. Unfortunately, for me right now, the word "buy" is about as operative as a thirty-year-old Land Rover in an overgrown field in Hong Kong. I literally have eight bucks and whatever I can scrounge in my garage and house.

I really don't trust steppers, though. No accountability. It could drop steps, and you would never know. As long as my Arduino knows how far everything turns and when, I think everything can be kept under the right kind of control. This stuff is pretty dang complicated from the outset, but I've done the right kinds of projects over the past five years that I'm confident I can pull it together with the quadrature feedback approach.