The process reaction curve is the picture a process draws of itself when you nudge it and stand back to watch. Put the loop in manual, step the controller output, and record how the measurement responds; that recorded trace is the reaction curve, and it holds everything you need to characterize the loop. This open-loop step test is the hands-on data-collection procedure that feeds the models and tuning rules everything else depends on. This guide covers how to run the test safely, how big a step to make, and how to read the three key parameters straight off the resulting curve.
Process Reaction Curve in one line: A process reaction curve is the trend of a loop's measurement over time in response to a single step change in the controller output, made with the loop in manual so the response is open-loop. It characterizes the process because its shape reveals the process gain, dead time, and time constant. Running the test, called the reaction curve method, is the standard hands-on way to gather the data that model-based tuning rules need.
The procedure is deliberately simple. First put the loop in manual, which breaks the feedback so the controller stops making its own moves and the output holds wherever you set it. This is what makes the test open-loop: you, not the controller, decide the output, so the measurement's response reflects the process alone rather than the controller reacting to it. Let the process reach a steady state at the starting output so you have a clean baseline. Then make a single, deliberate step change in the output, hold it, and record the measurement as it responds until it settles at a new value. That recorded response is the process reaction curve.
Running it safely takes some care, because in manual the controller will not protect the process if it heads somewhere unwanted. The step must be big enough to move the measurement clearly above the noise, so the response is readable, but small enough that the process stays within safe operating limits and does not trip an alarm, spoil product, or stress equipment. A common practical approach is to choose a step that produces a comfortable, visible swing in the measurement without approaching any constraint, and to have an operator watching, ready to return the loop to automatic if anything drifts toward trouble. Picking a quiet period free of other disturbances matters too, since anything else that moves the process during the test contaminates the curve.
Direction and repetition are worth thinking about. Stepping the output up and then, in a separate test, back down lets you check that the process responds symmetrically and helps average out noise, and it can reveal nonlinearity if the up and down responses differ. The whole test is only as good as its cleanliness: a well-executed step on a quiet process gives a crisp reaction curve, while a hurried step during an upset gives a muddled one that yields untrustworthy parameters.
The reason the reaction curve is so valuable is that three key process parameters are visible directly in its shape. The process gain is read from the total distance the measurement traveled: divide the settled change in the measurement by the size of the output step you made, and you have how much the process moves per unit of output. A large measurement swing from a small step means high gain; a small swing means low gain. This single ratio tells you how strongly the process responds and how hard a controller should push.
The dead time is read from the beginning of the curve. After you make the step, there is typically a flat interval during which the measurement does not respond at all, the pure transport or reaction delay before the process registers the change. The length of that flat stretch before the measurement starts to move is the dead time, and it is the parameter that most limits how tightly the loop can be controlled. The time constant is read from the rising or falling part of the curve after the delay: it describes how quickly the process approaches its new value once it has started moving, typically gauged by how long the response takes to complete most of its journey. A steep, quick approach means a short time constant; a slow, gentle one means a long time constant.
Together these three, gain, dead time, and time constant, are exactly the parameters of the first-order-plus-dead-time model, which is why the reaction curve is the natural companion to that model. This is also where the reaction curve complements a bump test: the bump test is the act of making the step and gathering the data, and reading the curve is the analysis side that extracts the numbers. Once the three parameters are off the curve, they feed straight into tuning methods such as lambda or IMC, which compute controller settings from them. The reaction curve, in other words, is the bridge from a physical test to a tuned loop.
A process reaction curve is just a recorded trend of output and measurement around a step, which is precisely what a SCADA historian captures at high fidelity. Running the test benefits enormously from good historization: the cleaner and finer the recorded trend, the more accurately the gain, dead time, and time constant can be read. When the SCADA system logs the step and the response, the reaction curve is preserved for analysis and reference rather than living only on a strip chart or in an engineer's memory.
In a cloud SCADA platform such as Merobix, this lets an engineer set up and capture a reaction curve on a remote loop and then analyze it from the office. The output step can be coordinated with an operator on site, the response historized, and the parameters extracted from the stored curve, so characterizing a loop no longer requires standing at its console for the duration of the test. Because the curves are archived, a loop can be re-tested later and its new reaction curve compared against the old one, revealing whether the process has changed through fouling or wear.
This matters across every industry the platform serves, since reaction curves are how loops get characterized everywhere. Oil and gas separators and heaters, water treatment loops, power plant thermal systems, and manufacturing processes all give up their gain, dead time, and time constant to a well-run step test. Having the curve captured and stored centrally turns a one-off field procedure into repeatable, reviewable data, letting a small team characterize and tune loops across many sites from the same historized trends operators rely on day to day.
Put the loop in manual so the feedback is broken and the output holds where you set it, let the process settle at a steady baseline, then make a single deliberate step change in the controller output and record the measurement until it settles at a new value. Doing it in manual makes the response open-loop, reflecting the process alone. Choose a quiet period free of other disturbances so the curve is clean.
Big enough that the measurement moves clearly above the process noise so the response is readable, but small enough that the process stays within safe limits and does not trip an alarm, spoil product, or stress equipment. A comfortable, visible swing that stays well clear of any constraint is the goal, with an operator watching and ready to return the loop to automatic if it drifts toward trouble.
Three parameters. The process gain is the settled measurement change divided by the size of the output step. The dead time is the flat interval before the measurement begins to respond. The time constant is read from how long the response takes to complete most of its journey once it starts moving. These three are the inputs that model-based tuning rules like lambda and IMC use to compute controller settings.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.