# PID parameters are given in percentage

I have a question about setting the parameters of PID controller if they are given in range between 0.1% and 100%.

I have a device with the PID controller and P, I and D shares can be set separately in a range from above. I want to make the setting process automatic. But in all algorithms about PID controllers I read the gain, reset and derivative are not given in percentage.

How can I calculate gain, reset and derivative from this parameters? What is the best strategy to automate my device if I use all parameters in percent directly? Any answers and links are appreciated.

With best regards, mikhail14

• So, provide a link to the controller you use. Alternatively set integral and derivative to zero and set with proportional percentage to see what the gain is when the system is stable. – Andy aka Oct 24 '16 at 8:36
• It is a TEC controller from Thorlabs thorlabs.com/drawings/… I understand how to set the value manually but i am interesting to make this automatically. So if the temperature changes, the PID corrects it without my help – mikhail14 Oct 24 '16 at 9:03
• That's still not clear to me, mikhail14. And I used to write PID controller software. What do you mean by "if the temperature changes, the PID corrects it without my help?" With a properly tuned PID, that's what it does. If the temperature changes (for example, someone opens a door or the air conditioner turns on or some other disturbance takes place), the PID is supposed to keep things in control. That's what it does. It's why there is a set point. It's why you have integrals. It's why you have a derivative term. So what MORE do you want it to do, than that? Are you changing the set point? – jonk Oct 24 '16 at 9:25
• I think I express myself not clear. I reask my question. If I have P in percent, what does it exactly mean? For example, I have I=20%. 20% of what and how does it depend on the reset? – mikhail14 Oct 24 '16 at 10:30
• Ah. I couldn't tell from your earlier writing. Looks like you have some answers, though, that should help some. – jonk Oct 24 '16 at 17:57

Only the P coefficient is a dimensionless quantity, and therefore can be expressed in percent.

The I coefficient is in units of 1/time, and the D coefficient in units of time. Percent therefore makes no sense unless the percent of what exactly is clearly specified and in the right units. There is no way for us to guess how percent may be applied to these in your case. Consult the documentation for the specific controller.

Responding to your

I reask my question. If I have P in percent, what does it exactly mean? For example, I have I=20%. 20% of what and how does it depend on the reset?

If the controller has a full scale of 500 units of something at the input, and the output swings from 0 to 100% of something else, then a 20% P means that a +/-50 input unit change in error will swing the output over the full range.

So if the setpoint is 250 degrees C on a 0-500°C controller, a 20% P means that -if there is no manual reset term and no feed forward- and the controller requires 40% power to maintain 250°C it will sit at a temperature that is 2% of full scale (10°C) high (assuming the integral term is disabled). That +10°C error generates a difference from 50% of power of -10/500 * 100%/20% = -10%.

Reset (aka integral) is in units of 1/time so % only make sense in a context where some full scale time (be it 1 hour or 1 microsecond) has been defined, as @Olin says.