UPS Sizing and Runtime Calculator

Ask what a battery backup is for and most people answer that it keeps the machine running. It does not, at any price you would pay for a desk. What it buys is the two or three minutes between the lights going out and the operating system finishing a shutdown it started on its own.

What the protected equipment actually draws at the wall, not the sum of the power supply labels. A wall meter beats a nameplate.
From the unit label
Also on the label, and always lower than the VA figure. If only VA is printed, leave this blank and the calculator assumes a power factor.
Only used when the watt rating is blank. Watts equals VA times power factor.
Two 12 V blocks in series is 24 V. Count the blocks and multiply.
Per string, at the rating discharge time below
Watt-hour mode only
The discharge duration the amp-hour figure was measured over. Twenty hours is the usual convention for sealed lead-acid.
How steeply capacity falls at high discharge rates. Around 1.1 to 1.3 for sealed lead-acid, near 1.05 for lithium. Set to 1 to switch the effect off and see the naive figure.
A backup unit stops at its low-voltage cutoff rather than emptying the pack
Sealed lead-acid loses capacity steadily and is often near half by year four. A unit that has never been load tested is an unknown.
Time your own shutdown with a stopwatch. Flushing large caches or closing a database takes far longer than a desktop.
Most units wait to see whether the power comes straight back before triggering anything
Multiplier on the total shutdown time. Two is a common choice and is not a rule.
UPS Sizing Calculator — Runtime, VA and Shutdown MarginBuildFigure

VA and watts are not two ways of saying the same thing

Every backup unit carries two ratings and they are not interchangeable. VA is apparent power, the product of volts and amps. Watts is real power, the part that actually does work. On a purely resistive load the two are equal; on a switching power supply they are not, and the ratio between them is the power factor. A unit labelled 1500 VA and 900 W has a power factor of 0.6, which is why the watt figure is the one that binds.

The practical consequence is that people size against the larger number and come up short. A load of 1000 watts is well under 1500 VA and comfortably over 900 watts, so the unit refuses it. The calculator above shows the load against both and tells you which one you hit first. If the label only gives VA, it assumes a power factor and says so, but the label figure is always better than the assumption.

Runtime does not scale the way division suggests

Divide watt-hours by watts and you get hours. That is arithmetic, and it is wrong by a wide margin for a lead-acid pack under a heavy load. Two things intervene. The inverter loses something turning DC into AC, typically ten to twenty percent, which raises the draw the battery sees. And the battery itself delivers less total energy the faster you take it, an effect described by the Peukert exponent, which is around 1.1 to 1.3 for sealed lead-acid and much closer to 1 for lithium.

The result is a curve, not a line. Halving the load more than doubles the runtime, which is the single most useful thing to know when you are trying to buy time. Moving a monitor and a printer off the protected outlets often does more for your shutdown window than a bigger unit would, and costs nothing. The calculator shows the half-load figure next to the full-load one so the shape is visible rather than asserted.

Sizing for a shutdown, not for an outage

The instinct is to ask how long the machine can keep working. For a desk-sized unit the honest answer is a handful of minutes, and buying enough battery to change that answer meaningfully costs more than most people expect and takes a floor-standing cabinet. So the useful goal is different: keep the machine alive long enough for it to notice the power is gone, decide the outage is real, and shut itself down in an orderly way.

That budget has three parts. There is a detection delay, because the unit sensibly waits to see whether the power comes straight back rather than triggering on every flicker. There is the shutdown itself, which for a desktop with nothing open is seconds and for a machine flushing a large write cache or closing a database can be minutes. And there is a safety factor, because the pack is older than you think and the load may be higher than you measured. Time your own shutdown with a stopwatch rather than guessing at it — the number is usually a surprise in one direction or the other.

What you are protectingWhat the reserve is really for
Desktop with unsaved workEnough time to save and stop, which is mostly about you noticing
Machine set to shut down automaticallyDetection delay plus the full shutdown, doubled
Storage array mid-writeFlushing caches before the power goes, which is the whole point
Network gear that must stay upA small constant load for a long time, which is a different sizing problem entirely

Batteries age, and they do it quietly

A sealed lead-acid pack loses capacity from the day it is made, faster in a warm room, and by year three or four a unit that self-tests as healthy can deliver half its original runtime. The unit rarely tells you this in a way you will notice, because its self-test is brief and light and passes long after the pack has stopped being useful for a real outage.

The condition field in the calculator lets you model that, but the only figure worth having is measured. Once a year, with the real load connected and everything saved, pull the mains plug and time it until the machine shuts down. Write the number on the unit in marker. When that number stops covering your shutdown budget, the pack is finished regardless of what the panel says. Chemistry, discharge behaviour and depth of discharge are covered in more depth in the battery runtime calculator, which handles larger banks and other chemistries.

What sits on the protected outlets

Nearly every unit has two groups of outlets, one with battery backup and one with surge protection only, and putting the wrong thing in the wrong group is the most common setup mistake. Anything with a motor or a heating element — a laser printer above all — has no business on the battery side, where it will consume the runtime you were saving for the machine. Monitors are a judgement call: you need one long enough to see what is happening, and you do not need three.

Work out the load figure from a wall meter rather than from labels. Power supply labels state a maximum the machine will never draw, and adding those maxima together produces a number two or three times the truth, which leads to buying a unit far larger than needed. The power supply wattage calculator gets you an estimate for a build that does not exist yet, and the circuit load calculator handles the question one level up, where the whole room shares a breaker.

Questions people ask

Why is the watt rating so much lower than the VA rating?

Because they measure different things. VA multiplies volts by amps without regard to whether the current is doing useful work; watts counts only the part that is. Switching power supplies draw current in a way that makes those two figures diverge, and the ratio is the power factor. A unit's watt rating reflects what its inverter and transformer can actually deliver. Size against watts, treat VA as secondary, and if only VA is printed on the label, the manufacturer's own datasheet will have the watt figure somewhere.

My unit says it will run for 20 minutes but it lasts five. Why?

The published runtime chart is measured with a new pack at a stated load, usually a fraction of the unit's rating, at room temperature. Your pack is older, your load is probably higher than the chart point, and if the room is warm the pack has aged faster than the calendar suggests. Runtime also falls faster than linearly with load, so being at 70 percent of the rating rather than the chart's 30 percent costs more than the ratio implies. Five minutes from a unit rated for twenty at a lighter load is unremarkable rather than faulty.

Should I get a bigger unit or reduce the load?

Reduce the load first, because the return is better and it is free. Runtime rises faster than the load falls, so taking a monitor and a peripheral off the protected outlets buys more minutes than the wattage change suggests. Only once the protected load is genuinely minimal does buying capacity make sense. The exception is when the load is above the unit's watt rating, in which case no amount of load shuffling fixes a unit that cannot carry the load at all.

Does a battery backup protect against surges and lightning?

Partly, and less than the box implies. Units include surge suppression, and the ones that continuously regenerate the waveform do more than the ones that switch over only on failure. None of them are a defence against a direct strike, which will find a path regardless. Treat surge protection as a secondary benefit and the shutdown window as the reason you bought it. And note that the protection components degrade with each event they absorb, invisibly, which is another argument for periodic replacement rather than indefinite trust.

Can I use one to keep a storage array safe?

It is one of the better reasons to own one, because an array caught mid-write during an abrupt power loss is exactly the scenario that turns a redundant array into a damaged one. The requirement is that the array is actually told to shut down, which means the data cable between unit and machine is connected and the monitoring software is configured and tested. An untested shutdown script is a shutdown script that does not work. Sizing the array itself is a separate exercise, covered in the storage array capacity calculator.

Related