Best IoT Sensors by Category: Temperature, Humidity, Vibration, Air Quality

Sensors are deceptively simple — they look like a five-dollar BOM line, until you discover the cheap ones drift 2°C in six months and the calibration paperwork delays your production run by a quarter. Here is what we actually pick on real projects.

Quick orientation

Temperature ranges from PT100/PT1000 RTDs (±0.1°C, defensible in regulated settings) to Sensirion STS3x/SHT4x for low-power embedded designs. Best when accuracy class is picked at least 2× tighter than the application needs, since real-world drift eats the margin.

Humidity accuracy depends on environment — Vaisala for traceable industrial calibration, Sensirion SHT4x for building/consumer use. Best when the design accounts for condensing conditions up front; cheap sensors saturate above 95%RH and don’t recover cleanly.

Vibration sensing comes down to noise density (µg/√Hz) — MEMS accelerometers like the ADXL355 for sub-1kHz monitoring, IEPE piezoelectric sensors for high-frequency bearing-fault detection. Best when the sensor is matched to the fault frequency being hunted, not just picked for “more accurate.”

Air quality sensors measure one specific gas, not a generic quality score — NDIR CO₂, laser PM2.5/PM10, or VOC-index gas resistance. Best when the marketing claim matches the actual measurement.

Current & power sensing spans Hall-effect ICs for submetering to isolated ΔΣ ADCs for in-drive industrial sensing. Best when accuracy class matches what the reading has to defend — billing, safety, or just trend monitoring.

Motion/orientation defaults to a 6-axis IMU for wearables and gesture products, with a magnetometer added only when true heading is required.

Category	Go-to pick	Key spec	Watch out for
Temperature	PT100/PT1000 (industrial), Sensirion STS3x/SHT4x (embedded)	Accuracy class vs. real-world drift	Spec accuracy is lab-only; budget 2x margin
Humidity	Vaisala HMP (industrial), Sensirion SHT4x (consumer/building)	Drift (%RH/year)	Saturation above 95%RH on cheap sensors
Vibration	ADXL355/357 (sub-1kHz), IEPE piezo (>5kHz)	Noise density (µg/√Hz)	Wrong sensor for the fault frequency you’re hunting
Air quality	SCD4x (CO2), SPS30 (PM2.5/10), BME688 (VOC)	Which gas is actually measured	”Air quality” claims without a named gas
Current & power	ACS770/ACS37800 (general), isolated ΔΣ ADC (industrial)	Accuracy class	Using submetering-grade parts for billing-grade claims
Motion/orientation	ICM-42688/BMI270 (IMU) + magnetometer for heading	Noise/power for wearables	Needing true heading without adding a magnetometer

Principles before picks

Three rules we apply to every sensor choice before we look at vendor part numbers:

1. Accuracy class is a contract, not a marketing number. “±0.5°C accuracy” means in lab conditions, at one operating point, after factory calibration. Real-world drift over a year often dwarfs the spec. Pick at least 2× tighter than your application requires.
2. Calibration drift compounds. If a humidity sensor drifts 0.5%RH/year, a 5-year deployment is reading nonsense by year three. Plan for re-calibration, sensor replacement, or design-in field calibration.
3. Sample rate matters more than absolute accuracy for many use cases — vibration, current, motion. A noisy fast sensor outperforms a clean slow one for fault detection.

Temperature

Industrial / process

• PT100 / PT1000 RTD — when accuracy needs to defend itself in court. ±0.1°C achievable, 30+ year deployments common.
• Type K thermocouple — high-temperature applications (kilns, exhaust). Cheap, robust, but cold-junction compensation is a real engineering job.

Embedded / consumer

• Sensirion STS3x / SHT4x — ±0.1°C, I²C, low power. Default for indoor temperature/humidity products.
• TI TMP117 — ±0.1°C medical-grade in a 2×2 mm package. Pricier; worth it for healthcare wearables.
• Bosch BME280 / BME688 — combined T/H/P (and gas in 688). Convenient for environmental products; not the most accurate option but trades accuracy for integration.

Humidity

The accuracy you need depends on the application:

• Industrial process control — Vaisala HMP series (±1.5%RH or better). Includes traceable calibration. Expensive but the deployment outlives the engineer.
• Building / HVAC / consumer — Sensirion SHT4x (±1.8%RH typical, much better than older Honeywell HIH series).
• Outdoor / agricultural — sensors with a hydrophobic membrane and replaceable filter. Drift is measured in 1%RH/year for the better units.

The trap: cheap humidity sensors saturate at >95%RH and never recover correctly. If your deployment sees condensing conditions, design for the saturation event — heater pulse, hydrophobic filter, or a different sensor technology.

Vibration

For predictive maintenance (see our IIoT post):

• Analog Devices ADXL355 / ADXL357 — high-resolution MEMS accelerometer, low noise, ±2g to ±40g. Default for sub-1kHz vibration monitoring on rotating equipment.
• TDK ICM-42688 — wider bandwidth, lower power, smaller package. Right call for wearable or handheld vibration monitors.
• Piezoelectric IEPE accelerometer — for high-frequency analysis (bearing fault detection above 5kHz). Pair with a 24-bit ADC and a real signal-conditioning chain.

The number that matters is noise density (µg/√Hz). Below 25 µg/√Hz catches early-stage bearing wear; above 100 catches only catastrophic faults.

Air quality

A growing category as IAQ regulation tightens:

• Sensirion SCD4x — true NDIR CO₂ at the price of a PCB sensor. Best fit for residential and small commercial.
• Sensirion SPS30 — laser-based PM2.5/PM10. The mass-flow design beats the cheap optical PM sensors.
• Bosch BME688 — gas resistance for VOC trends. Useful as a relative indicator, not as an absolute concentration measurement.
• Senseair Sunrise — single-channel NDIR CO₂ with self-calibration baseline. Good for outdoor or unconditioned spaces.

If your product claims to measure “air quality” without specifying which gas, the sensor is probably a VOC index, not an absolute reading. Be explicit in your marketing — regulators increasingly are.

Current and power

For energy submetering and condition monitoring:

• Allegro ACS770 / ACS37800 — Hall-effect, high-current, solid-state. The ACS37800 integrates a 24-bit ΔΣ ADC for true RMS measurement.
• Continental Control CTs + Wattnode — when you need utility-grade accuracy class 0.5 or better. Pricier, but defensible in court.
• Texas Instruments AMC1304 isolated ΔΣ — for inside-the-drive current sensing on industrial machinery. Galvanic isolation, high CMRR, MHz bandwidth.

Motion / orientation

• TDK ICM-42688 / ICM-45686 — 6-axis IMU, low noise, low power. The default for wearables and gesture-controlled products.
• Bosch BMI270 — 6-axis with built-in gesture recognition, ultra low power. Specifically tuned for fitness wearables.
• Asahi Kasei AK09918 — 3-axis magnetometer when you need true heading; pair with an IMU for sensor fusion.

What we hand over

For every sensor decision on a real project we deliver a sensor specification document that includes:

• Required accuracy class with derivation
• Drift budget over deployment lifetime
• Calibration method (factory, field, none)
• Failure modes and mitigation
• Replacement plan and life-cycle cost

That document is the deliverable, not the part number. The part number is the easy part.

If you have a sensor decision wrestling with accuracy class versus BOM versus drift, we will help you scope it.