Closed Loop example in position mode and velocity mode #97
Replies: 4 comments 20 replies
-
|
In speed mode: and INI: |
Beta Was this translation helpful? Give feedback.
-
|
in position mode: |
Beta Was this translation helpful? Give feedback.
-
|
Is the HAL configuration correct in both modes? |
Beta Was this translation helpful? Give feedback.
-
|
The system you describe is not a true closed loop example. A real closed loop example would take an input from and linear encoder and using that as feedback for LinuxCNC. Would love to try that out with a simple setup if somebody would be willing to sponsor the materials... In your example you use the PID component to translate between position as commanded by LinuxCNC and velocity as required by the FPGA (note: there is no trajectory planner in the FPGA! The FPGA relies on a steady stream of new velocities to form the position profile. this is very close to the architecture of LinuxCNC stepgen component.). I have no experience in tuning these components and also believe that it might be machine dependent. The FPGA does have a position mode. In this case the feedback loop is closed within the component. One of the known limitations is humming at very low speeds or stand-still. The reason for this is that the packages arrive at random times, i.e. the control loop of LinuxCNC has some jitter around it. When a position is commanded and held, there might be a small overshoot at each step, which is compensated and leads to a new overshoot, etc.... In LinuxCNC stepgen this is compensated by just doing the remaining step when the difference is small enough. Maybe I've to add the same functionality in the FPGA to prevent this humming. But this might require sending positions instead of velocities instead and a large re-write of the stepgen... |
Beta Was this translation helpful? Give feedback.
-
|
hello peter |
Beta Was this translation helpful? Give feedback.
-
The configuration of the stepgen will be: "modules": [
{
"module_type": "stepgen",
"instances": [
{
"pins" : {
"stepgen_type": "step_dir",
"step_pin": "j7:0",
"dir_pin": "j7:1",
"index_pin": "j7:2"
},
"soft_stop": true
}
]
}
],When the |
Beta Was this translation helpful? Give feedback.
-
|
I get the logic. It looks good! |
Beta Was this translation helpful? Give feedback.
-
|
I think the version is ready to be tested. This branch now contains:
When this test goes well, I think both this issue and #91 can be closed. |
Beta Was this translation helpful? Give feedback.
-
|
thank you for your work! |
Beta Was this translation helpful? Give feedback.
-
|
@hmnijp : did you have had any chance to test the changes? |
Beta Was this translation helpful? Give feedback.
-
|
And I remembered that I found another small bug concerning HOME_USE_INDEX that I forgot to write about: when index is found, linuxcnc resets
At the end of the homing, position and feedback are zero, as they should be, there is no error. But if HOME_OFFSET =/= 0 is used, then at the end of the procedure
It's the difference between pos and feedback that causes the follow error after homing. I think when using index-enable we should set This pin (encoder.00.position_offset) is mentioned in the documentation, but it doesn't seem to exist in the current version of the encoder component...
|
Beta Was this translation helpful? Give feedback.
-
|
You're correct that the |
Beta Was this translation helpful? Give feedback.
-
|
See #111 The Both |
Beta Was this translation helpful? Give feedback.
-
How to create a closed loop system in Litex?
Beta Was this translation helpful? Give feedback.
All reactions