R
I'm designing a charcoal grill/smoker controller. Output is a duty cycle for a small blower/fan. Input is a thermocouple inside the grill.
In reading about PID methods it seems that the Integral term is the suggested way to arrive at the steady-state output. For example there is going to be more integral error trying to reach 400 degrees F than there is trying to reach 200 degrees F, so tuning the I factor should be sufficient to set the output higher at steady-state for the higher setpoint.
One problem with this is that the integral gets reset sometimes. Example, when the controller is power cycled. User may not even turn it on until the temp is already close to the setpoint.
So the integral sum may not have been gathered leading up to the higher setpoint or it may have been reset.
So I'm thinking of adding a bias term. However, the only article I read mentioning bias referred to it as a constant. But I think I want it proportional to the setpoint. Higher setpoints need a higher bias.
Is there some literature on this idea of including a setpoint factor in the PID equation? SPID?
I'd eventually like to have the whole equation auto-tune. This factor would be the easiest to auto-tune. Observe oscillation and just take the average of output over whole periods of oscillation. Divide by Setpoint and you have you're new setpoint factor.
I'm a bonafide newbie so feel free to say RTFM, but just tell me where the M is.
Thanks
In reading about PID methods it seems that the Integral term is the suggested way to arrive at the steady-state output. For example there is going to be more integral error trying to reach 400 degrees F than there is trying to reach 200 degrees F, so tuning the I factor should be sufficient to set the output higher at steady-state for the higher setpoint.
One problem with this is that the integral gets reset sometimes. Example, when the controller is power cycled. User may not even turn it on until the temp is already close to the setpoint.
So the integral sum may not have been gathered leading up to the higher setpoint or it may have been reset.
So I'm thinking of adding a bias term. However, the only article I read mentioning bias referred to it as a constant. But I think I want it proportional to the setpoint. Higher setpoints need a higher bias.
Is there some literature on this idea of including a setpoint factor in the PID equation? SPID?
I'd eventually like to have the whole equation auto-tune. This factor would be the easiest to auto-tune. Observe oscillation and just take the average of output over whole periods of oscillation. Divide by Setpoint and you have you're new setpoint factor.
I'm a bonafide newbie so feel free to say RTFM, but just tell me where the M is.
Thanks