pysoem
pysoem copied to clipboard
bnjmnp
PySOES
Are there any plans to begin a PySOES project? If not, any recommendations of where to get started?
I have no plans to start a PySOES project.
To build an EtherCAT slave, one needs more specialized hardware. For example when building a slave that is based on an Raspberry Pi, you need a shield like this: https://www.sg-electronic-systems.com/ethercat-shield-for-raspberry-pi-etherc-v1-6/.
PySOEM has it's difficulties when hard real-time is required, which is the case for most industrial appliances. On problem is, that Python is not meant to be used for real-time computing. Check this SO answer, where the issues are explained a little bit. PySOEM has it's use cases at (automated) testing of non-real-time dependent features of EtherCAT slaves, or some simple appliances that don't require hard real-time. Further, PySOEM can be useful for someone that tries to get familiar with the EtherCAT technology. PySOES will have similar problems when the real-time question is asked. I'm not sure if MicroPython can be used for real-time computing, if so, looking in that direction would be worth considering.
Where to start is still a good question. It depends on where you are already good at, and which hardware you will start with...?
If it comes to writing a wrapper for a C library there are some options. When the wrapper and the library is combined into one binary (even when compling is done during the install), than you call this a binary (Python) extension. A good read is this guide in the official Python packaging documentation, where the options on how the implement a binary extension are listed. For PySOEM I chose Cython. The packaging.python.org is always the first read when creating a Python package, it also covers how to upload your package to the Python Package Index (PyPI), to make you package available to pip.
Thanks for the detailed reply @bnjmnp. My plan was to develop an EtherCAT slave on a linux machine that was running some heavy image processing. In this case, the slave response will not be real time (typically >500ms). Would it make sense to have a RaspberryPi act as the intermediary between the linux PC and the network? Considering that the application isn't real time anyway?
Okay, yes, maybe this could make sens for some use cases. The EtherCAT network could run in real time alongside your special slave that is running not in real-time. I guess your image processing slave will provide some calculated results to the master, that's fine. But beware that a traditional EtherCAT network is not meant to carry lot of image data.
I'll close this. It was good to talk about this, but would like to clean up the open issues a bit. Lets anywas try to drop a link her if anyone creates a PySOES project at some point.
I personally searched for pysoes, but for my purpose pysoes is not what we need (I wrote the comment for my further reference).
https://github.com/OpenEtherCATsociety/SOES/tree/master/soes/hal lists up the hardware list SOES supports. It will not work on other hardwares (including normal PC NIC).
https://www.acontis.com/en/ethercat-simulation.html sells the ethercat slave simulator, but it is not free and not something I can put to CI.
For my purpose, I decided to mock master to use different communication layer.
