Your Battery Backup Looks Fine. That Is Not the Same as Working.

A battery backup — a UPS, an uninterruptible power supply — is the only piece of equipment in a house or a shop whose entire job is invisible. A printer that has stopped working tells you it has stopped working. A router that has died takes the internet with it. A UPS with a completely ruined battery does none of that. It sits there with a green light, the equipment plugged into it runs perfectly, and the first thing you learn about the battery is that there is no battery, at the precise moment you needed one.

We see this constantly, and almost always in the same shape: someone did the sensible thing four, six, eight years ago, bought a good unit, plugged the important things into it, and then quite reasonably never thought about it again. Nothing in the intervening years asked them to. The unit did not complain, the equipment did not misbehave, and the power did not go out for long enough to expose anything. Then a Public Safety Power Shutoff arrives, or a car takes out a pole on the corner, and everything drops instantly — which is the same thing that would have happened with no UPS at all, after several years of believing otherwise.

This is a short, cheap job that almost nobody does, so it is worth doing properly. Below: why the failure is silent, what the automatic self-test genuinely checks and the specific blind spot that applies to most home setups, the manual test that gives you a real number, how long these batteries actually last and what shortens them out here, and how to replace the battery rather than the whole unit.

Why a dead battery makes no noise and changes nothing

The reason the failure is silent is structural rather than a defect. In normal operation your equipment is running on utility power passing through the unit. The battery is not in the path; it is standing by. It only carries the load in two situations — a real power event, and the brief self-test the unit runs on itself.

Schneider Electric, which owns APC, describes what that test involves: the unit "will force itself to run on battery for approximately 10 seconds," then, if the result is acceptable, "the UPS will go back to its previous online state." Ten seconds a week, and the rest of the time the battery is a lump in a box that nothing is asking anything of.

So the daily evidence available to you — everything works, nothing beeps, the light is green — is evidence about the mains supply and the pass-through, not about the battery. A unit whose battery has been useless for two years produces exactly the same evidence as one that was replaced last month. There is nothing to notice, which is why nobody notices.

The self-test is real, and it has a blind spot that fits most homes exactly

The automatic test is a genuinely good feature and it does catch batteries. Schneider documents three ways it runs — manual, automatic, or initiated by software — with the automatic test happening "upon power on" and then "every after 7 days after a Back UPS is powered on." If the battery cannot hold up, the unit "will report a bad/replace battery alarm."

Here is the part that is not on the box, and it is the reason we wrote this page. During those ten seconds the unit calculates runtime from whatever is plugged into it, and that calculation has a working range. Schneider states it plainly: the assessment is accurate when "the load attached is within 10% to as much as 30% of the unit's capacity." Below that, the page is equally plain — loads that are too low prevent proper runtime calculation. Above roughly 80 percent, you get battery drain and false alerts instead.

Now think about how a typical household UPS is actually loaded. The advice we give, and stand by, is to put the modem and the router on the battery outlets, because network gear draws very little and keeping it alive changes the character of an outage. A modem and a router together are a couple of dozen watts. On a unit sized for a desktop computer, that is often two or three percent of capacity — comfortably below the ten percent floor where the self-test can compute a meaningful runtime at all.

The conclusion is not that the advice is wrong; a lightly loaded unit is exactly what you want for a long outage. The conclusion is that on the setup most likely to matter during a shutoff, the automatic test is the least able to warn you, and the burden of finding out shifts to you. If your UPS is running a small network load and you have been treating a green light as a passed exam, you have not had one.

What the lights and the sounds actually mean

Two signals get misread in opposite directions, and both are worth knowing before you test anything.

The first is the replace-battery warning, which is not subtle once you know its vocabulary. Schneider describes it as either "a constant tone alternating red and green light" or "a rapid chirping sound," depending on the model, alongside a lit Replace Battery indicator. If your unit is doing this, it has already told you the answer and you can skip the testing entirely — although there is one honest caveat, below.

The second is the beeping during an actual outage, which is the single most commonly misinterpreted noise in home technology. That is a healthy unit doing its job: it is telling you it has taken over and you are on a clock. People routinely respond by unplugging the UPS or switching it off to stop the noise, which ends the backup they are standing in the middle of. If the power is out and the box is beeping steadily, the box is fine. Save your work.

The caveat on the replace-battery light: do not condemn a battery on a unit that has been unplugged, moved, or sitting in a cupboard. Schneider's own troubleshooting sequence is to charge the battery to 100 percent for a minimum of eight to twenty-four hours first, then run a self-test, and only then conclude that "the battery needs to be replaced" if the light stays lit. A deeply discharged battery can report as bad and recover.

The ten-minute test that gives you a straight answer

The self-test answers "can this battery hold up for ten seconds." The question you actually have is "how long will this thing keep my equipment alive," and there is only one honest way to find out, which is to run the real load on the battery and time it.

Do it in this order. First, save and close everything on any computer that is plugged in, because the entire point of the exercise is that the answer might be zero. Second, make sure the unit has been powered from the wall for at least several hours so the battery is actually charged. Third, with the equipment running as it normally runs, pull the plug out of the wall — not the switch on the unit, the plug — and start timing.

What you are looking for is one of three results. The equipment drops instantly or within a few seconds: the battery is gone, and you have just learned it for free rather than during a shutoff. The equipment runs for a while and then drops much sooner than you expected: the battery is tired and partway through its life, which is useful to know and worth writing down for comparison next year. The equipment runs long enough that you get bored and plug the wall back in: healthy, and now you have a number.

Write that number and the date on a piece of masking tape on the side of the unit. This is the only way this test gets more useful over time, because the figure that tells you the most is not this year's minutes, it is the fact that they are half of last year's.

One warning about a result that looks like a dead battery and is not. If the equipment drops the instant you pull the plug, check what is plugged into which socket before you order anything. On most units only some of the outlets are battery-backed and the rest are surge-only, and a computer in the wrong half behaves identically to a computer on a dead battery. Schneider lists that exact mix-up first among the causes of connected equipment turning off unexpectedly, ahead of overload, a tripped breaker and a faulty unit. We have written elsewhere about the four ways people set a UPS up wrong; if the test fails, read that before you spend money.

Runtime calibration: the precise version, and why not to keep doing it

Units with a management interface offer a runtime calibration, which deliberately runs the battery down under load so the unit can recalculate what it can actually deliver. It is the accurate version of the test above and it will give you a better estimate than a stopwatch.

It also costs you battery. Schneider lists calibration among the factors that shorten battery life and recommends that runtime calibrations be performed only once or twice a year. That is a good rhythm for the manual test as well. Testing a battery is a small, deliberate withdrawal from the thing you are testing, so twice a year is diligence and monthly is damage.

If you want a date rather than an intention, use the two clock changes. They are already the standard prompt for smoke-alarm batteries, they are six months apart, and adding one plug-pull to that ten-minute round of the house is the entire maintenance regime this equipment needs.

In Southern California, heat is most of the story

The published lifespan for these batteries is not long to begin with. Schneider states that "most APC batteries should last three to five years," and separately recommends replacing before the battery reaches five years old rather than waiting for it to fail.

What decides where you land in that range, more than anything else, is temperature — and the rule is much harsher than people expect. The optimal operating range for a sealed lead-acid battery is 20 to 25 degrees Celsius, which is 68 to 77 Fahrenheit. Above that, Schneider gives a rule of thumb: "every 8°C (15°F) increase in temperature will cut the battery's life in half." Their own worked example is a battery that would last four years at 77 degrees lasting about two at 95.

Then consider where these units actually live around here. In the garage, on the floor, next to the water heater. In a closed-up back office in Riverside or the Coachella Valley over a summer weekend, with the air conditioning off from Friday evening to Monday morning. In a cabinet with no ventilation, under a desk against a sunny wall, or stacked directly beneath the equipment it is powering. A garage in the Inland Empire spends a good part of the year well above 95 degrees, and by that rule of thumb a five-year battery in it is a two-year battery, or less.

This is the most actionable thing on this page, because it is free. Moving the unit off a hot floor and out of a hot room, giving it air on both sides, and keeping it out of direct sun buys you years of battery for the price of walking across the room. If your UPS lives in the garage alongside the router and the backup drive — a very common arrangement out here and one we have written about separately — that corner is worth rethinking as a whole rather than one item at a time.

It is a battery cartridge, not a unit — the part number is the whole job

The most expensive mistake at this point is assuming a dead battery means a dead UPS. On the great majority of home and small-office units the battery is a user-replaceable cartridge with its own part number, and the electronics — which is where the real cost and the real engineering sit — are fine.

Finding the right part is mechanical if you do it in the right order. Schneider's instruction is to get the model number of the UPS itself, from the white barcode sticker on the back or the underside of the unit, where the top row of the sticker is the model number. Put that into APC's official replacement battery selector and it returns the cartridge number. Two things they flag are worth repeating because both catch people out: do not search using the model number of the battery currently inside the unit, and note that where a unit contains several batteries the selector returns a single part number covering all of them.

We are not going to quote a price, because cartridge and unit prices both move and a number written today is wrong by spring. The comparison to make on the day is the current price of the correct cartridge against the current price of an equivalent new unit with a warranty. On a decent line-interactive unit the cartridge usually wins comfortably. On a very old, very cheap unit it sometimes does not, and if yours predates the equipment it is protecting, replacing the whole thing is a reasonable call.

One thing to check before ordering either way: whether the unit is old enough that its capacity no longer suits what you have plugged into it. Equipment changes over eight years. A unit bought for a tower desktop and now running a laptop, a mesh node and a modem is oversized, which is harmless; a unit bought for a small setup and now carrying a desktop, two monitors and a network switch may be near its limit, and running above 80 percent of rated capacity shortens the life of whatever battery you put in it.

The old battery does not go in the household bin

The cartridge you have taken out is sealed lead-acid, and in California it is not ordinary rubbish. The state's Department of Toxic Substances Control treats spent lead-acid batteries as hazardous waste, and the public route is a household hazardous waste collection facility — most Southern California counties and cities run them, and they take these without argument.

There is also a retail route worth knowing about, because it is free and it is the law rather than a courtesy. California's Rechargeable Battery Recycling Act defines a rechargeable battery to include "a small, nonvehicular, rechargeable nickel-cadmium, nickel metal hydride, lithium ion, or sealed lead-acid battery," and requires retailers to provide "the take-back at no cost to the consumer of a used rechargeable battery, the type or brand of which the retailer sold or previously sold." In practice, ask at the counter where you buy the replacement.

What not to do is leave it in the garage next to the new one, which is the outcome in a surprising number of the houses we visit, sometimes several deep.

Where we come in

Most of this is a job for a Saturday morning and a stopwatch, and we would rather you did it yourself than paid anyone for it. The part worth calling about is the layer above it: whether the things that need to stay up are actually on the battery outlets, whether the unit is the right size for what it is now carrying, and whether the plan survives the length of outage you actually get out here rather than the ten-minute one everybody imagines.

We work across the San Gabriel Valley, Orange County, Long Beach, the Inland Empire, the Coachella Valley and the coast, and this comes up every autumn as the winds and the shutoff notices start. If you want it looked at, the useful appointment is now, in September, and not in the middle of a red flag warning when the answer is whatever is already plugged in.