Stereolithography file to GCode

[rweatherly]

I have been working on decoding .stl files to eventually obtain Z heights for machining. So far, I have translated the binary .stl to extract the 3 coordinates for each facet's vertex. Now I am trying to decide how best to slice through a section to find the machining coordinates.

Anybody else tried this?

(I know there are commercial programs that will do this, but I am a DIY type and cheap, too)

[lakeside]

no I would just open the .stl file( in mastercam) and verify location from cad system if you need to send me an e-mail and I could do it for you
mikelakeside@comcast.net

[rweatherly]

Thanks, but this is more of a generic problem. I have some .stl files that I want to be able to route on my CNC router. I want to write my own program to do this. I already have written a routine for machining based on a bitmap, and I would like be able to do a .stl file too.

[lakeside]

try downloading a copy of mastercam demo .stl is one of the formats that they open

[High Seas]

stlworks (www.Imsrv.com as I recall)
it works and there is /was a trial option. Nearly as good as a free for a DIY....

[jed102]

I use MeshCAM for stl files. Really easy! Low cost too, and demo available.

Frank

[klick0]

rweatherly,

Don't worry, i'm not gonna tell you to download something, I understand you want to write it yourself I've written a program to do exactly what yer talkign about, it's not fancy, but it gets the job done. My file format of choice was DXF though, but basically my program loads the DXF into my Model class, which is composed only of Triangulated Polygons, which sounds like what a STL file has, just triangulated polygons. For a while i was going to support both quads and tri's but realized there was no benefit and a waste of time hehe...

anyway, a brief way of how I did it was load the stl into a Model class (granted i'm talking more java/c++ fashion here then straight c, not sure what you are programming in, doesn't really matter). Once all the polygons are loaded, you need a Ray intersection algorithm. Which I attempted for about an hour to do and mine only worked 99% of the time, which isn't enough for G-Code unless you want random deep spots poked into your model. So I downloaded a library that had ray->polygon intersection routines and just used that.

Of course it's all easy in theory, I had a gridsize setup, and basically the G-codes are done based upon that, then there's an optimization routine to remove G-code that is milling large 0 depth areas... also i wrotee a routine to optimize the G-code so there are way fewer lines, basically if you take 3 points of a g-code, detect the degree difference between them, if it's low enough remove the middle point, interating through the entire g-code file... LinuxEMC was having a fit with my enourmous g-code file and wasn't processing enough, so this helped a lot, and the model remains intact.

I'm not really sure i'm answering anything here, basically the key is a ray->polygon intersection routine, which you run against EVERY polygon, unless you want to get into BSP's and all sorts of complex routines, but lets keep it simple, just go through every polygon. and of course the ray start point is like +1000Z X Y (whereever you want to calculate) then it points straight down -1000Z same XY... and you loop through them and find the intersection height. then XY you move around in a grid fashion of course.

My advice is to find library's to help you out.. unless you want to learn 3d math, which is fun, but will take a while to learn. game developer sites is a great resource for this stuff though..

good luck to ya,
Ross

[rweatherly]

Thanks, that's basically what I am trying to figure out how to do. I will check out the sites.

[rweatherly]

I think I figured it out. I really did not understand the ray intersection equations, so I simplified it a bit (at least for me).

Once I converted the .stl from binary to decimal (not a small task), I set up a matrix that had a position for each G code machining move. I looked at each .stl triangle to see what machining points lied inside it, then calculated the Z height for those points, filling in the matrix as I went through each triangle. The completed matrix was used to generate the G-code.

This seems to work, although I have not machined it. (I have a routine that converts the z heights to a bitmap so I can get an easy impression of what the final machining looks like -- it's a lot quicker than using an nc simulator)

[kfb006]

It is truely a good idea to use STL model to produce G code. I am also intrested in this project.

[holger]

Hello,
i am trying same thing with a stl. But i try to build a offset surface wuch will be the tiilpath for end ballmill.
Slicing through the model is very easy. If you slice along the z- axis take any z coo you want. Then sub fro, every triangle this z value. If you get for a corners of the triangle the same sign (+ or -) this will nit cit the plan. If the a different you must calculate the crossingpoints from the triangle edge and the plan. This point will give a line on the plan. U can look fir the source if you want. Written in C.

regards Holger

[Death Adder]

Originally Posted by holger Hello, i am trying same thing with a stl. But i try to build a offset surface wuch will be the tiilpath for end ballmill. Slicing through the model is very easy. If you slice along the z- axis take any z coo you want. Then sub fro, every triangle this z value. If you get for a corners of the triangle the same sign (+ or -) this will nit cit the plan. If the a different you must calculate the crossingpoints from the triangle edge and the plan. This point will give a line on the plan. U can look fir the source if you want. Written in C. regards Holger

Yes, what you described sounds like how I would do it:

You define a plane for a given Z Level. Then you intersect every triangle in the model with the plane. Most triangles will not intersect. Some will intersect at a point and some will intersect at a line. All of the intersections then need to be sorted in either clockwise or CCW order and linked together. This gives you a profile around the model. What complicates things is that there could be multiple pockets and each one needs to be independently sorted and linked.

For a ball nose cutter and Z level machining you can pretend the cutter is a sphere of the same radius as the cutter. Then calculate the offset based on the radius and depth of cut. Hopefully you can make the assumption that the model being cut is convex with respect to Z and thus every deeper level is at least as large as the higher level. This simplifies gouge checking and will be true for most any model which could be cut in 3 axis.