What a Control Loop Actually Is

Why this matters

Half the calls that get written up as "control problem" are not loop problems at all, because the function that misbehaved was never inside a loop. Techs spend an afternoon adjusting settings on a controller that was doing exactly what it was told, while the actual culprit is a timer, a permissive, or a limit that has no idea what the process is doing. Getting the boundary right on arrival is worth more than any tuning skill, because it decides whether you spend the visit measuring or guessing.

Before you open anything to look at a loop

Most of what you need is readable from the controller's own display, its status output, and the visible position of the final device, with nothing open. Take those readings first.

Where you must measure inside an energized control enclosure, 29 CFR 1910.333(a)(1) allows energized work only where de-energizing would introduce additional or increased hazards or is infeasible because of equipment design or operational limitations, and live troubleshooting is one of the recognized cases; that permission covers the measurement itself, not the parts of the job you could have done dead. Use a meter and leads rated CAT III at or above the circuit voltage, work to the arc-flash and shock boundaries and protective equipment the task requires under NFPA 70E-2021, and keep your free hand off the enclosure.

Where the check can be made dead, open the disconnecting means and apply your own lock and tag - 29 CFR 1910.333(b)(2) in general industry, 29 CFR 1926.417 in construction - then prove dead by NFPA 70E-2021, 120.5: test the meter on a known live source, test the conductors, re-test the meter. If the work means opening a machine rather than a panel, isolate and lock or tag under 29 CFR 1910.147, discharge capacitors, relieve pressure to zero on a gauge, and block or release spring tension before a linkage comes loose.

The circle that has to close

A control loop is three things wired into a circle: a measurement of some physical quantity, a decision made by comparing that measurement against a target, and an action that changes the quantity being measured. The third part is the test. If the action does not eventually show up at the measurement, you do not have a loop, whatever the drawing calls it.

That circle is why loops have behaviour that individual components do not. A loop can be stable or unstable, fast or slow, accurate or offset, and none of those words describes a sensor or a valve on its own. They describe the round trip. It is also why a loop can be built entirely out of healthy parts and still perform badly: the fault lives in the relationship, not in any one box.

The circle closes through the process, not through the panel. The wire from the controller to the actuator is not the feedback path. The feedback path is the water, the air, the pressure, the product, whatever is actually carrying the effect back to the sensor. Anything that changes that physical path - a fouled coil, a bypass left open, a sensor relocated during a remodel - changes the loop even though nothing electrical was touched.

What is in the panel that is not in the loop

This is the useful list, because these are the parts that get blamed for loop behaviour and cannot produce it.

The distinguishing question for anything in that table is simple: does its output change based on what it just caused? A timer that runs 90 seconds runs 90 seconds whether the tank filled or not. A limit that opens at its trip point does not open a little bit at a value slightly under it.

Open-loop functions fail in a shape closed-loop functions cannot

A closed loop degrades. It hunts, it offsets, it responds slowly, it overshoots and recovers. There is an error signal in the middle of all of that, and the loop is visibly working against it.

An open-loop function does not degrade, it is simply wrong by the same amount every cycle. That is the field signature worth carrying: repeatability is evidence of open loop. If the overshoot is 8 degrees on Monday, 8 on Tuesday and 8 on the coldest day of the year, no feedback is acting on it, because feedback would have produced different errors under different loads. A complaint that is identical every single cycle is pointing at something running on time, count, or position rather than on measurement.

Interlocks and limits deserve their own line here, because they are the ones people defeat. A limit or an interlock exists to stop the machine, not to control it, and it is not part of the loop's job to be satisfied. When one of them is open, the question is never how to get past it. The question is what condition put it there. Jumpering a correctly-operating limit and replacing a correctly-operating limit reach the same end state, one of them slower and with a part number on the invoice; both leave the customer exposed to the condition the device caught.

Worked example: the loop that was blamed for a timer

A heated storage vessel holds product at a target of 140 F. It has a modulating heat input at the bottom, a circulating pump, a temperature sensor near the top, a high limit, a flow interlock, and a make-up valve driven by a fixed-duration timer. The complaint is that after every make-up event the temperature overshoots badly, and two previous visits adjusted controller settings without improving it.

Start with the make-up event, because it is the thing that repeats. The timer admits a volume equal to about 25% of the vessel at a supply temperature of roughly 60 F. If that charge mixed completely, the vessel would land at 0.75 x 140 plus 0.25 x 60, which is 105 plus 15, or 120 F: a 20 F step drop presented to the loop in the time it takes to open a valve.

Two things about that 120 F figure. It is a fully-mixed calculation, and this vessel does not fully mix - cold make-up enters low and the sensor is high, so the sensor sees less than the full 20 F at first and more of it later, as the pump works the charge through. Treat 120 F as the bound the loop will eventually be asked to recover from, not as a temperature you expect to read. Second, it assumes the two streams are the same fluid with the same specific heat, which is true here and would not be if make-up were a different product.

Now the diagnosis writes itself. The step is created by an open-loop device. The timer runs the same duration regardless of level or temperature, which is why the disturbance is the same size on every cycle. The loop is not producing the overshoot; it is recovering from a disturbance nobody told it about, with its sensor at the far end of a stratified vessel, so it sees the recovery late and drives past it. The last two visits tuned the recovery and left the disturbance in place.

The fix is at the open-loop function, in this order: make the make-up quantity smaller and more frequent so the step is smaller, or gate it on level rather than on time, so the amount admitted reflects the amount actually needed. If neither is available on that equipment, the honest recommendation is that the loop cannot do better than it is doing, and the customer is choosing between a control upgrade and living with the swing.

What would flip this. If the overshoot varied with load or season, the repeatability argument collapses and the timer stops being the prime suspect. If the sensor sat low, near the heat input, the loop would see the disturbance immediately and the overshoot would be a tuning story after all.

The failure mode. Tuning a loop to absorb a repeating open-loop disturbance means detuning it for everything else. You slow the loop until it rides through the make-up event, and now it also rides slowly through every real load change the rest of the day, which shows up months later as a separate, vaguer complaint that nobody connects back to this visit.

Confirming you drew the boundary correctly

Two checks, both cheap, and neither is a settings change.

Break the circle on purpose at the measurement, not at the safety devices. Put the controller in hand at a fixed output and watch whether the measured value moves and settles. If it does, the action reaches the measurement and the circle is real. If the measured value does not respond at all to a large commanded change, the loop was already open somewhere - a stuck valve, a closed isolation valve, a sensor reading a place the action never reaches. Do not use a limit or an interlock as your test point; leave every protective device in service while you do this.

Log one full cycle of the complaint with the time of every event. Write down when the open-loop functions fire and when the measured value moves. An open-loop function that fires at a fixed time will line up with the disturbance to the second, every cycle, and that alignment is the evidence. A loop-driven excursion will not hold that alignment across two different load conditions.

References

• 29 CFR 1910.333(a)(1) and (b)(2), OSHA general industry selection and use of work practices for electrical work
• 29 CFR 1926.417, OSHA construction lockout and tagging of circuits
• 29 CFR 1910.147, OSHA control of hazardous energy for mechanical isolation and stored energy
• NFPA 70E-2021, 120.5, process for establishing and verifying an electrically safe work condition
• See related: The Sequence of Operation as a Diagnostic Instrument; The Danger of Just Resetting a Safety Device Without Diagnosing Why