There once was a company called LumenLab. They had designed a CNC mill called the micRo. It was really pretty in pictures. Just because I think it’s a beautiful robot, and as I’m waiting on parts for EasyMaker, I’d like to try and revive the micRo as a real robot people can make via self-sourcing.
I’ve broken the problem down into 2 stages:
1. Bootstrapping — building a robot that can make the monolithic blocks of the micRo.
2. Reverse engineering and building a micRo ( I’m targetting the M3 ).
Why the micRo?
It’s pretty and it’s a challenge. The micRo drove LumenLab out of business because it was really hard to make — even with professional CNC mills, lathes, and custom-made robots. From what I’ve read on the Internet, reject rate on parts was very high since the micRo required tolerances of .005″ on all parts with almost no square error over a span of 2.5″. From a machining perspective, this is really difficult to achieve. Most CNC designs allow for a lot more error, and use pre-manufactured parts like sleeve bearings, O-Rings, and retaining rings to compensate. Monolithic block is very difficult and expensive to achieve, and cannot really be made in quantity. Monolithic block design is very pretty — it creates a minimal, sleek look. My desire to reverse engineer the micRo is purely aesthetic. I would have bought a micRo, but LumenLab had a bad reputation for not shipping products, and I didn’t want to end up paying the money and holding the bag.
What’s already done?
Some reverse engineering. I’ve figured out some of the BOM, but not all of it. I’ve also created an OpenSCAD version of the X axis motor holder( I call it the Y axis, since I like to think cartesian. Technically, the longest axis is X. ). I can’t draw for beans, so OpenSCAD is nice. I can then compile to STL, and from there, to whatever format people need. It takes about 8 hours per part to create a reasonable drawing for any part. I don’t know where to keep the OpenSCAD file — I need a place people can get to it, isn’t public domain, and can be modified and changed by me or others. A source control system — but a free one. Not Git — I want Perforce, because I’m used to it.
What’s the plan?
Pretty complex, actually. There are 2 options I can think of.
- Reverse engineer all the parts and see if I can get the parts made somehow. The raw material for the monolithic blocks is cheap — no more than $15.00 /ft, and I think only 2 ft are required.
- I’ve been examining this plan, and I don’t think it’ll work. I reverse engineered a single block and sent it our for quotes. The machine shop quotes I’m getting are around $400/block. The machine has 9 blocks. This would make it cost around $4K or so just in machine time. I’d have to find people with professional CNC machines willing to machine for me. Most shops are charging about $70-100 an hour, with a fixed fee per setup. The blocks have machined almost every face! There’s a lot of setups.
- Which makes using a commercial service impractical, even with a group buy. There’s too many setups per part to get good pricing with the design. That’s probably why Lumen went under. They simply couldn’t make the blocks cheap enough to be profitable.
- Getting the plate machined, however, is practical. That can be done with one setup — so a group buy of this item will work. It can even be water-jet cut, which would probably be the cheapest way to get it machined in a group buy.
- Build a boot-strapping bot with the tolerances needed to machine the blocks. This is hard, but 2 technologies now exist to make this possible:
- Laser cutters.
- 3D printers.
- This plan has a huge downside — design time. It takes about 8 hours/part to design. I’m already spending 80 hours or so recreating the blocks for the micRo. It would likely take another 80-100 hours to design a machine that uses all the same parts as a micRo m3, but can be made quickly and cheaply.
For the second options, the idea is to buy the parts from the BOM, and use them as much as possible. Those parts can then be transferred to the completed machine. Though the option of building a bot to build the micRo will require some additional parts. The whole thing must be kept cheap. I’m thinking of under $500 for both bots.
I’m thinking of using misumi 2040 extrusions( they can cut them to precision length — but even their non-precision cuts are precise ) and laser-cut acrylic/delrin to jig parts for the “Y” axis ( the one bolted to the plate. ) I can then 3d-print a basic block and use Igus 3/4″ plane bearings from Grainger as a liner. This would bolt onto Misumi Rail segments, with a few 1/4″ thick acrylic laser-cut re-enforcing plates for the X axis blocks ( the ones that make up the gantry. )
The misumi rail segments would be very short — maybe 20-40mm long! They’re not there to serve as a base, but rather, to be the vertical supports in the blocks. The alignment of the blocks is done by laser-cut acrylic. 3D printed parts are accurate to .3mm. With some work, this should be able to make a machine that can make the micRo blocks. However, I’d prefer not to go down this path. 80-100 hours to design this machine, a few hundred dollars in raw materials, and many hours of assembly/calibration time. It’s the only path I can see that can be achieved at the moment, though.
What I need help with
- Generating complete 3D drawings of all the parts.
- This includes finding a place to keep them as we do them.
- 3D drawings allow us to make the parts with many different technologies — 3D print, machine in a shop, generate GCode and CNC mill ourselves, etc…
- Manufacturing the blocks, or getting them cheap enough.
- manufacturing the plate
- figuring out all parts in the BOM.
- Figuring out the drive screws.
- I know these are 3/8-12 ACME screws.
- Getting screws is cheap. Machining them without a large lathe is hard. Stepped screws are expensive.
- Getting taps for the blocks is expensive.
- How do we kill backlash?
The priority should be the Gantry. I can think of many ways to handle the Y axis( attached to the base plate ). But the Gantry — aka the X axis in my words — is difficult because it doesn’t use any existing bearing system.
I hope I can find helpers on this project — it’s far too much for me to take on.