DHT22 vs BME280 vs SHT31: Picking the Right Temperature and Humidity Sensor

Every weather station, greenhouse monitor, and room logger starts with the same decision. All three of these sensors will give you a temperature and a humidity number. What separates them is how accurate that number is, how much work it takes to read it, and how they behave after six months in a real environment.

The short answer

• DHT22 — cheapest, easiest to find, worst accuracy and slowest. Fine for "is the greenhouse roughly 25 °C or roughly 35 °C".
• BME280 — the default choice for most projects. Good accuracy, I2C or SPI, and it throws in a barometric pressure reading for free.
• SHT31 — the accuracy pick. Best humidity performance of the three, with an onboard heater for condensing environments. Costs more.

If you are building a weather station and cannot decide, buy the BME280. The pressure channel alone makes it worth it, and the accuracy is good enough for nearly everything.

Specifications side by side

Specification	DHT22 (AM2302)	BME280	SHT31-D
Temperature accuracy	±0.5 °C	±1.0 °C (±0.5 °C typical, 25 °C)	±0.2 °C
Humidity accuracy	±2–5 % RH	±3 % RH	±2 % RH
Pressure	No	Yes, ±1 hPa	No
Interface	Single-wire proprietary	I2C or SPI	I2C
Max sample rate	0.5 Hz (one reading / 2 s)	Up to 157 Hz	Up to 10 Hz
Supply voltage	3.3–6 V	1.7–3.6 V	2.4–5.5 V
Typical current	~1.5 mA active	~3.6 µA @ 1 Hz	~1.5 mA active, 0.2 µA idle
Onboard heater	No	No	Yes
I2C addresses	N/A	0x76 / 0x77	0x44 / 0x45

DHT22: cheap, and you get what you pay for

The DHT22 is the sensor most tutorials use, largely for historical reasons. It works, and for coarse measurements it is fine.

The problems are real though:

The protocol is proprietary and timing-sensitive. It is not I2C, not SPI, not UART — it is a custom single-wire scheme with microsecond-level timing requirements. On an ESP32 running FreeRTOS and a WiFi stack, an interrupt at the wrong moment corrupts the read. You will see intermittent nan readings. Libraries handle retries, but the underlying fragility does not go away.

It is slow. One reading every two seconds, maximum. That rules out anything responsive.

Humidity drift is significant. DHT22 units commonly drift several percent RH over a year, and there is no calibration path. Two DHT22s side by side frequently disagree by 3–4 % RH out of the box.

Buy a DHT22 when: budget is the dominant constraint, you need one reading every few minutes, and ±5 % humidity is acceptable.

BME280: the sensible default

Bosch's BME280 is the sensor we recommend most often, and it is what our ESP32 weather station build uses.

Standard I2C. Two wires, a well-supported library, no timing games. Runs happily alongside an OLED on the same bus — see the I2C communication guide if you are new to sharing a bus.

Pressure for free. ±1 hPa barometric pressure enables altitude estimation and short-term weather trend prediction. Neither of the other two sensors offers this.

Genuinely low power. In forced mode, sampling once per second, it draws around 3.6 µA. Paired with ESP32 deep sleep, a battery-powered node can run for months.

Configurable oversampling and filtering. You control the noise/power/speed tradeoff in software rather than accepting a fixed behaviour.

The self-heating gotcha

The most common BME280 complaint is that temperature reads 1–2 °C high. This is almost always self-heating, not a faulty sensor.

Causes and fixes:

• Continuous ("normal") mode at high sample rates warms the die. Use forced mode: take one reading, return to sleep. This alone usually fixes it.
• Mounting on a board next to a warm regulator or an ESP32 conducts heat in. Physically separate the sensor, or use a breakout on a short cable.
• No airflow in a sealed enclosure. Ventilate the housing.

The address gotcha

BME280 breakouts ship with the address strapped to either 0x76 or 0x77 depending on the vendor. If begin() fails, try the other one before assuming the board is dead. Run an I2C scanner first.

BMP280 is not BME280

Watch the part number when ordering. The BMP280 has no humidity sensor — only temperature and pressure. They look nearly identical and are often mislabelled by cheap sellers. The BME280 is a 8-pin package; genuine Bosch parts are laser-marked. If humidity is why you bought it, verify the marking.

Buy a BME280 when: you want a good general-purpose environmental sensor, especially for weather, altitude, or battery-powered logging.

SHT31: when the humidity number has to be right

Sensirion's SHT31-D is the accuracy option. ±2 % RH and ±0.2 °C, with much better long-term stability than either alternative.

The onboard heater is the killer feature. In high-humidity or condensing environments — greenhouses, bathrooms, outdoor enclosures, cold rooms — moisture condenses on the sensing element and readings saturate at 100 % RH until it dries. The SHT31 can pulse an integrated heater to drive that moisture off, then resume measuring. Neither the DHT22 nor the BME280 can recover from condensation on their own.

Wide supply range. 2.4–5.5 V, so it works directly on both 3.3 V and 5 V systems without level shifting — one of the few sensors where that is true.

Fast and repeatable. Up to 10 Hz, with selectable repeatability modes trading speed against noise.

The cost is price and the lack of a pressure channel.

Buy an SHT31 when: humidity accuracy actually matters — curing, storage, HVAC, agriculture — or when condensation is likely.

Which to pick, by project

Project	Recommended sensor	Why
Indoor room logger	BME280	Accuracy is plenty, pressure is a bonus
Outdoor weather station	BME280	Pressure enables trend prediction
Greenhouse / mushroom tent	SHT31	Condensation recovery via heater
Battery sensor node, months of runtime	BME280	Lowest sleep current, forced mode
Altitude / vertical position	BME280	Only one with a barometer
Cold storage monitoring	SHT31	Accuracy plus heater at low temperature
Learning project, tight budget	DHT22	Cheapest path to a working reading
Fast-responding control loop	SHT31	10 Hz vs 0.5 Hz on DHT22

Wiring notes that save an afternoon

All three on an ESP32 are 3.3 V devices in practice. The DHT22 and SHT31 tolerate 5 V supply, but if you power them from 5 V their data lines will output 5 V logic, which the ESP32 does not tolerate. Power them from 3.3 V.

I2C pull-ups. Most BME280 and SHT31 breakouts include 4.7 kΩ pull-ups on SDA and SCL. Stack three or four such breakouts on one bus and the parallel pull-ups get too strong, and the bus stops working. Remove the pull-ups on all but one board.

Cable length. I2C is designed for on-board distances. Beyond about 50 cm you will see errors. For a sensor 10 m away, either drop the bus speed substantially, use an I2C extender, or put a second microcontroller at the sensor and send readings over MQTT or LoRa.

Calibration reality check

None of these are laboratory instruments. If your readings need to be traceable, you need a reference and an offset table.

A practical field check: seal the sensor in a small airtight container with a saturated salt solution. Sodium chloride gives 75 % RH at 25 °C; magnesium chloride gives 33 %. Let it stabilise for several hours, then record the offset. This costs nothing and tells you far more than a datasheet tolerance figure.

Summary

The DHT22 is a teaching sensor. The BME280 is the right answer for most real projects. The SHT31 is what you buy when the humidity reading is the point of the project rather than a nice-to-have.

All three, along with the ESP32 boards and OLED displays to pair them with, are stocked at Wavtron with datasheets, GST invoicing, and same-day dispatch across India.