The number that decides whether your fridge starts is the surge rating, not the continuous rating. A domestic fridge compressor idles at roughly 100 to 200 W, but for a fraction of a second on every start it pulls several times that, and an inverter sitting comfortably at a third of its continuous capacity all evening can still fold the moment the spike lands. Continuous output tells you whether the backup circuit can hold a steady load. Surge output, and the number of seconds the inverter can hold it, tells you whether the motor will turn over at all.
I learned that in a kitchen at half past nine at night, in the middle of stage 4.
The inverter display was glowing on the passage wall. Inside, the lights were on, the TV was on, the Wi-Fi router, the decoder, two phone chargers. The fridge was humming the way a fridge does. Then it wasn't, because the compressor had reached temperature and cut out, and the house went quiet in that particular way that makes you notice how many things in a home are always making noise. Twenty minutes later it wanted to start again. What you hear is a click from the back of the unit, then a hum that climbs and then bogs down, a low electrical grunt, the sound of a motor that has current but not enough torque to break free. Then the inverter beeps twice, the lights dip, and every socket on the backup circuit goes dead. The fridge's own thermal overload trips a few seconds later. The inverter restarts, the lights come back, and the compressor, still hot and still pressurised, tries again and does the same thing.
That minute is the whole sizing problem, and almost no quote explains it, because a load list is easier to build than a start event is to understand. The fridge appears on the list as 150 W. For most of the day it is 150 W. What trips you is the number that exists for a fraction of a second.
A motor at standstill is almost a short circuit
Before an induction motor spins up, its rotor is not turning, the magnetic field is sweeping past it at full speed, and the current that flows is limited mostly by the winding impedance. Locked rotor current on a small compressor is typically several times the running current, and it is heavily inductive, which means the power factor during that instant is poor. That last part matters more than people realise when they compare a datasheet's surge figure in VA against its continuous figure in watts. The inverter's apparent-power ceiling tends to bite before its real-power ceiling does, so a surge rating quoted in VA is not simply a watt rating wearing a different hat.
Running watts you can add up on paper with a plug meter. Start surge you have to go and find, appliance by appliance, and the biggest one in a South African home is usually not the fridge.
Someone else's motor is often the real problem
Rank the motors on your property by how badly they start. A borehole pump is the worst offender in most houses I have been in, and a fixed-speed aircon compressor is not far behind. A chest freezer, a fridge, a water booster pump, a gate motor, a garage door operator, a vacuum cleaner with a universal motor, all of them are start events, and none of them announce themselves.
The trap is simultaneity. Your load list adds running watts, and running watts do not care about timing. Surges do. If the freezer is already running when the fridge compressor kicks in, you are not looking at one spike against a quiet baseline, you are looking at one spike stacked on top of another motor's running load, and possibly on top of a second spike if the two happen to catch within the same second.
Then there is the grid return. When load shedding ends and the municipal supply comes back, the fridge, the freezer and the aircon all want to start within the same breath, and if the inverter is still carrying the backup circuit at that moment, the pile-up is worse than anything that happened during the outage itself. Changeover is a surge event in its own right, and a slow transfer can stall a compressor that was running happily a second earlier.
What the surge figure on the datasheet actually promises
Surge ratings come with a duration attached, and the durations are not interchangeable. A unit rated at twice its continuous output for five seconds and a unit rated at three times for one second are making different promises, and for a fridge the second one can be the worse buy. The compressor needs a hard pull for a few hundred milliseconds and then a softer period while refrigerant pressures equalise. A short, deep surge window does not cover that second phase.
Two other things get left off the marketing line. Ambient temperature matters, because a garage in January is not a laboratory bench, and thermal derating quietly shrinks both the continuous and the surge figure. And many inverters do not trip when they are over their limit, they fold back, reducing output voltage and frequency to protect themselves. The motor then sits there stalled, drawing current and heating, and nothing beeps. That is a worse outcome than a nuisance trip, because you never notice it happening, and the compressor insulation pays for it.
Three ceilings that sit either side of the inverter rating
The pass-through ceiling
When the grid is present, municipal current flows through the inverter and out to the backup circuit. That AC transfer or pass-through rating is a hard limit on how much of the house you can put behind the unit, even on a sunny afternoon with the grid up. It is one of the reasons a backup circuit ends up being a subset of the house rather than the whole DB, and that decision feeds straight back into the surge problem, because what you leave off the backup circuit determines which motors you have to start.
The battery-side ceiling
An inverter nameplate is a promise the battery has to keep. At 48 V, a 100 A discharge limit works out to roughly 4.8 kW before losses, and a battery management system may clamp below that, or clamp for a shorter time than the inverter's surge window. A 5 kW inverter behind a battery that can only feed 3 kW is a 3 kW system with a bigger badge. The surge case is where this shows up first, because it is the moment of highest current draw in the whole day.
The grid-connection setting
Any inverter that runs parallel to the municipal supply has to satisfy NRS 097-2-1 for its protection and disconnection behaviour, and the installation itself sits under SANS 10142-1. Practically, the maximum feed-in or zero-export setting you register with your municipality also shapes how the backup circuit behaves, because it changes how the inverter manages its own output and headroom. A proposal that never mentions the export setting is a proposal that has not been thought through to the end.
Why oversizing is not the free fix
A bigger inverter draws more standing current, all night, every night, and it costs more to buy. It is also less efficient at the light loads that make up most of the backup circuit's day. And it does not solve anything if the battery behind it cannot deliver the current, or if the pass-through is still the bottleneck. Inverter sizing is not a case of going up a size to be safe. It is a case of finding the highest current moment the house will actually produce and making sure every component in the chain clears it.
What I check before I would sign anything
- List only the loads on the backup circuit, and their running watts. Not the whole property. The geyser element and the stove are almost always somewhere else for a reason.
- Find the single worst start event on that list. In my experience it is a pump, then a fixed-speed aircon, then a freezer, then the fridge. Use the manufacturer's figure if there is one, or have it measured, rather than guessing.
- Add that start surge to the running watts of everything else that could plausibly be on at the same moment, including a second motor. Compare the total against the surge rating and its duration, not against the continuous rating.
- Check the battery's continuous and peak discharge current against that same worst-case number. If the battery cannot feed it, the inverter's rating is decoration.
- Check the pass-through figure against the backup circuit's full demand while the grid is up.
The cost that never shows up as a beep
There is a second failure mode, and it is the one that costs the most. If your inverter is marginal rather than hopeless, you may not get trips at all. You get a compressor that starts every cycle against a supply that sags slightly under it. Nothing complains. Nothing gets logged. The fridge just dies at year five instead of year fifteen.
The nuisance trips are annoying, and the family adapts around them, waiting for the fridge before switching on the kettle, not running the vacuum while the freezer is cycling. That adaptation is a signal. It means the system is being managed by people instead of by its own headroom, and the appliance that pays for it is the one nobody thinks to check until the compressor is already gone.
