A DIY car presence alarm that can't flatten your starter battery

Last reviewed: 18 July 2026

You want to know when someone is hanging around your car at 3 am, and the commercial options are either a subscription tracker or a screaming siren the whole street ignores. The tinkerer's answer is a small radio node in the car that watches for movement and reports to your smart home. The reason most people don't build one is a fair fear: a badly designed gadget left plugged into a car will, sooner or later, leave you with a car that won't start. This build is designed backwards from that failure. Four parts — always-on power from the OBD port, a low-voltage cutoff, a small buffer battery, and a radar sensor that wakes a sleeping microcontroller — and the worst case is an alarm that stops reporting, never a flat starter battery.

The design problem is the power budget, not the alarm

Detecting a person near a car is easy. Doing it for weeks on a parked car is the hard part, because a parked car is a battery with no income, and that battery has one non-negotiable job: cranking the engine. Any accessory you add is spending from the cranking budget.

So the design rules write themselves. Draw near nothing most of the time. Never drag the battery low enough to threaten starting. Keep working, at least for a while, after you've stopped drawing from the car at all. Everything below is one of those rules made physical. The same budget-first thinking is why an ESP32 on WiFi is the wrong tool for battery jobs — and a car is the same problem wearing a number plate.

Power: the OBD port's always-on 12V

Every car of the last couple of decades has an OBD diagnostic port within reach of the driver's seat, and it carries a battery-voltage supply that stays live with the ignition off — conventionally on pin 16, which is what plug-in dongles use to stay powered while parked. That gives you clean 12 V power with no wiring, no fuse-box spelunking, and no holes in anything: the node plugs in like a dashcam dongle and unplugs when you sell the car.

Verify it on your actual car with a multimeter before trusting it — pin numbering and always-on behaviour is a per-vehicle fact, not a promise. And treat the port as a power outlet only. The moment a project involves splicing into vehicle wiring, adding a fused battery feed, or anything beyond a plug-in accessory, hand it to an auto electrician. Modern cars are networks of computers, and the price of a wiring mistake is not a blown fuse, it's a tow.

The guard: a low-voltage cutoff

Between the OBD plug and everything else sits the one component that makes this build trustworthy: a low-voltage disconnect, set to open somewhere around 12.2 V. A resting healthy battery sits a few tenths above that; a battery sagging to the threshold is telling you it's spent its spare capacity. At the threshold, the cutoff disconnects the node from the car completely, and your gadget's draw drops to zero before the cranking reserve is ever in play.

Set-and-forget matters here: buy or build a cutoff with hysteresis (it reconnects only after voltage recovers well above the threshold, i.e. after the car has been driven) so it doesn't chatter on and off around the limit. Treat the 12.2 V figure as a starting point to verify against your battery type and climate, not gospel — the right threshold is the one that leaves your particular car starting reliably.

The reserve: a small LiFePO4 buffer

A cutoff creates a new problem: the alarm dies exactly when things get interesting, and a thief with a door open and the interior lights running is actively dragging the battery down. The fix is a small LiFePO4 battery inside the node, charged from the OBD supply while it's connected, carrying the node after cutoff.

LiFePO4 suits the job: it handles the temperature swings of a parked car better than ordinary lithium cells, it doesn't mind sitting at full charge, and it's a stable chemistry — which matters in a cabin that hits oven temperatures in an Australian summer. It doesn't need to be big; a node that sleeps properly sips so little that a phone-battery-sized cell buys days of standalone operation. The buffer turns the cutoff from "alarm goes dark" into "alarm switches to its own tank and keeps reporting".

The trigger: radar presence on a wake pin

The reason the whole thing can sleep is the sensor choice. A radar (mmWave) presence module watches continuously for movement and holds a single output pin high when it detects something. Wire that pin to a wake-capable digital input on the microcontroller, and the architecture becomes: radar always watching, microcontroller asleep at microamp levels, radio off. Movement near the car wakes the micro, which sends one short LoRa message and goes back to sleep.

Radar beats PIR in a car because it sees through the plastic of an enclosure, and it doesn't mind cabin temperature swings the way PIR does. Mount the node under the dash or on the parcel shelf, and expect an evening of tuning the sensitivity so a person at the window triggers it and a passing truck doesn't — every cabin is its own radar environment.

LoRa is the right radio for the same budget reasons: a message costs a brief burst instead of WiFi's expensive connection dance, and the range covers a driveway, a street park, or a paddock shed. At home, a receiver node feeds your automation — a phone alert, the driveway light, a camera snapshot. Escalation is a smart-home decision, made indoors, by the part of the system with mains power.

What the failure modes look like now

Run the design forward. Normal weeks: node powered by the car, radar watching, micro asleep, battery cost near nil. Car sits unused too long: cutoff disconnects, the buffer carries the alarm for days, then the node goes silent — and a node that stops checking in is itself an alert your smart home can raise. Someone at the car: wake, message, your house reacts. In no branch does the starter battery go flat because of the alarm. That's the whole trick: pick the worst case first, then build so it's the only one you can afford.

The short version

Power the node from the OBD port's always-on supply, put a ~12.2 V low-voltage cutoff between the car and everything else, add a small LiFePO4 buffer so the alarm outlives the cutoff, and let a radar presence module wake a sleeping microcontroller that sends one LoRa message. The design's promise is in its failure modes: the alarm can go quiet, but it can never be the reason your car won't start. Verify the OBD pin and the cutoff threshold on your actual vehicle, and anything beyond a plug-in accessory belongs with an auto electrician.

General guidance for hobby builds only — verify voltages on your own vehicle, and use a licensed auto electrician for anything beyond a plug-in accessory.

Want the receiver end wired into a smart home that actually reacts — the driveway light, the camera snapshot, the phone alert? That's exactly the kind of setup we do. No lock-in, no upsell to gear you don't need. Tell us what you want to automate and we'll map it out.