Anyone written a simple PID routine for a PIC???

[Carlos Barberis]

HI
I am looking for a simple pid routine to control a small heater using the pwm output of the pic. Any hints would be appreciated.
___________________________
This message was ported from CCS's old forum
Original Post ID: 2637

[Eric Minbiole]

:=I am looking for a simple pid routine to control a small heater using the pwm output of the pic. Any hints would be appreciated.

I recently finished a project that leviated a metal ball using an electromagnet. It used a PIC to control a PID loop which kept the ball floating stabily.

Here's the code for my PID loop. I'm sure it's not the best PID code out there, but it should give you a starting point to work from. The only tricky part of the code is in calculating the differential term. Rather than taking the difference of two adjacent samples, I averaged 4 old samples, and 4 new samples, and took the difference between the averages. This made the loop more immune to noisy samples.

Hope it helps!


// control loop constants
float Kp; // proportional gain
float Ki; // integral gain
float Kd; // differential gain

// terms
float Tp; // proportional term
float Ti; // integral term
float Td; // differential term

// circular queue vars
int i = 0;
signed long error_history[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
int8 queue_pos = 0;
int8 temp_pos;
signed long prev_ave;
signed long cur_ave;

signed long error;
signed long DeDt; // change in error over time
signed int32 error_sum = 0; // sum of errors, for integral term
signed long desired_power;
int8 power;
int8 setpoint;
float temp_float;

// **** set the loop constants (Kp, Ki, Kd) to something useful

// main control loop
while (1)
{
// calculate the raw error
// negative = disc too low
error = (signed long)read_adc() - setpoint;

// calculate the proportional term
Tp = -Kp * error;

// calculate the integral term
error_sum = error_sum + (signed int32)error;
temp_float = error_sum;
Ti = Ki * temp_float;

// use a circular queue to save a history of the last 8 samples
// this will be used to calculate the differential term
error_history[queue_pos] = error;
queue_pos++;
queue_pos &= 0x07; // keep in 0..7 range
temp_pos = queue_pos;

// calculate the average for the 4 oldest samples
for (i = 0, prev_ave = 0; i < 4; i++)
{
prev_ave += error_history[temp_pos];
temp_pos++;
temp_pos &= 0x07;
}

// calculate the average for the 4 most recent samples
for (i = 0, cur_ave = 0; i < 4; i++)
{
cur_ave += error_history[temp_pos];
temp_pos++;
temp_pos &= 0x07;
}

// calculate the differential term
DeDt = prev_ave - cur_ave;
Td = Kd * DeDt;

// calculate the desired power
desired_power = (signed long)(Tp + Td + Ti);

// set the correct power
if (desired_power < 0)
power = 0;
else if (desired_power > 255)
power = 255;
else
power = desired_power;
set_output_power(power); // this could be pwm duty, etc

// wait between samples
delay_ms(2);
}
___________________________
This message was ported from CCS's old forum
Original Post ID: 2639

[Neutone]

This is turned out to be a really easy way to impliment a PID control. It works good if it's tuned right. I use this to heat a heat exchanger with a variable heat load. My pulse width is 20 seconds and sampel rate is 5 times per pulse. Any of the number can be changed.

All variables are floats
//PID PWM calculation routine for SupplyTemperature

errorNOW = Setpoint - SupplyTemperature;
PWMonTIME = PWMonTIME + (PWMp * errorNOW ) + (PWMi * (errorNOW - (PWMd * (errorNOW - errorLAST))));
errorLAST = errorNOW;
if (PWMon > 190) PWMon = 190; //maximum on time
if (PWMon < 10) PWMon =10; //minimum on time
PWMoff = 200 - PWMon;



:=HI
:=I am looking for a simple pid routine to control a small heater using the pwm output of the pic. Any hints would be appreciated.
___________________________
This message was ported from CCS's old forum
Original Post ID: 2658

[future]

Are PWMp, PWMi and PWMd values from 0 to 1?

I want to use PID to control a stepper motor, so it has to know how many steps and direction to apply to correct the error.

[Neutone]

[sonicdeejay]

thx guys...been looking for this too...I will study the code...

[sonicdeejay]

[andibaciu]

@Eric Minbiole
can you put the principle how to set the Kp, Ki, Kd ???

any help is good ....

[newguy]

[kilarn]

A good place to start for values is to do a quick search for the Ziegler-Nichols Decay ratio technique... this is only a starting place though...

The best method I have found is to use Root Locus with a program like MATLAB... but this again depends on how well you have modeled your design.

Also to find the Kp you can use a Routh Array and/or what is called the Magnitude condition.

These all come from classical control theory.

Hope this helps.

[mmprestine]

here is what was posted in the past. try the search.

http://www.ccsinfo.com/forum/viewtopic.php?t=32995&highlight=