If you need to design an optimized control loop for a hardware control problem， consider trying a proportional integral derivative （PID） controller. In this article， Robert explains that PID regulations are simple to code， wire， and tune to find the results you‘re looking for.

Welcome back to the Darker Side. The vast majority of real-world systems are based around feedback loops， which are used to manipulate the inputs to a system to obtain a desired effect on its outputs in a controlled way. Your DVD player uses a feedback loop to drive its spinning motor for a precise rotation speed. Your mobile phone has a feedback loop to adjust its transmit power to the required level. And， of course，your car has plenty of feedback loops（not only in the cruise control module）。Systems incorporating feedback work even if the relationship between the desired value （e.g.， the spinning speed of a motor） and the controlled one （in that case， the current applied to the motor） is not straightforward. Delay effects， inertia， and nonlinearities make life more interesting， and so do external conditions. For example，a cruise control device needs to apply more torque on the motor if you are climbing a hill. It should keep overshoots as small as possible to avoid getting a fine after you get to the top of a hill.

On the theoretical side， these problems have been well-known for years and are covered by the Control System theory. The theory explains how to design an optimized control loop for a given problem， at least if the problem is well formalized. The theory-which started with the works of James Clerk Maxwell in 1868， followed by a lot of inspired mathematicians，such as Alexander Lyapunov and Harry Nyquist， to name a few-is not simple. The first pages of control system books usually start with mathematical notions like pole placement，Z-transforms， and sampling theorems，which nonspecialists may find difficult to deal with， even if good books have adopted a more engineer-oriented approach （e.g.， Tim Wescott‘s Applied Control Theory for Embedded Systems，Newnes， 2006）。 Should you give up？No. Fortunately， some classical control algorithms are applicable to a lot of problems. More importantly， for design guys like us， they are easy to implement in firmware or hardware. The most usual is the ubiquitous proportional integral derivative （PID） control， which has one interesting characteristic： it is tuned by only three or four parameters. As you will see， the parameters can be determined by empirical methods， without even knowing the exact behavior of the controlled process. Don‘t get me wrong. I am not saying that a PID will solve any control problem， especially if the system has multiple inputs and outputs. But you may want to give it a try before digging into more complex solutions. In any case， it's a must-have for every engineer. So， let‘s go with PID！

A BASIC CASE
The most basic example of a feedback loop is a temperature controller used to drive a heater or cooler to get a precise temperature （or temperature profile） on a temperature sensor on a device. Regular readers may remember my article about a home-made reflow oven controller based on a firmware PID loop （"Easy Reflow：Build an SMT Reflow Oven Controller，"Circuit Cellar 168， 2004）。This month， I will take an even simpler example. Twenty years ago， I bought a Weller VR20 soldering iron，which had a built-in resistive temperature sensor. At that time， I couldn‘t afford the corresponding power supply，so of course I built my own （see Photo 1）。 It was easy. I used a 24-V transformer， a potentiometer to tune the temperature， a digital voltmeter （Remember the old CA3161/CA3162 three-digit voltmeter chipset？）， and a crude "control system." I used an LM311 comparator to drive an output TRIAC on and off， whether the measured temperature was above or below the threshold. For control system specialists，this is the most basic form of a "bang-bang control" algorithm-and it is so for obvious reasons. （Just imagine yourself driving your car with only two settings： throttle fully open or brakes fully engaged.） It was working，but the temperature regulation was oscillating around 3°C above and below the preset value. That corresponded to an 18-mV oscillation on the sensor measurement （see Photo 2）。