This is the first installation in my Adventures in Motion Control series.
$JOB we build industrial CNC machines, and while developing a simulator
for our machines I noticed a distinct lack of online resources on how
they work under the hood. Hopefully this series will address the situation.
The goal for this series is to develop a simulator which will accurately reflect how an embedded motion controller is implemented internally.
This motion controller will be designed to control a 3D printer, with the eventual idea being to compile everything to WebAssembly so the simulator will run in the browser and users can explore it using a basic web UI.
The first step in implementing any project is to specify exactly what you want it to do (and not do) and the constraints imposed by hardware.
This project will simulate a 3D printer. Most 3D printers have 3 orthogonal axes (hence the 3D bit) controlled using stepper motors. The printer will have a pre-defined “working area”, with limit switches at the ends of each axis to make sure we don’t go past the end of travel.
The only way an outside user can interact with the printer will be via a RS-232 connection (which we will model as a simple, bi-directional byte stream). RS-232, often referred to as Serial, provides no guarantees around framing (breaking bytes into individual “messages”), detecting corrupted data, or that the other end has actually received our message. This will all need to be accounted for.
The printer won’t have any other physical inputs (e.g. push buttons or sensors) or outputs (e.g. an LCD display).
Jobs will be sent to the motion controller over the serial connection in the form of a gcode program.
The motion controller will need to expose some of its internal state to the user for diagnostic purposes.
Users should be able to inject “errors” to see how the simulator would react.
The motion controller should be able to execute pre-defined automation sequences. For example, bed levelling and axis calibration.
The motion controller will have a limited amount of non-volatile memory (e.g. an on-board flash chip) which can be used for caching jobs and settings.
The simulator won’t need to support flashing new “firmware”. This means we won’t need to worry about having a bootloader, or deal with all the complexity around rewriting firmware in-place and worry about “bricking” our simulator.
The motion controller won’t have access to nice things like multi-threading and may only have access to a small amount of RAM (e.g. 192KB).
Now we’ve got a rough understanding of the project the next step is to start implementing it. The way I like to do things is by stubbing out just enough of the top-level architecture to get a Hello World working then dive into one area, preferably something fundamental so other systems can start making progress.
To see the code, check out the Michael-F-Bryan/adventures-in-motion-control repository on GitHub.