What it is

This is the call nobody wants: the customer swears the AC quit twice yesterday afternoon, but when you arrive it's cooling like a champ and every reading is textbook. You can't fix what you can't see fail, and a parts-changer's instinct here is to either guess-and-replace something or write "no problem found" and leave. Both get you a callback.

An intermittent fault is still a fault. It's just one that needs a trigger you weren't there for — usually heat, time, vibration, or a specific load condition. Your job changes from "diagnose the failure" to "recreate the trigger or capture the fault while it happens." That's a different skill, and it's the one that separates a real diagnostician from somebody who only fixes things that are conveniently broken on arrival.

How it works

Intermittents almost always trace to one of a handful of physics-driven triggers. Understanding the trigger tells you where to look:

  • Heat soak / thermal. A marginal part works cold and fails once it's been running an hour and the cabinet, the attic, or the component itself heats up. Weak capacitors, marginal compressors approaching their overload trip point, control boards with cold solder joints, and contactor coils all do this. This is the single most common reason "it works when you arrive" — you cooled it off by opening it up, and the trip point reset.
  • Connection / mechanical. A loose lug, a corroded spade, a backed-out wire nut, or a contactor with pitted contacts conducts fine until thermal expansion, vibration, or a high-current event opens the path. It's a coin flip every cycle.
  • Load-dependent. The system only fails on the hottest part of the day, at peak head pressure, or when a second stage kicks in. The fault needs that operating point to express itself — a high-pressure switch that trips only at 115°F outdoor, a TXV that hunts only at high load.
  • Time-of-day / utility. Low incoming voltage during a utility brownout at 4 PM drops the compressor below the voltage it needs to run, and it trips on overload. Comes back when the grid recovers. Your 9 AM visit sees perfect 244V.

In the field

You generally can't out-stubborn an intermittent by staring at it. Work the trigger:

  1. Interview like a detective. When exactly does it quit — time of day, after how long running, indoor or outdoor temp, any noise or smell? Does it come back on its own after a while (thermal/overload reset) or only after a power cycle (a latching fault)? "Comes back by itself in 20 minutes" screams thermal overload. "Have to flip the breaker" points at a latching safety or a board.
  1. Try to force the trigger. If they say it dies after an hour in the afternoon heat, then run it for an hour and block some condenser airflow or close it up the way the install sits. Push it toward the operating point where it fails instead of testing it in the easy condition.
  1. Heat-test suspect components. A heat gun (gently) on a suspected cold-solder board, a capacitor, or a relay can provoke a thermal fault on the bench-side of the visit. Watch for the symptom to appear as the part warms.
  1. Wiggle-test the connections. With the system running (and respecting live-panel safety), gently flex the wiring at lugs, the contactor, the capacitor spades, and the disconnect. If the unit hiccups when you move a wire, you found it. Pay attention to the disconnect and the whip — outdoor connections corrode.
  1. Leave a recorder when you can't reproduce it. This is the move that wins. Drop a recording clamp/voltage logger on the compressor common, or use the equipment's own monitoring if it has it. Let it run a day. When you come back, the log shows you the exact moment it tripped and what the current/voltage did at that instant — overload trip, voltage sag, or a hard open. On systems with onboard fault memory or a connected monitor, pull the history before you do anything else.
  1. Check the cheap chronic stuff while you're there. Even if you can't catch the fault, a weak run capacitor that's testing low-but-passing, a contactor with light pitting, low refrigerant, or a dirty condenser all predispose a system to intermittent trips. Bringing those back to spec often kills the intermittent even if you never see it fail.

Normal values & targets

You're hunting deviations under stress, not absolute numbers — but a few "marginal but passing" thresholds to flag:

  • Run capacitor: if it measures more than about 6% under its rated µF, it's weak. A cap that reads in-spec cold but a hair low is a prime heat-soak suspect — don't trust a single cold reading.
  • Compressor current: sustained draw creeping toward nameplate RLA as the unit heats up means it's marching toward an overload trip. Note whether amps rise over time during a long run.
  • Voltage at the contactor, under load: should hold near nominal (e.g. ~240V single-phase) while running. A sag to the low-220s or below under load points at a connection or utility problem. Voltage that's fine at no-load but drops when the compressor pulls in is a connection telling on itself.
  • Voltage imbalance / drop across a connection: measurable voltage drop across a closed contactor or a lug while current flows means resistance where there shouldn't be any. More than a volt or two across a "closed" connection is a problem.

Common faults & what they mean

  • Quits after ~30–60 min, restarts on its own: compressor or motor tripping on internal thermal overload. Find why it's overheating — low charge, high head, weak cap, low voltage, dirty condenser — don't just wait for it to "do it again."
  • Quits at the hottest part of the day only: high-pressure trip (condenser can't reject heat at peak ambient) or a voltage sag during peak grid load. Check head pressure at peak conditions and log incoming voltage.
  • Quits randomly, returns when you wiggle a wire: loose/corroded connection or a pitted contactor. Mechanical, not refrigerant.
  • Dead until breaker reset, then fine for days: a latching safety tripped (float, high-pressure lockout) or a board fault. Find what tripped it; the safety is reporting a real event you missed.
  • Only fails when a second stage or the heat strips energize: load- or sequencing-related — a marginal transformer, an overloaded circuit, or a staging control. Look at what's different in that operating mode.

Tech tips & gotchas

  • Opening the unit IS the reset. The act of pulling the disconnect, removing the panel, and letting it sit cools the very part that was tripping. By the time your gauges are on, the trigger is gone. Account for that — recreate the running, closed-up, heat-soaked condition.
  • "No problem found" is a diagnosis you have to earn. It's legitimate only after you've tried to force the trigger and left monitoring. Writing it because the unit happened to behave is how you lose a customer.
  • One intermittent can have multiple contributing factors. A weak cap and a dirty condenser together cross the trip threshold on a hot day; either one alone wouldn't. Fixing both is sometimes the real repair.
  • Believe the customer's pattern, not the moment. If they describe a clean thermal signature ("an hour, then 20 minutes off, all afternoon"), trust that over your one perfect snapshot. Their data covers days; yours covers minutes.
  • Document the marginal findings. Even if you can't reproduce it, recording "cap 4% low, contactor lightly pitted, voltage sags to 228V under load" protects you and gives the next tech (or the next visit) a head start.

Safety / code notes

  • Wiggle-testing and heat-testing happen in a live panel — use properly rated meters and PPE, keep one hand clear, and bleed capacitors before touching motor terminals.
  • A safety that trips intermittently (high-pressure switch, float, limit) is protecting against a real condition. Never jumper it out to "see if that's the problem" and leave it that way — diagnose what it's guarding against.
  • If you leave a logging device, make sure it's secured and rated for the location (outdoor units need weather-appropriate placement) and doesn't compromise a panel cover or strain relief.