How It Ticks

Around the house · Exhibit 9

How does a smoke detector work?

Updated

A smoke detector watches the air inside a small chamber and sounds its horn when that air changes. An ionization detector makes a tiny electric current flow through the air, and smoke chokes the current. A photoelectric detector shines a light across a dark chamber, and smoke scatters some of it into a sensor that normally sits in the dark. Each kind has traditionally been quicker with a different fire, which is why fire services recommend having both, or one alarm with both sensors inside. Below you can fill both chambers with smoke yourself.

If something on the ceiling is beeping right now: loud beeps in sets of three mean smoke or fire and sets of four mean carbon monoxide, and NFPA says to get out, call for help and stay out. A single chirp every 30 or 60 seconds “means the battery is low and must be changed,” and on a sealed 10-year alarm the same chirp means the alarm itself has to go. The chirp guide further down covers the other sounds.

How does an ionization smoke detector work?

The ionization chamber of a smoke detector, opened up. Add smoke with the slider and pick the kind: tiny soot from a flaming fire or larger particles from a smoldering one, at the same weight of smoke in the air. Watch the gauge on the right fall toward the alarm line.

Ionization smoke detectors hold a speck of radioactive americium, one microcurie or less in most units sold today, says the NRC, the US nuclear regulator. Americium gives off alpha particles, and each one tears electrons off the air molecules along its path. That leaves ions behind, air molecules with an electron too few or too many, and two charged plates “attract the ions, creating a small but steady current.” An ion that bumps into a smoke particle sticks to it and never reaches a plate. The current drops, and when it drops far enough, the horn goes off.

One microcurie is 37,000 decays a second, and with my guesses about how much of each alpha’s energy stays in the air, the toy makes about 880 million ion pairs a second and 142 picoamperes, 142 trillionths of an ampere.

The kind of smoke matters more than I expected. At the same weight, smoke made of 0.1-micrometer soot has a thousand times as many particles as smoke made of 1-micrometer particles, and every particle is a trap for ions. A micrometer is a thousandth of a millimeter. In the toy, 0.1 milligram of fine soot in a cubic meter of air cuts the current by one and a half percent, and a full milligram by 13 percent. A full milligram of big smoldering particles barely touches it: 0.15 percent.

Is the radiation in a smoke detector dangerous?

No, as long as it stays closed. “Smoke detectors are very safe” is the NRC’s heading over the dose, and the number is small: “people with two of these units in their homes receive less than 0.002 millirems of radiation dose each year.” The NRC then says an East Coast resident gets that much from natural background in about 12 hours. My own check came out smaller still. CDC, citing NCRP Report 160, gives an average American 33 millirems a year from space and 21 from the ground, 54 in all, which delivers 0.002 millirems in about 20 minutes. Either way, a year under two smoke detectors adds less radiation than an ordinary afternoon. The same NRC study “found that 10 million unwanted smoke detectors each year can be safely put in the trash.”

How does a photoelectric smoke detector work?

The other common kind uses light. Inside a dark chamber an infrared LED shines a beam that the receiver can’t see directly. When particles drift in, they scatter some of that light sideways into the receiver. How much they scatter depends on their size and on their color.

A photoelectric chamber from above, and below it the main picture: the same weight of smoke made of smaller or bigger particles. Slide the size: tiny particles barely light up the receiver (top) but choke an ionization chamber (bottom). The spread buttons show how much the light depends on how mixed the sizes are.

A particle much smaller than the wavelength of light scatters very little of it, and the loss is steep: at the same weight of smoke, halve the size of the particles and the light they send to the receiver falls about eightfold. The LED in my toy shines at 0.88 micrometers, so 0.1-micrometer particles sit deep in that steep part. On the top chart, which compares pale smoke with pale smoke, the light peaks at about 0.3 micrometers and falls again: at 90 degrees, 1-micrometer particles send 1.7 times the light of 0.1-micrometer ones. With the widest measured spread the peak almost disappears, and 1-micrometer smoke sends less light than 0.1-micrometer smoke.

Color matters as much. Real soot from a flame is black and swallows much of the light: in the toy, black particles of 1 micrometer send about a sixth of the light of black 0.1-micrometer soot. Pale smoke against black soot depends on how widely the particle sizes spread, and my first guess there was wrong. Smoke is never one size. With a narrow spread, pale 1-micrometer smoke beat black soot 2.2 times at 90 degrees. NASA and NIST measured wider spreads in real smoke, 1.6 to 1.9, and with those the black soot sends 1.2 to 3.2 times as much light as the pale smoke at 90 degrees. At the widest spread, pale smoke comes out ahead only when the receiver looks well forward, at 30 degrees, and then by 4 percent. The toy’s status line shows this comparison for the angle and spread you pick. So the light side of the toy shows why tiny particles are hard to see. It doesn’t prove which fire a real detector notices first. The ionization side holds with every spread I tried: soot cuts the current about 90 times as much as the same weight of smoldering smoke.

Ionization or photoelectric: which smoke detector is better?

For years the answer was “each is better at one kind of fire.” According to NIST, ionization detectors “tend to detect small black soot particles from flaming fires more quickly because they are produced in greater numbers,” and photoelectric detectors “tend to be more sensitive to particles that are larger in size and white or light-colored, and thus more reflective, like those emitted by smoldering fires.” Alarms built since June 30, 2024 to the 8th edition of UL 217 have to pass tests that, in NIST’s words, “ensure that all detectors are good at detecting smoke from both kinds of fires.”

The US Fire Administration still recommends “Both ionization AND photoelectric smoke alarms, OR Dual-sensor smoke alarms.”

Why is my smoke detector beeping or chirping?

Loud beeps that repeat in sets of three are the fire alarm itself and sets of four are carbon monoxide; NFPA says to get out for both. A short chirp every 30 to 60 seconds is different: the detector is asking for attention, and what it wants depends on its battery. Each answer is my summary; under it are the exact words of NFPA, Kidde or First Alert, with links. German alarms from Ei Electronics, Hekatron and ABUS chirp on their own clocks, every 32, 48 or 40 seconds, so they get a page of their own in German. French ones from Kidde, Lifebox and FireAngel have theirs in French.

Quotes from NFPA, from Kidde and First Alert support pages, from Kidde’s hardwired alarm user guide and from Energizer’s battery handbook, checked on October 14, 15 and 21, 2026. I’m not connected to any of these companies; I name them so you know whose words you’re reading. If none of them describes a sound, I don’t guess.

Why does a smoke detector chirp at night?

At 3 a.m. the reason is cold. Kidde explains that a battery near the end of its life builds up internal resistance, and “A drop in room temperature increases this resistance.” Most homes, Kidde adds, “are the coolest between 2 a.m. and 6 a.m.” Energizer’s handbook gives the chemistry: “Cold temperatures cause the electrochemical reactions that take place within the battery to slow down.” So the tired battery holds out through the warm evening and dips below the alarm’s limit in the coldest hours. There’s no third toy for this: I found numbers for cold and resistance only for an AA cell, as a chart in a picture, and a 9-volt battery is a different thing.

Why is my hardwired smoke detector chirping?

Usually because it has a battery too. Most hardwired alarms run on mains power (120 volts in the US) and keep a backup battery for when the power goes out; Kidde lists a few models that have none. Its guide for the hardwired 20SAR and its relatives says they use “either AA replaceable backup batteries or a 10 yr sealed battery backup”. When that backup runs low, they chirp every 60 seconds. Models with a voice also say “Low Battery”, or “Replace Alarm” on the sealed ten-year ones, whose battery can’t be changed.

A brand-new hardwired alarm can chirp as well. Kidde lists a false low-battery chirp when “AC power is connected before backup batteries activated, or when battery is missing”, and the same happens if it is wired up without removing the battery pull tab or with a battery put in backwards. Three chirps every 60 seconds mean a fault on these models: a smoke sensor error, a CO sensor error on combined alarms, a fatal fault or a stuck Test/Hush button. Two chirps every 60 seconds mean the alarm has reached the end of its life. Older hardwired alarms keep other clocks. Kidde’s discontinued i4618AC chirped “approximately every 30-40 seconds” on a low 9-volt battery, so check which alarm it is, from the name printed on the back, before you go by the rhythm.

Why does my smoke alarm go off when I cook?

Cooking makes particles too, and a detector only sees particles. Even a toaster can do it, according to NIST: “Sometimes something as simple as turning on a toaster can set them off.” Since June 30, 2024, new alarms built to UL 217 have to pass a cooking test, in which “two standardized hamburger patties are broiled 10 feet away from a smoke alarm.” NIST’s verdict after cooking next to new and old alarms: “Compared to the old alarms, the new alarms weren’t universally better at not going off during cooking.” That matters because of what people do next. NIST’s Mensch: “Studies have found that people who repeatedly have these nuisance alarms tend to disable their devices,” and “when an alarm is disabled, it can’t do its main job of alerting people of a real fire.”

Why does my smoke alarm go off when I shower?

What fills a bathroom after a hot shower is a mist of water droplets. A study at the Lovelace Respiratory Research Institute measured them: with hot water, a mass median diameter of 5.2 to 6 micrometers in the bathroom. A photoelectric alarm sees those droplets the way it sees smoke, because they scatter its light. In a 2012 University of Maryland test over steam, “the ionization units performed much better than the photoelectric units”, and “Nearly half of the photoelectric units alarmed to steam exposure.” Alarms that combined several sensors did better there than plain photoelectric ones.

The simpler fix is where the alarm hangs. Kidde says alarms “should not be installed within 3 ft (0.9m) of the door to a bathroom containing a tub or shower”, and that in very humid air “moisture or steam can cause nuisance alarms”. If yours is closer than that, moving it is the fix (for a hardwired alarm, a job for an electrician), not taking the battery out.

How often should you replace a smoke detector?

Every 10 years, whatever battery it takes. USFA says “Replace the entire smoke alarm every 10 years” and to test it monthly, and NFPA agrees: “All smoke alarms must be replaced after 10 years.” The americium hardly notices those 10 years: its half-life is 432.6 years, so it loses 1.6 percent of its activity.

What do the toys simplify?

The ionization toy takes the NRC’s upper limit of one microcurie and gives every decay the energy of americium’s main alpha line, 5,485.56 kiloelectronvolts. In the NNDC tables that line carries 84.8 percent of the decays, and the others are within 2 percent of it. For the energy it takes to make one ion pair I used 34.44 electronvolts, the ICRU Report 90 value for protons in air. ICRU Report 31 gives 35.08 electronvolts for 5 MeV alpha particles, which would lower the current by 2 percent. The rest are my guesses: that half of the alphas fly into the backing of the source and the other half leave about 30 percent of their energy in the air before hitting the wall, 15 percent in all; 0.08 seconds for an ion to reach a plate; how readily ions stick to a particle, in proportion to its diameter; and an alarm line a quarter below clean air. The toy also lets every ion reach a plate in clean air. In a real chamber many would meet an ion of the opposite charge first and cancel out, so the real current is smaller. These guesses move the 142 picoamperes and the alarm level. I checked the direction only against the size spread, where soot always wins; the other guesses could change the ratio, though as far as I can tell not which smoke cuts the current more.

The photoelectric toy computes the scattering with the standard Mie theory for spheres, in my own code, checked to six digits against the tables Warren Wiscombe published in 1979. The guesses: an LED at 0.88 micrometers, all particles with the density of water, black soot with a refractive index of 1.57 + 0.56i and pale smoke with 1.5, sizes spread around the slider value with a geometric spread of 1.6, the low end of what NASA and NIST measured, and a receiver that sees 10 degrees either side of its angle. Real soot grows as fluffy chains of tiny grains, far from a sphere.

Sources