The full PID controller is not always the right tool. Sometimes an engineer deliberately switches off the integral action and runs proportional only, and sometimes they switch off the proportional action and run integral only. Each stripped-down mode has a distinct personality and a set of jobs it does well. This guide explains what P-only and I-only control do, the classic applications such as surge and averaging level control, and the tradeoffs that make each choice deliberate rather than a mistake.
P-only / I-only control in one line: P-only control uses proportional action alone, giving a fast, stable response that accepts a permanent offset from setpoint, while I-only control uses integral action alone, giving a slow, smooth response with no proportional kick and eventual return to setpoint. Engineers choose them deliberately for tasks such as surge and averaging level control, trading tight setpoint holding for smoothness and simplicity.
Proportional-only control drives the output in direct proportion to the current error. The bigger the deviation from setpoint, the more the output moves, but only while the error exists; the moment the process is exactly on setpoint the proportional contribution to the correction disappears. Because a real process usually needs some standing output to hold a load, a P-only loop settles at a point where a residual error remains, just enough error to generate the output the process demands. This standing error is the well-known offset of proportional control. In exchange for accepting it, P-only control is fast, simple, and inherently stable, with no integral to wind up and nothing to slowly build.
Integral-only control drives the output based on the accumulated error over time rather than its instantaneous value. As long as any error persists, the integral keeps moving the output in the direction that reduces it, and it stops moving only when the error reaches zero. This gives I-only control its defining trait: it eliminates offset and eventually parks the process exactly on setpoint. But it does so slowly and smoothly, ramping the output gradually, and crucially it produces no sudden output change when the setpoint or process variable steps, because there is no proportional term reacting to the instantaneous error. Its motion is always gentle.
So the two modes are almost opposites. P-only is quick to respond and never overreacts, but leaves a standing offset. I-only removes the offset and moves smoothly, but is slow and has no fast reaction to sudden changes. Recognising these personalities is what lets an engineer match the mode to the job, rather than reflexively reaching for full PID everywhere.
The showcase application for both stripped-down modes is level control on vessels used as buffers, where the surprising goal is often not to hold level tightly but to let it wander. A surge drum or averaging tank exists precisely to absorb flow variations so that its outflow to the next unit stays smooth. If a tight controller held the level rigidly on setpoint, it would translate every inflow disturbance straight into an outflow disturbance, defeating the vessel's purpose. Averaging level control instead uses a gentle P-only or I-only strategy so that the outflow changes slowly while the level is allowed to swing within the tank, using the vessel's capacity as the shock absorber it was built to be.
For that job the personalities pay off directly. A P-only averaging controller is deliberately given a wide proportional band so the output changes only a little as level moves across much of the tank, and the offset that would horrify a temperature engineer is welcome here, because holding level exactly on setpoint was never the aim. An I-only averaging controller, with its slow ramping and absence of any sudden kick, likewise smooths the outflow and, because it has no proportional term, will not jerk the downstream flow when the level or setpoint moves sharply. Both convert an abrupt inflow upset into a gentle, spread-out outflow change.
P-only appears in other places too. Some simple or fast inner loops are run proportional only where a small offset is tolerable and the priority is speed and stability, and some protective and override elements are proportional only by design. I-only, being slow and kick-free, is chosen where smoothness matters more than speed and where any proportional reaction to noise or steps would be undesirable. The common thread is that these are conscious selections driven by what the loop is really for, not shortcuts.
The central tradeoff of P-only is offset for speed and simplicity. You get a fast, well-behaved loop that cannot wind up, but you must accept that it will sit off setpoint by an amount that grows as the load moves away from where you set the output bias, and tightening the proportional band to shrink that offset eventually makes the loop oscillate. If holding the exact target value matters, P-only is the wrong choice; if a predictable standing offset is acceptable or even desirable, it is often the right one. The offset here is a designed-in consequence of the mode, not a fault to be fixed.
The central tradeoff of I-only is speed for smoothness and zero offset. You get a loop that always returns to setpoint and never delivers a sudden output move, but it is sluggish, slow to correct a large deviation, and can be slow to settle because integral action alone tends toward a rolling, oscillatory approach if pushed too fast. Where a quick reaction to disturbances is needed, I-only is too slow; where gentle, offset-free motion is the whole point, it fits. Choosing between P-only and I-only for an averaging duty comes down to whether a small permanent offset or a slower approach to setpoint is the more acceptable price.
Because these choices are about behaviour rather than a fault, cloud SCADA trending is the natural place to confirm them. A P-only loop's steady offset is plainly visible as the process variable settling a consistent distance from setpoint, and an I-only loop's slow, kick-free ramping shows as a smooth output with no steps at setpoint changes. On buffer vessels spread across remote sites, being able to watch level swing within the tank while outflow changes gently is how an operator confirms the averaging strategy is doing its job. If instead a buffer level is being held rigidly and the downstream flow is jumpy, the trend reveals that the loop has been tuned or moded for tight control when averaging was wanted.
Because sometimes the offset is acceptable or even wanted. In averaging level control, holding the level exactly on setpoint would defeat the vessel's job of smoothing flow, so a standing offset is fine. P-only is also fast, simple, and cannot wind up, which suits some inner and protective loops. The offset is a designed tradeoff, not a defect, in these applications.
Integral-only control eliminates offset and returns the process to setpoint, and because it has no proportional term it produces no sudden output kick when the setpoint or measurement steps. Its motion is always smooth and gradual. This makes it useful where gentle, offset-free control is wanted, such as averaging duties, at the cost of being slow to respond to disturbances.
Both are used, and the choice depends on what you can tolerate. P-only with a wide proportional band gives smooth outflow but leaves the level offset from setpoint. I-only gives smooth, kick-free outflow and returns level to setpoint eventually but responds more slowly. The decision comes down to whether a small permanent offset or a slower approach to setpoint is more acceptable for the vessel.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.