RTU Short Cycling: Why Your Rooftop Unit Starts and Stops Constantly

Stand in the parking lot and listen. A healthy rooftop unit runs in long, steady stretches. A short-cycling one kicks on, runs a few minutes, shuts off, and does it again, dozens of times an hour, all day. From inside the restaurant you might only notice rooms that never quite settle at temperature. On the roof, the unit is dying in installments.

What's actually happening

Compressors are built for sustained running, not for starting. Every start is a stress spike: high current draw, oil that hasn't circulated yet, refrigerant pressures slamming around. A unit that should start a handful of times an hour and run long cycles is instead starting hundreds of times a day. Nothing fails today. Everything fails sooner.

The likely causes, roughly in order

Control and thermostat issues. A thermostat mounted where a draft or heat source hits it, a mis-set anticipator or dead band too narrow, or a building controller with conflicting setpoints. The unit is fine, the instructions it gets are not. Common and cheap to fix.

Safety switches tripping and resetting. High-pressure or low-pressure cutouts opening mid-cycle, then closing after pressures settle, so the unit restarts and repeats. The cutout is doing its job; something behind it (dirty condenser coil, low charge, failing fan motor) is the real story. This one is the classic "it still works, so we ignored it" path to a dead compressor.

Low refrigerant charge. The low-pressure switch trips, pressures equalize, it resets. Runs shorter and shorter cycles as the leak progresses.

Dirty condenser coil or failing condenser fan. Head pressure climbs until the high-pressure switch says no. On a restaurant roof, where the condenser breathes kitchen exhaust grease, coils foul faster than anyone expects.

Oversized unit. If the unit was sized generously (or the space use changed), it satisfies the thermostat in minutes and shuts off, over and over. No component is broken, which is why nobody ever fixes it. It still wears the machine and it never dehumidifies properly.

What it costs if you wait

Short cycling is the highest-wear duty a compressor can run, and compressors are the expensive organ: a commercial RTU compressor replacement typically runs into thousands, and if the unit is old enough, the quote becomes a full unit replacement measured in five figures. Our guide to the signs a compressor is failing covers how that decline progresses. The secondary bill is comfort: a short-cycling unit controls temperature badly and humidity worse, and in a dining room that reads as "it's always either cold or stuffy in here."

The pattern in industry service data is consistent: most compressor failures had a long observable run-up, and cycling behavior is one of the most visible parts of it. This is rarely a lightning strike. It's a slow-motion event with a paper trail, if anything is recording.

What the data shows before it becomes a failure

Cycling is one of the easiest failure precursors to see in data, because it's arithmetic, not judgment:

• Starts per hour climb. A monitored unit shows its cycling rate drifting up over weeks. No human notices three extra starts an hour. A counter does.
• Cycle length shortens. Average run duration trends down before any cutout starts tripping.
• Current draw at start-up changes. Compressor start current rising over its own baseline is an early flag for the hard-starting that accompanies this pattern.
• Supply air temperature stops holding. The gap between return and supply air narrows during runs. The unit runs, but does less each time.

The point of watching these isn't elegance. It's that every one of them moves weeks before the Friday-afternoon no-cool call, when the fix is a coil clean or a control adjustment instead of a compressor.

What to do now

Count it: ten minutes near the unit (or a plug-in runtime logger) tells you if you have a cycling problem. If the unit starts more than roughly 6 to 8 times an hour under steady conditions, get it looked at. Ask the tech specifically what tripped, thermostat logic, a safety, or charge, because "it's running now" is not a diagnosis.

If you'd rather the counting happened automatically, that's what monitoring does, and what we're building at SmartHVAC. Get the failure library as it grows.