Local CAM

A CNC mill is a computer-controlled cutter that carves a shape out of a solid block. Mine is a small benchtop machine with three axes of motion — left-right, front-back, up-down — and an off-the-shelf PC for a controller.

Two pieces of software stand between a 3D model and that machine. CAM decides the toolpath: the route the spinning cutter takes through the material, and how fast and how deep it goes at every point. A post-processor then translates that route into G-code, the plain list of moves the machine reads. G-code is the only thing the mill ever sees. Everything else is a plan.

CAM is normally something you rent. I wanted it to run on my own hardware, offline, on a machine whose source I can read. It went wrong in a series of specific ways, and those are the part worth writing down. This is an ongoing story; the parts are below, in order, and new ones land as the work does.

A 3D mill simulator showing a rectangular recess being carved into a block of plastic clamped in a vise, the cut path drawn as yellow and magenta ribbons coloured by tool load, with a readout bar below showing cut time, engagement percentages and material removed — My simulator part-way through cutting a rectangular recess. Each ribbon is one pass of the cutter, coloured by how hard it is working. The block is the raw material; the shapes under it are the clamp holding it down, modelled from the real one.

Where it stands

Three parts so far. The software is proven in simulation: the first cutting file passes every check, and no metal has been cut. The next part is the first real cut.

What it’s really about

I didn’t write my own CAM to make better toolpaths than Fusion. I wrote it so I could see the assumptions. Rented software makes its decisions out of sight — how deep is too deep for a cutter to plunge, whether your machine can look up its own tool lengths, whether a number in a dialog is a hard limit or an average. Writing my own put every one of those decisions in front of me.

Most of the wrong ones turned out to be mine.

When the mill finally cuts that coupon, I expect it to find something else I was sure about.

The parts, in order:

The picture was not the program

I wrote my own CAM — the software that plans a cutter’s route through the material — for a benchtop mill, and built a simulator to check its work. This is the first thing the simulator found. The first version drew its toolpaths in the browser. They looked fine, and I trusted them. Then I built a simulator that reads back the posted file — posted meaning it has been through the post-processor, the step that turns a planned route into the G-code the machine reads. So it is the file the machine will run, not the plan that produced it. I pointed it at a program every existing check had passed. ...

What the references taught

Part 1 built a simulator that grades a toolpath by engagement — how much of the cutter’s width is buried in material at each moment — with 55% as the limit. This is what happened when I graded everyone else’s toolpaths with it. The one that beat both has no corner code The next step was to stop guessing and read what everybody else does. I ran five open-source toolpath libraries over the same test shape and graded them with my own measure — including FreeCAD’s Adaptive, which holds under the 55% line just as Fusion does, and is free. ...

Half the program was air

The simulator from Part 1 plays a G-code file move by move and keeps a clock. This is what the clock said about the first real cutting file — and the most recent thing measurement corrected. When I watched the simulator play that file, the cutter crawled through open space. Those were repositioning moves — hops from the end of one pass to the start of the next, over ground already cleared — but they ran at cutting speed rather than travelling speed. Across the whole file, 48.8% of the job is the cutter moving through nothing, and the true cycle time was 8 minutes 57 seconds, not the 4:34 I had been telling people. I had read the “cutting time” figure and repeated it as the total, when the next line said how much air came with it. ...

Where it stands#

What it’s really about#

Where it stands

What it’s really about