This project bring CNC control for Raspberry Pi or any ARM based Linux boards.
Typically there is no way to control stepper motors from Linux runtime
environment due to the lack of real time GPIO control. Even kernel based
modules can not guarantee precise control of pulses for steppers. There is
always way out. DMA(Direct Memory Access) is a separated hardware module which
provides high precision for GPIO outputs. This module can copy bytes which
represent GPIO states from RAM buffer directly to GPIO with some clock based
on main chip internal oscillator without using CPU's cores. Using such approach
this project generates impulses for moving stepper motors and that is very
precise way regardless CPU load and OS time jitter.
This approach also allows to use Python language for this project. Typically,
Python is not good choice for real time application, but since project just
needs to set up DMA buffers and hardware will do the rest, Python become the
perfect choice for easy development of this project.
Video demo - YouTube video
Current command support
G0, G1, G4, G20, G21, G28, G90, G91, G92, M2, M3, M5, M30
Commands can be easily added, see gmachine.py file.
Config
All configs are stored in config.py and contain hardware
properties, limitations and pin names for hardware control.
Raspberry Pi implementation should be connected to A4988, DRV8825 or any other
stepper motor drivers with DIR and STEP pin inputs.
Default config is created for Raspberry Pi 2-3 and this wiring diagram:

So having Raspberry Pi connected this way, there is no need to configure
pin map for project.
Hardware
Currently, this project supports Raspberry Pi 1-3. Tested with RPI2. But there is a way to add new boards. See hal.py file. Note: Current Raspberry Pi implementation uses the same resources as on board GPU(memory). So video output will not work with this project. Use ssh connection to board. And do not connect HDMI cable, otherwise project would not run. Probably, increasing of GPU dedicated memory(at least to 64 MB) could solve it and allow to work project and GPU together, but it was never tested.
Usage
Just clone this repo and run ./pycnc from repo root. It will start in
interactive terminal mode where gcode commands can be entered manually.
To run file with gcode commands, just run ./pycnc filename.
Optionally, pycnc can be installed. Run
sudo pip install .
in repo root directory to install it. After than, pycnc command will be added
to system path. To remove installation, just run:
sudo pip remove pycnc
Performance notice
Pure Python interpreter wouldn't provide great performance for high speed
machines. Overspeeding setting causes motors mispulses and probably lose of
trajectory. According to my tests, Raspberry Pi 2 can handle axises with 400
pulses on mm with top velocity ~800 mm per min. There is always way out! :)
Use JIT Python implementation like PyPy. RPi2 can handle up to 18000 mm per
minute on the machine with 80 steps per millimeter motors with PyPy.
Note: Raspbian has outdated PyPy version in repositories(v4.0). Moreover v4.0
has issue with mmap module implementation. Use PyPy v5.0+, download it for
your OS from here.
PyPy installation:
wget wget https://bitbucket.org/pypy/pypy/downloads/pypy2-v5.7.1-linux-armhf-raspbian.tar.bz2
sudo mkdir /opt/pypy
sudo tar xvf pypy2-v5.7.1-linux-armhf-raspbian.tar.bz2 --directory /opt/pypy/ --strip-components=1
sudo ln -s /opt/pypy/bin/pypy /usr/local/bin/pypy
Project architecture
Dependencies
Nothing. Just pure Python code.
GCode simulation
Just a link, mostly for myself :), to a nice web software for gcode files emulation (very helpful for manual creating of gcode files): https://nraynaud.github.io/webgcode/
License
see LICENSE file.
Author
Nikolay Khabarov
