Aircon compressor lockout: why it refuses to restart
A unit that locks itself out is refusing to run, not failing to run. Something measured a condition it was built to avoid. Cutting the power changes the count of those refusals, and leaves the condition behind them untouched.
By Team Snowflake | Updated 6 Aug 2026
What the protection layer is watching
Protection on a split system is not one device. Several independent watchers sit around the circuit, each holding a single measurement and a single limit. Any one of them can end a run on its own, and none of them consult the others before doing it.
Each watcher covers a different route to a damaged compressor. Limits on either end of the refrigerant circuit cover heat that will not leave and refrigerant that is not arriving. A current or temperature device on the motor covers a machine labouring past what its windings tolerate. Supply voltage gets watched on much of the equipment sold here too.
One difference between the older and newer hardware matters later on. A mechanical switch has no memory at all. It opens, it closes again once the condition passes, and nothing anywhere records that it happened. A board reading a sensor can store what it saw, tally how often it saw it, and behave differently on the fourth occasion than it did on the first.
So protection describes a decision rather than a component. Nothing on that list is a part that broke. Each entry is a boundary that got reached, by something else, elsewhere in the machine.
| What gets watched | What pushes it past the boundary | Where the answer usually sits |
|---|---|---|
| What gets watchedPressure on the hot side of the circuit | What pushes it past the boundaryHeat failing to leave the outdoor coil | Where the answer usually sitsThe air path around the outdoor unit |
| What gets watchedPressure on the cold side of the circuit | What pushes it past the boundaryRefrigerant not arriving in the amount designed | Where the answer usually sitsInside the sealed pipework |
| What gets watchedCurrent drawn by the compressor motor | What pushes it past the boundaryA machine straining, or one that cannot break away from standstill | Where the answer usually sitsThe starting components and the motor itself |
| What gets watchedTemperature at the compressor body | What pushes it past the boundaryNothing cool arriving to carry the motor's heat off | Where the answer usually sitsCharge, airflow and room load together |
| What gets watchedSupply voltage seen at the board | What pushes it past the boundaryPower arriving outside the band the equipment accepts | Where the answer usually sitsUpstream of the unit, not inside it |
A protection stop and a breaker trip are different events
Both leave a room warm and they carry different meanings. A breaker trip is the electrical supply being cut off outside the machine. A protection stop happens with the supply perfectly intact, decided by the equipment itself.
The rest of the flat gives it away. Lights and sockets on that circuit going dark points to the supply. Everything else still working while only the aircon sits silent points to the machine making its own call. That split decides where a technician starts. What the timing of a breaker trip narrows down is covered separately.
Why nothing happens for a while after a stop
A compressor is not permitted to restart the instant it stops. Controllers hold it off for a set period, and that hold applies whether the stop was routine or a protection event. What an owner sees during the hold is a unit that appears to have died.
The state of the circuit at shutdown is the reason. One end of it is left carrying pressure the other end is not, and a motor asked to push off against that difference pulls a much heavier current than it otherwise would. Standing still lets the two ends come level, so the start happens against nothing.
Two other things use the same window. Heat that gathered in the motor windings has somewhere to go while everything is stationary. Oil thrown out into the pipework during the run gets an opportunity to drain back toward the parts that need it. Neither happens on request, which is why the hold cannot be talked down.
The hold also stops a fault from turning into a frantic repeat. Without it, a system tripping on something that clears the moment it stops would restart, trip, and restart again with no pause at all. The enforced gap turns that into a slow rhythm, which the machine survives, and which anyone watching can actually see.
The hold looks exactly like a dead unit
Nothing about it announces itself. The indoor fan usually keeps blowing, the display stays lit, the remote still answers, and the air arriving at the room is room temperature. Outside, the unit is silent.
That combination is what sends people to the power switch. Flicking it off and on partway through does not shorten anything. Most equipment starts the wait over from the beginning, so the second attempt sits further from a restart than the first one did. A unit that never seems to come back is often one that has never been left alone long enough to try.
Lockout counts the trips, and a reset clears the count
A single stop and a lockout are two different states, and arithmetic is the difference between them. A stop ends one run. A lockout declines to attempt any further runs until something clears it, and it arrives once the same stop has happened a set number of times inside a set window.
That escalation is deliberate. One stop on the fiercest afternoon of the year might genuinely be a one-off, so the controller lets the system have another go. The same stop arriving again and again is no longer a one-off, and carrying on would mean walking the compressor back into the identical condition each time. At that point the board stops offering.
A lockout is therefore a statement about frequency, not about severity. It does not mean anything got worse between the third stop and the fourth. It means the tally reached the figure where trying again stopped being a reasonable thing to do.
Here is where cutting the power does its real work, and it is not the work most people imagine. On a great deal of the equipment fitted in local flats, the tally does not survive losing supply. Take the power away and the latch drops, the code vanishes, the tally returns to zero, and a fresh set of attempts begins. Nothing outside the board moved while any of that happened.
The condition that produced those stops is sitting exactly where it was. Whatever carried a reading past its boundary before will do it again once the unit has run for a while. The tally then climbs back toward the same lockout, and a power cycle buys a repeat of the sequence rather than an end to it.
| What the unit is doing | What that state means | What a power cycle actually changes |
|---|---|---|
| What the unit is doingStops, then comes back on its own after a pause | What that state meansOne stop, cleared by the controller | What a power cycle actually changesNothing, since it was going to return anyway |
| What the unit is doingStops earlier in each run than it did in the last | What that state meansThe condition is worsening and the tally is climbing | What a power cycle actually changesWipes the tally, and hides how fast it was climbing |
| What the unit is doingSilent outdoors, code held on the display, no restart | What that state meansA latched lockout | What a power cycle actually changesDrops the latch, and takes the evidence with it |
| What the unit is doingRuns normally for a stretch after every power cycle | What that state meansSomething that only builds once the unit is under load | What a power cycle actually changesBuys another stretch at the compressor's expense |
| What the unit is doingLocks out again as soon as the power returns | What that state meansA reading already outside limits before any run begins | What a power cycle actually changesNothing, which is itself a useful finding |
The tally is worth more than the code
A code names which boundary got reached. The tally and the spacing between stops say how much effort the machine is spending to stay inside it, and that is the part deciding urgency.
Stops arriving further apart each week describe something drifting slowly. Stops arriving closer together describe something picking up speed. One stop that never comes back describes a day rather than a fault. None of that is legible from a code read once, and all of it disappears the moment supply is cut.
The spacing is the thing to record, because the machine will not keep it. Note the date of each stop, and how deep into the run it landed. A list of four dates says more about what is coming than any single reading taken on the day somebody finally visits.
When the protection reports something that is not there
A protection stop is usually honest and it is not always. The devices doing the watching can fail, and a system then locks out with nothing wrong in the circuit at all. Treating every lockout as real wastes money. Treating every lockout as a bad sensor wastes a compressor.
Three things go wrong on the reporting side. A sensor drifts, so it reports a value the circuit is not sitting at. A connector corrodes or works loose, so the board sees an open circuit and assumes the worst case. Or the board misreads a signal that arrived perfectly correctly.
Timing separates those from a true stop better than any single reading does. A genuine one is tied to conditions. It shows up after the machine has been working a while, or in the worst heat of the afternoon, or when the room is at its busiest. A reporting fault has no such manners. It turns up at start-up, on a cool morning, or the instant power returns, because nothing needs to build first.
Ask which of the two was established before agreeing to replace anything. A sensor swapped out on the strength of a code alone is a guess with a part number attached to it. What settles the question is a reading taken from the circuit while the system runs, set against what the device claims at that same moment. Disagreement makes the device the fault. Agreement means the device was telling the truth.
Codes point at the boundary, not at the cause
The characters on a display name which watcher acted. They do not name what carried the reading there, and most brands have no entry that could. The per-brand error-code pages set out what each entry means on that equipment, and they are the place to confirm the letter and number before anyone quotes against it.
Two units showing identical characters can therefore need entirely unrelated work. A choked outdoor coil and a dying fan motor both finish at the same boundary and both display the same thing. The code narrows the search to a branch. Something measured has to narrow it to a cause.
What a lockout is and is not saying about the compressor
A locked-out system is not a verdict on the compressor. That is the fear this state provokes, and it is usually the wrong fear. The protection exists so that the compressor comes through the condition that triggered it.
The exception lives in the pattern rather than in any one event. A compressor struggling to get going pulls heavy current on every attempt, and the current device ends each attempt. That produces a lockout too, and here the protected part is also the failing one. Telling a weakening compressor apart from a healthy one in bad conditions is what the guide on signs of compressor failure is for.
For everything else the useful frame is a queue. Something is holding heat in, or holding refrigerant back, or holding air still, and the protection sits at the end of that queue rather than the front of it. Work backwards from the boundary that was reached toward the condition keeping it there, and the compressor stays out of the conversation.
The decision on the day is narrower than it feels. A first lockout on an unusually hot afternoon can be watched. A lockout returning after a power cycle has already answered the only question a power cycle is able to ask. A lockout returning sooner each time is the machine spending its own remaining life to make the point, and that version is worth acting on before anything gets opened.
A visit that finishes with the supply cycled and the unit running has repeated what the owner already did. Three statements separate that from a finished job: which boundary was reached, what condition carried the reading there, and what was measured to establish it. Anybody who actually did the work can give all three without being asked twice.
Leaving it alone is itself a test
Keeping away from the switch produces information that touching it destroys. A system left sitting in its lockout still holds the code and still holds the tally. Both are readable by whoever arrives.
There is a narrow case for cycling the power and it deserves naming. A bedroom that has to be usable tonight is a fair reason, and the price is that the record goes along with it. Where the room can wait, waiting is the cheaper choice, because the next visit then begins from evidence instead of from a machine that has been put back to zero.
Ready to get started?
Tell us what’s going on. Symptoms, setup, photos, anything we should know. We’ll assess and come back with the right next step.