Tuesday, April 21, 2009
Designing for Toolheads
I am new to RepRap and want to give a heart felt thanks to everyone who has contributed to where we are today. I recently built a bootstrapped Darwin by making all the printed parts out of MDF and poplar. I had my first successful print last week and am looking forward to making my machine a true RepRap in the near future. I also look forward to being a contributing member of the RepRap community. You can read more about my progress at http://blog.onshoulders.org.
Now that you know my background I would like to shed light on what I feel is a fundamental flaw in the Darwin mechanical design. Please understand that I think the Darwin design is top notch. Still, everything has room for improvement.
Before I built my extruder V2.0 I got stuck with an RRRF that wasn’t selling stepper motor driver boards and no Makerbot Industries. Thus, I attempted to mill the PCB with my working Cartesian robot and a Dremel. Let me throw up a picture and then describe what happened.
As you can see, I got close. However, there are areas where the bit dug through the board. After a weekend of trial and error I narrowed the problem down to a flaw in design. I really should render a 3d animation to make this clear but let’s give words a try first. Imagine with me that you have a Dremel flex shaft mounted to the Darwin X carriage. Imagine that the bit is milling the PCB, moving in the –Y direction, keeping X and Z constant. Milling requires force on the board, which means the board applies force to the bit, in turn. The force on the bit is in the opposite direction of the motion of the bit. Put simply, I have found that this force is enough to twist the X axis guide rods. When the guide rods twist the X carriage rotates about the X vector and pushes the bit into the board.
How to fix? My solution (I have not tried this) is to redesign the X carriage in order to mount the Dremel flex shaft between the guide rods and maybe use heavier rods. Vik pointed out that the BitsFromBytes design has four guide rods. This might also work.
I feel that future designs (Darwin revisions and Mendel) should consider all tool head types, both additive and subtractive. I realize progress is being made with additive board construction, but we still need subtractive! Options are always good, right? Thanks for listening and I look forward to contributing in the future.
Feedback on this issue will be well received. There is more information on http://blog.onshoulders.org.
I've managed to do 'some' milling (your result is better, by the way):
I used an extra reaction bar just like tofletcher who had excellent milling results (PCB isolation routing). Light milling (styrofoam and shallow engraving) is certainly possible with the Darwin frame (without major adjustments to the X-carriage). The topic of Tony is here: http://forums.reprap.org/read.php?1,13435
Here are some pictures of my setup:
First let me say that I'm not a mechanical expert! But I think running the reaction bars across the underside of the other horizontal bars would be better than the wooden/plastic connector that I had made. Even the one with the bearings was complicated but didn't function well. Just make sure the bars are greased up a little bit. If you need a better explanation, just ask. The pictures in tofletcher's topic might illustrate it a bit better.
+1 for the hydra-raptor-like approach of building a machine that can mill as well as extrude.
I've built several machines-- and I'm in the process of building a new one now.
As you've discovered, it takes a surprising amount of stiffness to handle milling forces.
Darwin is designed to be small, compact, cheap, and easy to make-- but not to be stiff enough to mill. Do do that, you need a much stiffer machine-- IE not a darwin. You've got to compromise some of the design decisions to get the other benefits you want.
A couple of simple equations can really help-- a beam deflection equation and a section moment of inertia equation.
the deflection of a beam subject to a centered load, simply supported can be computed as d = w L^3/48*E*I, where w is the load, E is modulus of elasticity, I is the section moment of inertia, L is the length.
section moment of inertia is b*h^3/12 for a rectangular section, where b is the width and h is the height, or pi*a^4/4 for a round section
Putting these equations into a spreadsheet and playing with the numbers helped me get a good 'feel' for what kind of strength you need to handle milling forces with low enough deflection to mill accurately.
A good guess on the side forces on an engraving bit might be 10 lbs. Doesnt sound like much, but for a 24" long rod of steel, .25" diameter, a 10 pound load creates a surprising 0.031 deflection under even this light load.
Even a pretty beefy 0.5" diameter steel rod still deflects 0.002 ( barely acceptable0 under this load.
I'm working on a machine designed to mill non-ferrous metals plus carry a reprap head-- it is a monster compared to darwin, because it is designed to carry 50 lb cutting loads ( typical for aluminum) over a 24" span with 0.001 or less deflection. Takes a lot of material to do that...
Alex, thanks for the equations! I would really like to see some of your work! Any website or images? I would like to see what a machine capable of milling aluminum looks like!
Milling or fabbing circuits would allow a larger degree of specialised parts of a RepRap to be digitally fabricated. The most important thing about this is that anybody who wants to improve for example a PCB will be able to do so and subsequent RepRap makers will be able to choose the latest version instantly.
It would lessen the dependence on centralised vendors and reduce the waste of deprecated inventory. It is important for scalability once RepRap replication starts to take off!
But if you only want to do PCBs, you can take advantage of the fact that there are only two Z values: UP and CUTTING. Make a shoe like a sewing machine with the milling bit projecting from underneath by slightly more than the thickness of the Cu layer. Then mount that on the X carriage so it is flexible vertically, but stiff in X and Y - that is just put it on a beam that's a flat sheet in X and Y and thin in Z.
Then you just lower this 'too low' onto the PCB so the beam flexes slightly and cut away happily letting the shoe set the depth of cut. Lifting 'too high' will clear for in-air moves.
For milling PCBs, incidentally, get the paper-and-resin blanks (called Paxolin or Synthetic Resin Bonded Paper - SRBP), not fiberglass. Otherwise the milling bits won't last.
Raising the work so that it is nearer to the beams will reduce the twisting force. It does increase the rate of change of depth with deflection and so may need to be combined with Adrian's ingenious depth control.
My build is based on a bits from bytes kit with the four X rails. The downside of more rails is that more parts have to be perfectly aligned for the carriage to move. Darwin's design may not be strong but it is very forgiving, it doesn't even require the rods to be exactly parallel.
I think Ian found that too as version 3 of the bits from bytes repstrap has returned to using 2 X rails.
Instead of dragging the bit across the surface to create a line, could you just create a row of tiny dots drilled through the copper layer? Would this exert less force on the rest of the structure?
Yes - it would work in theory. The big problem would be ensuring no shorts between one hole and the next. Also, it would be jolly slow.
Sorry for the delay. Here are a couple of pucs of my first cnc-- it can cut mild steel but has a disappointly small work envelope:
it does milling on soft,small materials very well:
i have actuallly done several boards, one for a reprap extruder.
i use eagle and pcbgcode. here is a picture. it has smallest trace obout 0.018 or so.
...as you can seee, it was not 100% right the first time :(
here is the top
after doing quite a fer boards, i have come to prefer the "inkjet toner tranfer " method desribed here
best. using it, you have the negative of having to drill the holes via cnc(though only if u use through hole components), but the traces can be much smaller-- i,ve gotten as small as 0.012, while easier to solder and much masch faster.
Links to this post: