What a PID controller is
A PID controller holds a process at a setpoint. It reads a sensor, compares the reading against the value the operator has entered, and adjusts an output to close the gap. On a machine it is the instrument that decides how much heat, cooling or flow goes in, which is why it sits on extruders, moulding machines, packaging lines, ovens, baths and test chambers rather than only in a laboratory.
It is worth separating a controller from the things around it, because enquiries often conflate them. A temperature sensor measures and produces a signal. A transmitter conditions and passes that signal on. A data logger records it. A controller acts on it. The controller may display the measured value, and usually does, but the display is not what makes it a controller; the closed loop is.
Selection therefore turns on four questions: what the input is measuring, which output the load needs, how many alarm channels the process requires, and what panel space is available. Everything else follows from those, and the order matters, because the panel size on some instruments limits which outputs can be fitted at all.
How a PID controller measures
The difference between a PID controller and a thermostat is what happens near the setpoint. A thermostat switches the output fully on until the setpoint is reached and then fully off, so the process overshoots, falls back and oscillates around the value rather than settling on it. A PID loop looks at three things at once: how far the process is from the setpoint, how long it has been away from it, and how fast it is moving. It then adjusts the output proportionally rather than in steps, and a correctly tuned loop settles closer to the setpoint and stays there.
The output form decides how that adjustment reaches the load. A relay can only switch, so a loop driving a relay varies the proportion of time the contact is closed within a cycle. An SSR drive switches in solid state and can therefore proportion over a much shorter cycle, which suits a fast thermal process. A linear current or voltage output drives a power controller or a valve instead of switching the load directly, and is the form used where the load is too large for a contact. None of these changes the control algorithm; they change how finely its decision can be applied.
The input type is a menu setting rather than a hardware choice on most panel instruments, and that is a commissioning matter as much as an ordering one. A controller configured for one thermocouple type will display a plausible but wrong value when a different sensor is connected to the same terminals, because it is linearising the incoming millivolt signal against the wrong curve. Setting the input type to match the sensor actually fitted is part of bringing the loop into service.
Configuration choices
Most of the specification is fixed by these decisions rather than by the accuracy class. Each one constrains the next, which is why they are settled in order.
| Configuration | Options to fix | What it actually decides |
| Panel size and cut-out | Common square and rectangular DIN sizes, from compact to full width | The space available in the panel, and on some instruments the outputs that can be fitted at all, since a small case cannot carry every combination. |
| Input type | Thermocouple and RTD families, plus analogue current and voltage inputs | Which sensors can be connected without a separate transmitter. The range of types is usually wide but not universal, so an unusual sensor is worth checking rather than assuming. |
| Output form | Relay contact, solid-state drive, linear current or voltage, or a combination | How finely the output can be proportioned, and therefore how tight the control is on a given process. |
| Number of outputs | Single output, or a second output for cooling or for a second zone | Whether a process that needs both heating and cooling can be run from one instrument, which removes a second controller and its wiring. |
| Alarm channels | One or two alarm outputs, or none | Whether the controller can signal an out-of-limit condition independently of the control output, and whether the alarm is absolute or relative to the setpoint. |
| Alarm function | Deviation, band, absolute, timer, and heater-break types | What the alarm actually catches. A heater-break alarm detects a failed element rather than a high reading, and needs its two parameters set to the process to work at all. |
| Communication | Serial bus, with or without a host system | Whether setpoints and readings can be collected centrally, and whether a recipe change can be pushed to several controllers at once. |
| Supply and switching capacity | Mains or low-voltage supply, and the contact rating of any relay output | Whether the load can be switched directly or needs an interposing relay, which is the usual arrangement for an inductive load. |
How to choose a PID controller
| Factor | Why it decides the answer | How to settle it before enquiring |
| Sensor type and range | The input type has to match the sensor, and the measured range has to sit inside the instrument span | Give the sensor type and the temperature or process range it will cover, including the extremes rather than the normal point. |
| Load and its rating | The output form and the contact rating both follow from what is being switched | State the load type, its current and whether it is resistive or inductive. A contact rating is not a load rating. |
| Control quality needed | A process that must hold a tight band needs finer proportioning than one that only has to stay within a wide window | Describe the tolerance required and what happens if it is exceeded, rather than naming a controller type. |
| Alarm requirement | An alarm that catches a failed heater is a different function from one that catches a high reading | Say what the alarm is there to catch. Describe the fault, not the alarm name. |
| Panel space and cut-out | The cut-out is usually already fixed by the enclosure and cannot be enlarged without remaking the panel | Give the existing cut-out, or the space available if the panel is new. |
| Environment at the panel | Heat, dust and washdown act on the front face and on the life of the instrument | Describe where the panel is, not the plant in general, and say whether the front face gets hosed. |
| Integration | Whether the setpoint is set locally or pushed from a host decides whether communication is needed | Name what, if anything, reads the controller remotely, and in which protocol. |
The instrument span and the working range are different figures. A controller selected for the widest temperature the process might ever reach resolves less well at the temperature it normally runs at, so the working point is the one to state.
The nearest alternatives
These are the types an enquiry for a PID controller usually turns out to be, and the point at which the choice is genuinely decided.
| Nearest alternative | What separates it from a PID controller |
| PLC systems | A programmable controller implements a control loop only if one is written and if analogue I/O is fitted, and it then handles many loops at once as part of a wider program. A PID controller is a self-contained instrument for one or two loops, configured rather than programmed, and it keeps working as a controller if the machine program is stopped. |
| Signal converters | A converter changes the form or the range of a signal, for example a thermocouple to 4-20 mA, without acting on the process. A PID controller measures, decides and drives an output. Where a panel meter or a PLC needs a standard signal from a sensor, a converter is the part that supplies it. |
| Control modules | A control module is a building block inside a larger control system, without an operator interface of its own. A PID controller is a complete front-of-panel instrument with a display, keys and an output, which is what a machine operator normally adjusts. |
| Temperature sensors | A sensor produces a signal and nothing more. It is the controller that decides what to do about the value it reports. The two are usually specified together, because an input type chosen without regard to the sensor fitted is the commonest reason a new loop reads wrongly on first start. |
| Data loggers | A logger records what happened and is used to prove afterwards that a process stayed inside its limits. A controller acts in the moment. A process that has to be both controlled and documented needs both, or a controller with a logging host behind it. |
What to include in the enquiry
The first reply can be a quotation rather than a set of questions if the enquiry carries these. Where a figure is not known, send the duty description and we will confirm what it implies.
- Sensor type being connected, and the thermocouple or RTD designation if it is not a common one
- The process range to be controlled, with the working point as well as the extremes
- What the output drives: contactor, solid-state relay, thyristor unit or valve
- The current of that load, and whether it is resistive or inductive
- The panel cut-out, or the space available if the panel is not yet made
- The number of loops to be controlled, and whether heating and cooling are both needed
- What the alarm has to catch, described as a fault rather than as an alarm type
- Whether a host system reads the controller, and over which protocol
- Supply voltage available at the panel
- Whether process documentation is required separately from control