External reset feedback, sometimes called back-calculation, is the quiet mechanism that keeps a controller honest when it is not the one actually in charge. Instead of building its integral term from its own internal output, the controller builds it from a signal fed back from downstream - the position a selector actually chose, or the setpoint a secondary loop actually accepted. When that feedback tells the controller its wishes were ignored, its integral simply stops accumulating, so there is nothing wound up to unwind later. This page explains external reset feedback as the general anti-windup principle behind clean selector and cascade behavior.
External Reset Feedback in one line: External reset feedback is a controller design in which the integral (reset) action is computed from an external feedback signal representing the actual downstream value, rather than from the controller's own output. When a loop is overridden or its output is not being followed, the feedback holds the integrator in check, preventing reset windup so the loop resumes control without a bump.
A conventional PID controller integrates its own error over time and feeds that accumulation back into itself to form the reset action. That works fine when the controller's output actually drives the process. The trouble starts when it does not - when a selector picks a different controller's output, when a downstream secondary refuses the setpoint, or when the valve is already hard against a stop. The controller keeps seeing error, keeps integrating, and its reset term marches off to a value the process never honored. That stored-up value is windup, and it shows up later as a slow, ugly overshoot when the loop finally regains authority.
External reset feedback breaks that trap by changing where the reset term comes from. Rather than feeding the integrator from the controller's own computed output, the design feeds it from an external signal that reports what really happened downstream. In many implementations this is done with a positive-feedback lag network: the proportional term is added to a filtered version of the feedback signal, and the filter time constant equals the integral time. When the feedback matches the output, the loop integrates normally. When the feedback diverges - because someone else is in control - the reset action tracks the feedback instead of running away.
The practical result is that the integrator can never accumulate a demand the process did not accept. The moment the controller's output is being followed again, its reset term is already sitting at the right value because it was tracking reality the whole time. There is no wound-up error to bleed off, which is exactly why external reset feedback gives smooth, bumpless handoffs where a naive controller would lurch.
The classic home for external reset feedback is the override or selector scheme, where two or more controllers compete for one valve and a high or low selector passes just one output through. The controllers that lose the selection must not wind up while they sit idle, or they will slam the valve when they eventually win. Wiring the selector's chosen output back to every controller's reset feedback input solves this cleanly: the losing controllers see that their output was not selected and simply track the winner, poised to take over the instant the process crosses into their territory.
Cascades use the same idea for the primary controller. In a cascade, the primary's output is a setpoint for the secondary, not a direct valve command. If the secondary is in manual, is saturated, or is itself overridden, the primary's demand is not being met - and a plain primary would wind up against that unmet demand. Feeding the secondary's working setpoint or its process value back to the primary as reset feedback keeps the primary's integral aligned with what the secondary can actually deliver, so closing or re-opening the cascade does not produce a bump.
Because external reset feedback is a general property of the controller rather than a bolt-on feature of any one scheme, the same block handles all of these cases with the same wiring philosophy: always feed the reset term from the truest available downstream signal. That is why modern DCS and PLC PID blocks expose a dedicated reset-feedback or track input. Once you understand it as back-calculation, selector windup, cascade windup, and output-saturation windup all become the same problem with one answer.
On remote oil and gas sites, override and cascade schemes are common precisely because the equipment is unattended and has to protect itself. A pump discharge loop might be overridden by a suction-pressure limiter; a separator level loop might hand off to a downstream flow controller. Every one of those handoffs is a candidate for windup if the losing or upstream controller is not using reset feedback, and a bad handoff on an unmanned site can mean an overpressure trip or a slug that nobody was there to catch.
When a cloud SCADA platform such as Merobix trends these loops, external reset feedback is not something the platform performs - it lives in the field controller - but the platform is where its absence becomes visible. A loop that overshoots hard every time it comes off a limit, or a secondary that bumps whenever the cascade closes, leaves a fingerprint in the historized output and process-value traces. Trending the selected output alongside each competing controller's demand is the fastest way to spot a controller that is winding up because its reset feedback was never wired.
Configuring reset feedback correctly also makes remote operation safer during mode changes. When an operator far from the site switches a loop between auto, manual, and cascade over the SCADA link, the reset-feedback path is what guarantees the transfer is bumpless. Because the reset term was tracking the real output all along, the loop picks up exactly where the field left it, with no surprise motion sent down a wire to a valve that may be hundreds of miles away.
Clamping simply freezes or limits the integrator when the output hits a bound, which stops windup but is a blunt on/off action. External reset feedback instead makes the reset term continuously track a real downstream signal, so the integrator is always aligned with what the process actually accepted, not just held. The feedback approach handles cascades and selectors that clamping alone cannot, because the constraint is not always the controller's own output limit.
Because the controller effectively works backward from the true downstream value to figure out what its integral term should be, rather than forward from its own output. Given what actually happened, it back-calculates a reset value consistent with that reality. The two names describe the same mechanism from different angles - one names the signal path, the other names the math.
A plain single loop with a valve that follows the output faithfully gets little benefit, since the output and the true action already agree. It becomes important the moment the output might not be honored - in overrides, selectors, cascades, or when the final element can saturate or be put in manual. If a loop ever shares a valve or feeds a setpoint to another loop, reset feedback is worth wiring.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.