Direct and reverse controller action

In short. The controller must move its output in the direction that reduces the error. The direction is set by the sign of the process gain K: if raising the output raises PV (heating), you need reverse action; if it lowers PV (cooling), direct action. The wrong direction turns negative feedback into positive feedback and the loop runs into a limit.

Terms

• Process gain K = ΔPV/Δu — with its sign. Steam heating: +1.8 °C per +1 % of opening → K = +1.8 °C/%.
• Reverse action: when PV rises the output decreases. The standard choice for K > 0.
• Direct action: when PV rises the output increases. The choice for K < 0.

With the section convention e = SP − PV:

reverse action  ⇔  Kc > 0        (needed for K > 0)
direct action   ⇔  Kc < 0        (needed for K < 0)
negative feedback condition:  Kc · K > 0

All tuning rules in the application (IMC, SIMC) return Kc with the sign of 1/K, so the action follows automatically: for K = −1.8 SIMC gives Kc = −1.04.

Examples

The last row is a common trap: the sign depends not only on the process but on the actuator (fail-open/fail-close), a reversed drive, an inversion in the analog output. All of this is part of K "as the controller sees it".

How to check the sign before auto

1. In manual change the output by a small amount (e.g. +5 %).
2. Wait for PV to react, allowing for the dead time.
3. If PV rose, K > 0: choose reverse action (Kc > 0 with e = SP − PV). If it fell: direct.
4. The same test gives a model estimate: step test.

What a sign error looks like

With the wrong sign the controller amplifies the deviation: PV moves away from the set point, the output hits a limit and stays there. This is not "bad tuning"; reducing Kc does not help.

In the simulator: process K = −1.8, τ = 45 s, θ = 12 s.

• Kc = −1.04, Ti = 45 s — stable, PV reaches the set point;
• Kc = +1.04, Ti = 45 s — the simulator stops the run: PV grows without bound.

How a PLC sets it

PID blocks use one of three ways, sometimes together:

• a separate "direct/reverse action" (or "invert") parameter;
• the sign of the gain (a negative Kc is not accepted everywhere);
• the choice of the error formula (SP − PV or PV − SP).

If a block computes the error as PV − SP, the sign of Kc flips for the same process. Before entering settings find out from the documentation of the particular block which convention it uses.

Typical mistakes

• Double inversion: the sign is taken into account both in the "action" parameter and in the sign of Kc — the direction is wrong again.
• Replacing a fail-close valve with a fail-open one without changing the action.
• Copying settings between similar loops (inlet/outlet) without checking the sign.

Limits

The sign of K can change between operating points (for instance on processes with an optimum), which needs another control structure. For ordinary processes the sign is constant, but it must be determined for what the controller sees, including actuator and scaling.