How to Settle a Hunting Loop Without Guessing

Why this matters

Most loop tuning in the field is done from memory: change something, stand there, decide it looks better, change something else. It sometimes works, and it is never transferable, because nobody can say afterwards which change did what or what the loop looked like before anyone touched it. The result is a system carrying three or four undocumented adjustments, at least one of which is compensating for a defect that has since been repaired. A single sheet of paper with four columns removes all of that, and it also turns tuning from an opinion into a measurement.

Before you disturb a running loop

• Leave every limit, interlock and protective device in service for the entire exercise. If one opens while you are working, that is the finding: establish why it opened before anything else. Widening a limit, jumpering it, or replacing a correctly-operating one to keep a test running all reach the same end state, and the last one just costs a part on the way.
• Actuator and linkage work happens with the actuator isolated. A spring-return unit stores energy and will slam when released. Isolate it, release or restrain the spring, and lock or tag under 29 CFR 1910.147 before a coupling, set screw or pry bar goes anywhere near the travel path.
• A hot or pressurized line is a burn and scald route. Take temperatures at existing wells, let surfaces cool or use insulated gloves and eye protection, and if a line must be opened, isolate it, drain to a safe point, and confirm zero on a gauge before a thread moves.
• Where a setting or a measurement requires an energized control enclosure, 29 CFR 1910.333(a)(1) permits energized work only where de-energizing introduces additional or increased hazards or is infeasible due to equipment design or operational limitations; use a meter and leads rated CAT III at or above the circuit voltage and work to the boundaries and protective equipment NFPA 70E-2021 assigns. Where the panel can be dead, open the disconnecting means, lock and tag under 29 CFR 1910.333(b)(2) in general industry or 29 CFR 1926.417 in construction, and prove dead per NFPA 70E-2021, 120.5.
• On combustion equipment, a personal carbon monoxide monitor goes on your body before it fires and stays on you for the whole visit. Tuning changes cycle length, and startup is the least settled part of any combustion cycle.

The sheet

Six columns. Nothing else, and none of them optional.

The period and amplitude columns are the working pair. The period tells you where the delay lives, and the amplitude tells you whether the lever you just pulled was the right one.

The order rows go in, and why it is not negotiable

Rows one through three are defect corrections: things that stay fixed no matter how the tuning ends up. Rows four onward are settings: trades, where you give up one property to buy another.

1. Baseline. Change nothing. Without this row every later row is compared against a memory.
2. Measurement integrity. Terminations, sensor seating and heat transfer compound, the input filter or smoothing parameter. This comes before everything because every subsequent row is scored against a trace, and a trace that is smoothing or adding noise scores every one of them wrong.
3. Mechanical: stroke, linkage, positioner. A loop cannot be tuned to deliver an output the actuator does not actually produce. Slop makes every gain number meaningless, and it makes the meaningless number look like it worked at one load.
4. Timing protections: minimum on and off times, deadband. Before gain, because these are bounded by what the equipment documentation requires, and you need to know how much room you have before you start spending it.
5. Gain: throttling range or proportional band. The first genuine trade. Wider range means less swing and more standing offset.
6. Reset or integral time. Last, because it removes offset and it is also the parameter most able to reintroduce oscillation on a slow loop.

Tuning before repairing is the specific mistake this order prevents. Tune around a slipped linkage and the settings become wrong the moment somebody tightens it, which will be a different tech on a different day who has no idea a compensation exists.

How long to wait before writing a row

At least three full periods after the change, and at least one real load change. Three periods because a loop can look calm for one cycle by luck. A load change because a loop that is stable at one load can be unstable at another: gain through a valve and through a coil is rarely constant across the range, so the loop's effective gain moves with load even though the setting did not.

If the process is too slow to see three periods in the time available, write the row anyway and mark it provisional, then check it on the next visit. A provisional row is worth far more than no row.

The filled-in sheet

A modulating heating loop, complaint of continuous swing that the customer describes as waves of hot and cool.

Row 1 was a defect, not a cause. Amplitude went from 4.0 to 3.8 F, a 5% change, which is inside the noise of the measurement. The loose lead was a real fault and it stays fixed, but it was not driving the swing, and the sheet says so plainly. Without the baseline row, a tech who found a loose lead would have claimed it.

Row 2 was the biggest single win and it was mechanical. Amplitude went from 3.8 to 2.4 F, about 37%. The actuator had been travelling its full stroke while the valve stem lost roughly 12% of travel to a loose coupling, so the controller had to overdrive in both directions to get any effect and then overshot when the slop took up. Note that the output swing narrowed at the same time, from 50 points of range to 40, which is the corroborating evidence: the loop needed less output travel to do the same work.

Row 3 was the only genuine tuning change in the whole visit. Widening the throttling range from 5 F to 10 F halves the gain, and amplitude went from 2.4 to 0.9 F, a 62.5% reduction. That is more than proportional, and it should be: near the stability boundary amplitude is far more sensitive to gain than one-for-one, in both directions, so do not expect halving the gain to halve the swing and do not read a large improvement as proof you found the root cause.

Read the period column on its own. It sat at 12.0 minutes at baseline, 12.0 after the lead, and 12.0 after the linkage, then moved to 13.0 after the range change, an 8% shift on the last row only. A period that holds through three readings while amplitude falls by just under 78% overall, from 4.0 F to 0.9 F, is the signature of a real process delay that none of the work touched. Nothing done on this visit made the system faster; the work made it stop overshooting a delay that is still there.

What that costs the customer, stated honestly. The loop is now stable and it is also slower to recover from a load step than it was, because the range is twice as wide and the standing offset is correspondingly larger at any given output. If recovery speed matters on this system, the remaining work is to find and shorten the 12-minute delay, and that is a separate scope.

What would have changed the approach. If the period had moved substantially on any row, the oscillation was not a linear tuning oscillation and the sheet would have said so immediately. If amplitude had refused to fall at row 3, the gain was not the lever and reset time would have been the next row, not another gain change.

The failure mode. Doing rows 2 and 3 in the same trip to the panel. Both were real improvements, the loop settles, and the sheet records one row instead of two. Six months later, when somebody asks whether the range can be narrowed to get speed back, nobody can say whether the stability came from the linkage repair or the range, so the only safe answer is no.

What a completed sheet tells you that a settled loop does not

A flat period column with a falling amplitude column means the delay is untouched. The loop is stable because it has been slowed to accommodate a round trip nobody shortened. That is a legitimate finish, and it is also a quotable second phase.

A period that changes when a mechanical repair is made means the repair changed the process itself, most often by restoring stroke that changes how fast the actuator can move the process. Those rows are worth more than tuning rows because they improved capability rather than trading it.

An output swing that pins at 0% or 100% on any row is saturation, and every tuning conclusion drawn from that row is void. Redo it at a load where the output stays inside its range, and note the load at which saturation began, because that number is the real capacity finding hiding inside a tuning complaint.

Two rows with identical parameters and different results are a load effect, not an inconsistency. Compare them, and if the loop is stable at one load and hunting at another, the setting has to be chosen for the worse of the two and stated as such on the ticket.

References

• 29 CFR 1910.147, OSHA control of hazardous energy, for actuator spring tension and mechanical isolation
• 29 CFR 1910.333(a)(1) and (b)(2), OSHA general industry work practices for energized and de-energized electrical work
• 29 CFR 1926.417, OSHA construction lockout and tagging of circuits
• NFPA 70E-2021, 120.5, verifying an electrically safe work condition
• See related: Why a Control Loop Hunts; Proportional Response in Plain Terms