Synchronized sub-drivers for 4th order S-curve

[Juanduino]

Hey !

@winder @breiler

I have come across a S-curve algorithm for multi-axis machines, with the potential to enable SimpleFOC sub-nodes, with UGS.

Practically speaking, how can we sync sub-drivers running SimpleFOC?

Is it possible to have the first command reach each motor-controller MCU at the exact same time? If so, then we can use the zero time as the synchronizing factor. Any other idea/approach is welcome. At this point I’m just investigating.

I’m starting this new discussion, since the last discussion about closed loop virtual machine was getting a bit weird.

I believe this S-curve approach can actually be a viable way to have SimpleFOC nodes work w. UGS?

https://github.com/Juanduino/S4_CurvePlanner/tree/main

Best regards

JASEP

[breiler]

What controller are you planning to use for this?

[breiler]

This sounds like a job for a realtime controller, how does this apply to UGS?

[Juanduino]

I suppose there is a potential for UGS to enable SimpleFOC nodes, if the syncing is possible.

[Juanduino]

I mean, if Klipper can do it, why not UGS with 4th order S-curve?

[breiler]

It might be possible to implement a virtual controller in UGS in a similar way as Klipper, that keeps track of the gcode state, planning, positions and inputs. But I still don't think that UGS is the platform for this.

The whole Klipper ecosystem consists of many known parts which can be controlled. For instance, its host running on an dedicated single board computer with a known OS and dependencies. Each controller board is running their own firmware which is communicating with the host through a lightweight protocol. Etc.

UGS runs on Windows x86, Windows x64, Linux x64, Linux ARM, Linux aarch64, MacOSX x64, MacOSX aarch64. This is a lot of architectures to get this working on and to support.

You might as well implement this as an self contained application. Any integration with this controller application can be done using a known protocol (such as the GRBL protocol over TCP) which would allow you to use practically any sender you'd like.

[Juanduino]

Just a side note: The controller has a CAN FD transceiver, but I haven’t even started exploring it. Maybe it is the way to go, as it appears to have some inbuilt syncing in the protocol?

Edit: Ok, I’m quite sure it is the way forward. As I understand it, all nodes receive the CAN commands at the same time, which effectively means that no syncing on the host/nodes is necessary. We can just bundle the relevant move commands, with the expectation, that they will reach the MCUs simultaneously.