Steady-state offset, often called droop, is the permanent gap between setpoint and the actual measurement that a proportional-only controller leaves behind once the process settles. It is not a tuning mistake - it is a mathematical certainty of pure proportional action, and understanding it is the single clearest reason integral (reset) action exists. This guide explains where the offset comes from, why load changes make it worse, and how reset drives it to zero.
Steady-State Offset (Droop) in one line: Steady-state offset (droop) is the residual difference between setpoint and process value that remains after a proportional-only controller reaches equilibrium; because a P controller needs a non-zero error to produce a non-zero output, it can never fully close the gap, and integral action is added specifically to accumulate that lingering error away over time.
A proportional controller sets its output in direct proportion to the error, plus a fixed bias: output equals gain times error, added to the bias term. To hold a valve at, say, 60 percent open against the current load, the controller must generate a specific output. If the bias was set for a different load, the only way the proportional term can supply the missing amount is by carrying a standing error. In other words, the measurement has to sit slightly off setpoint forever so that the gain multiplied by that error produces exactly the output the process demands.
This is why the phenomenon is called droop: the process value droops away from setpoint and parks there. On a level loop it might settle two inches low; on a temperature loop it might hold three degrees under target. Nothing is broken - the loop is stable, quiet, and repeatable. It is simply obeying the proportional equation, which has no mechanism to remember or accumulate the leftover error. The gap is the price of using proportional action alone.
The size of the offset depends on two things: controller gain and how far the operating load has moved from the point where the bias was set. A higher proportional gain shrinks the offset, because a smaller error is now enough to swing the output the required amount. That tempts operators to crank the gain up - but push it too far and the loop starts to oscillate and hunt. So proportional-only control forces an ugly trade: either accept a visible offset at low gain or accept instability at high gain. There is no gain that gives both zero offset and good stability.
Load is the other driver. Bias is usually set so the offset is zero at one specific load - the design condition. As throughput, ambient temperature, or upstream pressure drifts away from that condition, the required output changes, the standing error grows, and the droop widens. This is exactly why a P-only heater that sits perfectly on setpoint at midday can run several degrees cold on a cold night: the load moved, and proportional action had to invent a bigger error to compensate.
Integral action - the I in PID, also called reset - is the direct cure for droop. Reset continuously integrates the remaining error over time and keeps nudging the output until that error is zero. Because it accumulates rather than reacts, it does not need a standing error to hold an output, so it eliminates the offset entirely and lets the process land exactly on setpoint regardless of load. The classic problem droop describes is precisely the problem reset was invented to solve, which is why nearly every real loop in the field runs PI or PID rather than P alone.
On a cloud SCADA platform such as Merobix, steady-state offset shows up plainly in the trends. A tag that consistently settles a fixed distance below setpoint after every disturbance - and whose gap widens as production rate changes - is the signature of a loop running with too little or no reset. Because Merobix stores long-horizon history for every process value and setpoint, an engineer can review months of a level or pressure loop from a browser, spot the persistent droop, and flag the loop for a controls tech to add or retune integral action. SCADA does not tune the loop, but it makes the offset visible so someone can.
Because its output is gain multiplied by error plus a fixed bias, it needs a non-zero error to produce any output different from that bias. To hold the valve where the current load requires, the measurement must sit permanently off setpoint. That leftover error is the offset, and proportional action has no way to remove it.
Add integral (reset) action. Integral action accumulates the remaining error over time and keeps adjusting the output until the error reaches zero, so the process lands exactly on setpoint at any load. This is why loops run PI or PID instead of proportional alone. Raising proportional gain only shrinks offset and risks instability - it never fully removes it.
No. Deadband is a deliberate zone around setpoint where the controller takes no action, used to stop valve hunting and nuisance alarms. Droop is an involuntary residual error left by proportional-only control even when the controller is actively responding. Deadband is a design choice; droop is a mathematical consequence of using pure proportional action.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.