I got this linear actuator, and I had good experience with AdvancedPID.h
However I'm getting weird spikes on the output for no reason.
They only occur downwards, meaning when I change direction.
I tried 2 forces and 2PIDs one for each direction and it was worse...
The documentation on the library is very basic....
Any help would be appreciated.
here is the implementation on a GIGA R1
if (!ActMode) { //only compensate in active mo
ACTUATOR_PID.setTunings(1, .3, 0.1, 0); // Kp, Ki Kd and Clamping
//ACTUATOR_PID.setDerivativeFilter(0.3);//Enable derivative filter (0.8 = strong smoothing, 1.0 = OFF)
// ACTUATOR_PID.setOutputRampRate(100.0); // Protect actuator: Output can change max 100 units per second
DeadBand_int = 6;
int Input = PotnActu - (PotnChair - 500);
Error_M1 = M1_target_pos - PotnChair;
ACTUATOR_PID.setMode(MODE_AUTO);
ACTUATOR_PID.setControllerDirection(DIR_DIRECT);
FORCE = ACTUATOR_PID.run(PotnChair, M1_target_pos); //input and Setpoint
if (Error_M1 >= DeadBand_int and PotnActu < 470 )
{ //in this case the plunger is pushing out
analogWrite(PUSH_PWM, FORCE);
analogWrite(PULL_PWM, 0);
PID_State=1;
} else if (Error_M1 <= -(DeadBand_int) and PotnActu > 20)
{ //Pulling in
analogWrite(PUSH_PWM, 0);
analogWrite(PULL_PWM, abs(FORCE));
PID_State=2;
}
else // This is the dead band between -4 to +4
{
PID_State=0;
ACTUATOR_PID.reset(); // 1. Stops the internal PID calculations
ACTUATOR_PID.setMode(MODE_MANUAL); // 1. Stops the internal PID calculations
FORCE = 0; //
analogWrite(PULL_PWM, 0);
analogWrite(PUSH_PWM, 0);
}
} // end of if !act mode
I'd bet that the mode/deadband changes would cause issues, but it would be hard to tell much without the rest of the code with the timing and mode-switching details.
And unsubstantiated WAG: if you need deadband behavior, maybe it might work better to do a deadband-like preprocessing on Input.
I think that the .reset() and .setMode(...) calls around the deadband edges could easily cause noise in the internals of the algorithm. They set the dT tiny, and lastInput=0.
...and a large, positive dInput over a tiny dT would explain a large negative spike due to the derivative term.
If you want a well-documented PID, use the regular:
And if you want more some advanced features, consider this small patch that exposes some internals to userspace so you can monitor and adjust/manage the integral term/windup yourself:
