For most winter DIY automation builds, you need three sensor categories and nothing more: a temperature and humidity sensor, a presence or state sensor (motion, contact, or limit switch), and one safety sensor matched to the hazard you are automating around. A combined temperature-humidity part such as a DHT22, SHT31, or BME280, plus a magnetic reed switch or PIR motion module, covers the large majority of heating, greenhouse, freeze-protection, and door-automation projects people start in cold months. Winter changes the requirements more than the project list. Condensation, sensor self-heating, battery chemistry, and the difference between air temperature and surface temperature all shift which part is the right one, and these are the failure points that turn a working bench prototype into an unreliable installation in an unheated garage.
Table of Contents
- The three-sensor core, and what each one actually measures
- Why air temperature is often the wrong measurement
- Condensation, self-heating, and other cold-weather failure modes
- Matching sensors to four common winter builds
- Interfaces, wiring, and where projects actually break
- Frequently Asked Questions
The three-sensor core, and what each one actually measures
A temperature-humidity sensor answers "what are the conditions?" The cheap DHT11 reads in whole degrees with poor humidity accuracy and is adequate only for coarse thresholds. The DHT22 improves on it modestly. If a decision depends on the reading – running a heater, venting a greenhouse, triggering a freeze alarm – use a digital I2C part like the SHT31, SHT41, or BME280, which report calibrated values over a bus rather than a timing-sensitive single-wire protocol. The BME280 adds barometric pressure, which is useful for weather logging and irrelevant for most indoor control. A presence or state sensor answers "did something change?" A magnetic reed switch tells you a door, window, lid, or vent is open or closed, and it is the most underrated part in home automation because it is a simple contact closure with no calibration and almost no current draw.
A PIR motion sensor detects a warm body moving across its field of view. A limit switch or endstop tells a moving mechanism where it is. Pick the one that matches the question you are asking – motion sensors are a poor substitute for knowing a state, because they report events, not conditions. A safety sensor answers "is something going wrong?" This is the category people skip, and it is the one that matters when the project drives a heater, a pump, or a motor. The specific part depends on the hazard: a water leak sensor under a pipe or water heater, a thermal cutoff or second independent temperature probe on anything producing heat, a current sensor to confirm a load actually switched, or a commercial CO alarm near any combustion appliance.
Why air temperature is often the wrong measurement
The sensor most beginners reach for measures air temperature a few centimetres from the board. Many winter problems are about surfaces, not air. A pipe freezes because the pipe wall reaches freezing, which can happen while the room air sits above it. A floor feels cold because of surface temperature. A plant seedling suffers from root-zone temperature.
For surfaces, a DS18B20 in a waterproof stainless probe is the common choice: it is a digital one-wire sensor, available pre-wired in a sealed metal sleeve, and can be strapped directly to a pipe under insulation or pushed into soil. Multiple DS18B20 sensors share a single data pin because each carries a unique address, which makes it practical to monitor several pipes or zones with one microcontroller. The trade-off is that it reads temperature only, so a humidity-dependent build still needs a second part. For non-contact readings, an infrared thermopile such as the MLX90614 reports the temperature of whatever fills its field of view. This is genuinely useful for spot-checking a radiator, a duct, or a window pane, but the field of view widens with distance, so a reading taken from across a room is an average of a large area, not the small target you intended.
Condensation, self-heating, and other cold-weather failure modes
Cold air holds less moisture, and a sensor moved from cold to warm – or sitting in an unheated space where the sun warms it each afternoon – cycles through the dew point repeatedly. Liquid water on an exposed humidity sensing element corrodes it and drifts the readings permanently. Parts with a protective membrane filter, or sensors mounted inside a vented enclosure that breathes but blocks splashing, hold up far better than a bare module zip-tied to a wall. Self-heating is the quieter problem. A sensor on the same small board as a powered regulator, a radio, or a bright LED reads warmer than the air around it, sometimes by several degrees.
This is why an all-in-one development board with an onboard temperature sensor often disagrees with a thermometer sitting beside it. Where accuracy matters, run the sensor on a short cable away from the board, or put the radio to sleep between readings. Batteries deserve the same scepticism. Alkaline and lithium-ion capacity drops in cold conditions, and a device that runs all summer on a coin cell can fall over in an unheated outbuilding in January. Primary lithium cells perform noticeably better in cold than alkaline. Where a reading is safety-relevant, mains power with a battery fallback beats batteries alone.
Matching sensors to four common winter builds
The common thread is that every automated output deserves an independent sensor confirming it happened. A relay that clicks is not proof a heater is drawing power, and a servo that received a command is not proof a vent opened.
- Freeze protection for pipes: one or more DS18B20 probes strapped to the pipe under the insulation, plus a leak sensor on the floor below. Trigger heat trace or an alert on a pipe-surface threshold above freezing, not on room air.
- Greenhouse or grow tent control: SHT31 or BME280 for air temperature and humidity, a DS18B20 in the root zone, and a reed switch or limit switch confirming the vent actually moved. Humidity accuracy matters here because you are managing condensation and mould, not just warmth.
- Smart heating or zone control: one calibrated temperature sensor per room at seated height away from radiators and exterior walls, plus motion or door sensors to tell occupied rooms from empty ones. A current sensor on the boiler or heater circuit confirms the relay did what it was told.
- Winter garage or workshop automation: contact sensors on doors, a temperature sensor for a frost alarm, and a humidity reading if tools are rusting. Motion sensing is unreliable in a cold space because PIR works on thermal contrast and a heavily dressed person in a cold room reads differently than a person in a warm one.
Interfaces, wiring, and where projects actually break
Sensor choice is partly a wiring question. I2C parts like the SHT31 and BME280 are clean on a breadboard but are intended for short runs; long I2C cables are sensitive to capacitance and noise, and runs beyond a short distance need lower bus speeds, stronger pull-ups, or a differential extender. One-wire DS18B20 sensors tolerate much longer cable runs, which is why they dominate multi-room and outdoor builds despite needing careful pull-up resistor selection. Analogue sensors – thermistors, many soil moisture boards, light-dependent resistors – are cheap and are affected by supply voltage, ADC reference, and cable length.
They are fine for threshold work and frustrating when you want a trustworthy number. Resistive soil moisture probes in particular corrode; capacitive versions last considerably longer in continuous use. Two practical habits prevent most grief. Log raw readings for at least a few days before writing any control logic, so you learn what normal drift and noise look like in that specific location. And validate every new sensor against a second independent device – a separate probe or a hardware store thermometer – before you let it control anything with a heating element behind it.
Frequently Asked Questions
Is the DHT22 good enough, or should I spend more?
It is adequate for coarse alerts where a degree or two does not change the outcome. For greenhouse humidity control, heating setpoints, or anything you will tune over a season, an SHT31 or SHT41 gives tighter, more stable readings for a modest price increase.
Can I leave a sensor outdoors all winter?
Only in an enclosure that keeps liquid water and direct sun off it while still allowing air exchange. A sealed box reads its own internal microclimate; an open board collects condensation and eventually fails.
How many temperature sensors do I need for a whole house?
One per zone you intend to control independently, plus one outdoor reference. Extra sensors in rooms you cannot control separately produce data, not better behaviour.
Do I need a humidity sensor if I only care about heating?
Not for control, but it is worth adding in any space where condensation, mould, or rust is possible. Humidity data is what tells you whether a cold room is merely cold or actively damaging what is stored in it.
