Sizing an Inverter or UPS for a Lab Fridge: Startup Surge, VA vs Watts, and Runtime

Size backup power for a laboratory refrigerator against three separate numbers, not one: the surge in volt-amperes the compressor demands at the instant it starts, the continuous power in watts it draws once it is turning, and the energy in watt-hours the battery has to return across the outage you actually get. The first decides whether the appliance starts at all, the second decides whether the inverter overheats or folds back, and the third decides what the installation costs. A cabinet that draws 90 W while running can ask for well over 1 000 VA for a fraction of a second at start, which is why a unit chosen on running watts alone trips every time the thermostat closes.

What follows is the arithmetic behind those three numbers: where each is written down, how to measure it in your own room rather than borrow it from a temperate datasheet, and which parts of the calculation are genuinely soft.

Why a 90-watt fridge needs a 1 000 VA inverter

A small refrigerator is driven by a hermetic motor-compressor: a single-phase induction motor sealed in the same shell as the pump. At the moment power is applied the rotor is stationary, the motor presents almost no back-EMF, and it draws locked-rotor current — a figure conventionally several times the rated-load current. Electrical codes treat the two as distinct design quantities rather than as one rating. Article 440 of the National Electrical Code requires a hermetic refrigerant motor-compressor to be marked with its rated-load current, and sizes the disconnecting means at not less than 115 % of the rated-load current or branch-circuit selection current, whichever is greater, with the locked-rotor rating considered separately ^[2]. Whatever national wiring code applies where you are, the same distinction is in it, because a device sized only for running current will not survive a start.

For small hermetic compressors, locked-rotor current is commonly around four to six times rated-load current. That surge collapses within a fraction of a second as the rotor accelerates — but only if the supply holds its voltage. If the source sags under the inrush, the motor accelerates more slowly and therefore draws high current for longer, which sags the supply further. The failure is self-reinforcing: the compressor stalls, the internal thermal overload opens, and the cabinet sits warm until someone notices, leaving a temperature excursion to assess afterwards.

• Ask for the surge window, not just the surge number. "3 000 W peak" printed with no stated duration is marketing; a rating held for 20 ms does not cover a motor start, and one held for two seconds does.
• Take locked-rotor amperes from the compressor plate where it is marked, and treat any multiple-of-running-watts rule of thumb as a starting point to be measured, not a result.
• The inverter is not the only thing being surged. Undersized cable and a marginal breaker also drop voltage at exactly the wrong instant.
• A soft starter or start-capacitor kit reduces the current demanded at start, and is often cheaper than the next inverter size up.
• Where an appliance is designed for constrained supply the specification says so: WHO prequalification requirements for cold-chain refrigerators cover compressor starting alongside power consumption and rated ambient ^[6].

VA and watts are not the same number

Apparent power, in volt-amperes, is simply voltage times current. Real power, in watts, is what actually does work: apparent power multiplied by the power factor. A kettle is resistive and has a power factor of about 1, so its VA and W are the same. An induction motor is not: current lags voltage, and a small hermetic compressor typically runs at a power factor somewhere between 0.6 and 0.8. This is why the box carries two ratings. An inverter or UPS marked 1 000 VA / 800 W has a rated output power factor of 0.8, and it can deliver its full watts only into a load at least that good. IEC 62040-3 is the standard that defines how UPS performance is specified and tested, including the output rating and the classification code that describes how far the output is independent of the input ^[1].

Measure your own load instead of trusting the plate

Nameplates give worst-case electrical ratings, not what a given cabinet does in a given room. The variable that decides energy consumption is duty cycle: the fraction of each hour the compressor actually runs. An appliance that runs perhaps a third of the time in a 22 °C laboratory can run two-thirds of the time or continuously in a 35 °C store room, because the heat it is rejecting is set by the temperature difference across its walls while its condenser rejects into hotter air. Domestic energy figures are measured under the defined laboratory conditions of the household refrigerating appliance standard ^[10], which a hot store room does not reproduce. Cold-chain prequalification handles this by requiring performance to be declared at the maximum ambient of the zone the appliance is rated for ^[6] — borrow that habit and never record an energy figure without the room temperature beside it.

• Log a full 24-hour kilowatt-hour total with an in-line energy meter, with an ambient temperature trace alongside it, in the hottest week you get rather than the most convenient one.
• Capture running watts and power factor while the compressor is on, not the hourly average — the average understates what the inverter must supply during a run.
• Capture the start surge with a clamp meter that has an inrush function. A plug-in energy meter averages over roughly a second and will miss the event entirely.
• Count starts per hour, and measure the inverter's own no-load consumption with the appliance unplugged.
• Characterise the supply as well as the load: mains voltage that sits low all afternoon turns a marginal start into a stalled compressor. South Africa's quality-of-supply standard defines the voltage characteristics, compatibility levels and assessment methods used at the point of supply ^[8]; elsewhere the equivalent document is published by the national electricity regulator, and is worth reading rather than assuming a figure.

Runtime: turning hours into watt-hours into a battery

Runtime is an energy problem, and the inverter's power rating has almost nothing to do with it. The quantity you need is battery energy: average load power times outage hours, divided by inverter efficiency, divided again by the fraction of the battery you are willing to use. Work an example. A cabinet measures 1.4 kWh across 24 hours in a 32 °C room — an average of about 58 W. During a four-hour afternoon interruption the room is hotter and the duty cycle rises, so take the average as 90 W: 360 Wh at the appliance. At 85 % conversion efficiency, plus roughly 12 W of inverter idle draw across those hours, the battery must give up about 470 Wh.

That is where the chemistry decides the price. Flooded and sealed lead-acid banks are conventionally sized so that a routine discharge takes no more than about half the rated capacity, which turns 470 Wh of demand into roughly 940 Wh of nominal capacity — a nominal 12 V, 100 Ah battery is 1 200 Wh, so it looks comfortable. Two corrections spoil that. First, rated capacity is quoted at a specified discharge rate, commonly the 20-hour rate, and available capacity falls as the discharge rate rises; the Peukert relation is the usual correction, and a critical review of it in the battery literature concluded that it cannot predict remaining capacity accurately unless discharge current and temperature are both constant ^[5]. Neither is constant behind a thermostat that switches. Second, capacity and service life are temperature-dependent: the IEEE recommended practice for sizing lead-acid batteries takes cell capacity at 25 °C as the reference and applies correction factors for other electrolyte temperatures ^[3]. A battery cupboard at 35 °C returns marginally more capacity and loses a great deal of life.

Lithium iron phosphate changes the arithmetic rather than removing it: a usable depth of discharge of 80–90 % rather than 50 %, less sensitivity to discharge rate, better tolerance of a hot room — offset by charge-temperature limits of its own and a management system that will disconnect the load to protect itself. Whichever chemistry, the selection, installation, maintenance and testing framework in the IEEE guide for UPS batteries is worth borrowing ^[4], particularly when the real choice is between a cheap flooded string nobody will top up and a sealed one nobody has to.

Two design points get missed. Size against the outage you actually get — the published block duration on a scheduled shedding grid, or whatever your own interruption log shows — not an imagined worst case that triples the budget. And check the recharge window as carefully as the discharge: a bank that cannot be brought back to full between two interruptions is flat by the third, and the charger rating usually limits that, not the battery.

Transfer time, waveform and restart lockout

A compressor tolerates a brief break in supply without complaint. What it does not tolerate is being restarted immediately against the head pressure still standing in the system — which is exactly what a supply that flickers off and straight back on does. A restart-delay timer enforcing a few minutes between stop and start prevents a stalled rotor, and belongs between the backup supply and the appliance whichever topology you buy.

Topology matters too, and IEC 62040-3 gives it a vocabulary ^[1]. A VFI unit conditions the output so that it is independent of both input voltage and input frequency; a VFD unit's output simply follows the input, which is acceptable for equipment with a switch-mode supply and much less so for a motor asked to start while the mains is sagging. Output waveform is the other half of the same question: a modified or quasi-square wave contains harmonics an induction motor converts into heat and noise rather than torque. For a compressor, specify a sinusoidal output and confirm it is declared rather than implied.

• A small desktop UPS bridges seconds, not hours. It is the right device for the data logger and the alarm, and the wrong device for the compressor.
• A battery inverter with automatic transfer is the usual answer for a scheduled multi-hour interruption, provided both the surge rating and the recharge rating are checked.
• A generator with auto changeover covers long or unscheduled outages, but has start latency, needs fuel present, and puts a second motor-starting problem in series with the first.
• Fit surge and undervoltage protection on the appliance circuit; restoration transients damage more compressors than outages do, and test the whole arrangement by pulling the mains deliberately with the logger running.

When holdover beats runtime

There is an alternative to storing energy in a battery, which is storing cold in the cabinet. WHO's prequalification specification for solar direct-drive refrigerators without battery storage requires a minimum holdover of 20 hours at the maximum ambient temperature of the rated zone, defined as the period during which all points in the storage compartment stay inside the band with no energy input ^[7]. That is a design philosophy as much as a product category: insulation and thermal mass do not degrade at 35 °C, need no charge controller, and cannot be found flat. The caveat is the band. Those figures are published for appliances holding +2 °C to +8 °C, not for a freezer at −20 °C, and they do not transfer to what a lyophilised solid is specified to need.

Know what your supply is before designing around it. The Multi-Tier Framework used to measure electricity access grades a connection on capacity, availability, reliability and quality rather than on mere existence, with Tier 3 defined at a minimum of 200 W available and at least eight hours a day ^[9]. A refrigerator plus its start surge sits above where many connections actually are, and the honest conclusion is sometimes that the appliance, not the inverter, is the wrong choice for the site — or that the material does not need refrigeration at all.

Where the numbers are soft

Some of this is firm. Apparent power and real power are defined quantities, the code distinction between running and locked-rotor current is written into wiring regulations ^[2], and the reference temperature and correction structure for lead-acid sizing are published ^[3]. Quote those with the source named and they hold up.

What is soft is most of what appears in a product listing. IEC 62040-3 provides a method for specifying and testing UPS performance ^[1], but low-cost inverter-chargers sold for household backup are frequently not declared against it, so surge figures from two vendors are not comparable and may not have been measured over the same window. The four-to-six-times multiple for locked-rotor current is a rule of thumb across a wide population of compressors, not a property of yours. Peukert-style corrections are approximations the literature itself calls unreliable outside constant current and constant temperature ^[5]. Published holdover performance exists for the vaccine band, not for research freezers ^[7]. And appliance energy figures come from standardised laboratory conditions ^[10] that no hot store room reproduces. Where a decision rests on one of these, the resolution is a logged measurement in your own room rather than a better guess.