Why Short Cycling Protection Looks Like a Fault

Why this matters

"It won't come back on for five minutes" is one of the most common complaints in field service and one of the most reliably misdiagnosed, because the thing the customer is describing is the equipment protecting itself, working exactly as designed, on a machine that has a real fault somewhere else entirely.

The claim: an anti-short-cycle timer converts a fast fault into a slow one, so the complaint it produces is a report of the protection working, and the length of the delay is diagnostic data rather than the problem. What makes this article's method usable is that it works on a machine that is running perfectly when you arrive, which is the state you will almost always find it in.

Before anything else: which device is cutting it out

Short cycling and repeated protective operation produce the same customer sentence and are not the same thing. A machine whose operating control satisfies early and re-demands is cycling. A machine whose LIMIT, pressure switch or interlock opens every cycle is not cycling, it is being stopped, over and over, by something reporting a condition that should not exist.

Establish which device opens at cut-out before you go further. If it is a protective device, the question is why that device sees the condition it sees, and treating the device as the fault by replacing it or adjusting it lands in the same place as jumpering it, one step slower. The sibling article on the difference between a limit and a control carries the separation stack that tells you which term moved.

And whichever it turns out to be: do not bypass or shorten the minimum-off timer to "see what it does". Defeating a protection to reproduce a symptom is not a test, and the reproduction it gives you is of a condition the machine was built to prevent.

Where reading a controller means opening an enclosure, that is 29 CFR 1910.333(b)(2) for de-energizing and lockout with live-dead-live verification per NFPA 70E-2021, 120.5. Readings that only exist with the circuit live fall under 29 CFR 1910.333(a)(1), permitted where de-energizing would introduce additional or increased hazards or is infeasible for the equipment design, with a meter rated at or above the system voltage and leads inspected for damage first per 29 CFR 1910.334(c).

What the timer actually does to the evidence

A minimum-off timer holds the machine off for a fixed interval after every stop, regardless of demand. Consider what that does to a machine whose real fault is a run period far shorter than it should be.

Without the timer, the machine would stop and restart almost immediately, producing a very high cycle rate that is obvious to anyone standing near it and destructive to contacts and driven components. With the timer, each of those cycles gets a fixed floor of dead time inserted after it. The cycle rate drops, the destruction slows, and the symptom transforms from "it's banging on and off constantly" into "it takes a while to come back", which sounds like a completely different and much less urgent complaint.

That is a good trade for the equipment and a bad one for the diagnosis, because the loud, obvious symptom has been traded for a quiet one that gets described as slowness.

The counters carry the case

You do not need to witness the cycling. The machine has been counting.

Two figures, read from the controller and recorded with the date: a cycle or start counter, and a run-hour counter. Divide run hours by cycles and you have the average run length per cycle, which is the single most useful number in this whole diagnosis and one almost nobody reads.

Two conditions on it. It is an AVERAGE over whatever period separates your two readings, so a machine that ran normally for three weeks and badly for one will report something in between rather than the bad figure. And both counters must come from the same period, which means either two readings separated by a known interval or a controller that reports both since a common reset.

The one observation that says who governs the off period

Here is the observation that turns the reconstruction into a diagnosis, and it takes one stopwatch and one cycle:

Time the gap between the machine stopping and the machine restarting, and compare it against the published minimum off time.

Take that timing from outside the swept path of every rotating or driven part, with all guards in place and the enclosure closed, and clear of the flue and relief-opening path with a personal CO monitor on your collar if the machine burns fuel. The whole point of the observation is that the machine restarts on its own, with no warning and nobody's hand on a switch, so treat the entire off period as a machine about to start.

If the gap is longer than the timer, demand went away and came back on its own schedule. The timer expired at some point during the gap and nothing happened, because nothing was asking. That machine's cycling is set by the process.

If the gap equals the timer, to within a second or two, and does so every time, then demand was present throughout the off period and the machine restarted the instant it was permitted. The off period is entirely timer-imposed. That machine is asking to run continuously and being held off, and its run length is the fault.

Those two readings look identical to a customer and mean opposite things.

The reconstruction, worked

A machine whose published minimum off time is 300 seconds, which is 5 minutes. Complaint: a five-minute wait before it comes back. Machine running normally on arrival.

Counters read against a record from a service visit 30 days earlier:

Average run per cycle: 54 hours / 1,260 cycles = 0.0429 hours, which is 2.57 minutes. Call it about 2.6 minutes.

Cycles per day: 1,260 / 30 = 42.

Run time per day: 54 / 30 = 1.8 hours. Check that against the other two figures, because a cross-check catches a mis-read counter: 42 cycles at 2.571 minutes each is 108 minutes, which is 1.8 hours. Consistent.

For this equipment and this load a healthy run period sits in the range of 10 to 15 minutes, so 2.6 minutes is roughly a fifth to a quarter of what it should be. That is the finding, and it has nothing to do with the five minutes the customer described.

Now the stopwatch observation. The machine stopped, and restarted 5 minutes and 2 seconds later. Repeated twice more: same result within a second or two each time.

So the off period is governed entirely by the timer. Demand never went away. The machine is being asked for continuous output, delivering 2.6 minutes of it, stopping for reasons of its own, and then being held out for 5 minutes by a protection doing precisely its job.

The customer's complaint describes the 5 minutes. The fault is in the 2.6.

What the timer's length tells you on its own

Match the observed gap against the published figure before you conclude anything, because a mismatch is informative in both directions.

A gap consistently SHORTER than the published minimum off time means either the setting has been changed from published, or the protection is not the device you think it is. Both are worth knowing before you diagnose anything else, and the first one is a finding in itself.

A gap consistently LONGER than the published minimum, on a machine with continuous demand, points at a second delay you have not accounted for, such as an equalization permissive or a staged restart, layered on top of the timer.

Confirming the repair with the counters, not with a stopwatch

The cause on this machine turned out to be a control sensing location that satisfied early on a local condition rather than on the load it was supposed to represent, which the sibling article on where a sensor reports from covers in full.

The confirmation is the part worth copying, because standing beside the machine for one good cycle proves nothing on a fault that only shows up over hours. Read the counters again after seven days:

Average run per cycle: 12.6 x 60 / 62 = 12.19 minutes, inside the 10 to 15 minute expectation. Cycles per day: 62 / 7 = 8.9, down from 42, a reduction of about 79 percent.

And the figure that confirms the diagnosis rather than just the improvement: run time per day is 12.6 / 7 = 1.8 hours, the same as before the repair. The machine is doing the same amount of work per day. It is simply doing it in 8.9 cycles instead of 42, which is exactly what a cycling fault looks like when it is corrected, and it is not what a capacity fault looks like when it is corrected. Had the daily run hours moved substantially, the repair would have changed the machine's output, and that would mean the original diagnosis was incomplete.

The failure mode this avoids. The common wrong ending is to accept the customer's framing, find the 5-minute delay in the documentation, explain that it is normal, and leave. That explanation is true. The machine still cycles 42 times a day, the contacts and the driven components still absorb it, and the next call is a failed component that reads as unrelated. Being right about the timer is not the same as being right about the machine.

What would change the answer. If the stopwatch gap had come back at 8 minutes against a 5-minute published timer, demand was not continuous and the run length figure, while still low, is being driven by something in the load rather than in the control. If the cut-out device had turned out to be a limit rather than the operating control, the entire counter analysis still applies but the question becomes why that limit reaches its setpoint in under three minutes, which is a much more urgent conversation. And if the two counter readings straddle a season change, the 10 to 15 minute expectation itself may not hold for the earlier part of the window, so take the shortest period that gives you a usable cycle count rather than the longest one available.

References

• 29 CFR 1910.333(a)(1) for the conditions permitting energized measurement, 1910.333(b)(2) for de-energizing and lockout of electrical circuits, and 1910.334(c) for test instrument and lead inspection
• NFPA 70E-2021, 120.5 for live-dead-live verification
• Manufacturer service documentation for the published minimum off time, expected run period for the equipment and load, and counter reset behaviour
• See related: What a Time Delay Is Protecting; The Difference Between a Limit and a Control; Where a Sensor Is Reporting From and Why It Matters