Chris Collins

Tools · radio

IoT wireless chooser

Which radio for your IoT device? A comparison of fifteen technologies, from Wi-Fi, BLE, Zigbee, Thread/Matter and Z-Wave to LoRaWAN, Sigfox, NB-IoT, LTE-M and sub-GHz FSK, with typical range, data rate, latency, currents, cost and rules. It ranks them for your answers, works out a link budget and range per technology (free space, walls, Okumura-Hata) and the battery life for your message, and explains the protocol stacks, Matter commissioning and mesh against star.

Added

Works with

How to use it

Picking the radio for an IoT device is a trade between range, battery, data, latency, the hub you already own and what it costs over the years. This puts the usual candidates side by side, ranks them for your answers, and lets you check the two numbers that decide most projects: how far the link reaches and how long the battery lasts.

How to use it

  1. Rules — choose Europe (ETSI, 868 MHz) or North America (FCC, 915 MHz). It sets the sub-GHz frequencies, the LoRaWAN data rates, the TX power defaults and the duty-cycle checks.

  2. Compare — the matrix of all fifteen technologies. Switch the columns between Radio (band, range indoors and out, data rate, latency, topology), Power (typical module, TX, RX and sleep currents, and the battery life for 20 bytes every 15 minutes on two AA cells), Cost (module, gateway, subscription, and the cost per device over five years) and Ecosystem & rules (ecosystem, hub support, regulation). Filter by family, and click a row for its details, its PHY presets with their sensitivities, the sources and links to the deeper tools.

  3. Recommend — answer the questions: distance to the hub or gateway, indoors or out, message size and interval, battery or mains, whether the device receives commands, the latency you need, your hub (Home Assistant, or Apple, Google or Amazon through Matter), the number of devices and the budget. Or start from an example: a farm sensor 10 km out, a smart plug for Apple Home, a coin-cell thermometer, a van tracker, a camera, a door lock, a basement water meter or a garden station. The ranked list shows every rule that added (+) or took away (–) points, and what ruled the others out (×).

  4. Range — choose a technology and PHY (data rate); the TX power, antenna gains and cable loss start at that technology’s typical values. Choose a path-loss model:

    • Free space: the Friis formula, the best case with a clear line of sight;

    • Indoor: log-distance, L = FSPL(1 m) + 10·n·log₁₀ d + (d / spacing)·Lwall, with a path-loss exponent (ITU-R P.1238 gives about 3 for offices, 2.8 for homes) and a wall every few metres of a type (plasterboard to reinforced concrete, frequency-dependent);

    • Outdoor: Okumura-Hata up to 1500 MHz and COST-231 Hata to 2000 MHz, for open, suburban, urban and metropolitan areas with the gateway’s and the device’s antenna heights.

      The fade margin is subtracted from the maximum path loss, and the range is where the model’s loss reaches what is left. The first chart shows the path loss against distance with the allowed loss; the second shows every technology’s range in the same environment.

  5. Battery — choose a technology and PHY, the message size, the interval and the battery. The tool builds one message’s activity state by state (for Wi-Fi: boot, association, DHCP, TLS, MQTT; for BLE: three advertising channels per event; for LoRaWAN: the airtime and the two receive windows; for cellular: RRC connection, data, release) and shows the battery life, the average and peak current, the charge per message and where it goes. The chart compares every technology for the same message and battery. Options: Wi-Fi reconnecting or staying associated, TLS and static IP; the poll interval of a Zigbee or Thread sleepy end device; BLE repeats; NB-IoT/LTE-M release assistance.

  6. Stacks — the protocol layers (PHY, MAC, network, transport, security, application) of Thread + Matter, Zigbee, BLE, LoRaWAN and Wi-Fi with MQTT/HTTP/CoAP, two side by side; click a layer for what it does. Matter commissioning steps through BLE discovery, PASE, device attestation, the operational certificate, the Thread or Wi-Fi credentials and CASE as a message sequence. Mesh vs star sends a packet hop by hop from any device to the hub; change the radio range, switch routers off and watch the mesh route around them.

Method and assumptions

  • Typical figures. Every current, sensitivity, range and price is typical, for orientation. Currents come from the datasheets of representative parts: Espressif ESP32-C3 and ESP32-C6 (Wi-Fi, ESP-NOW), Nordic nRF52840 (BLE, Thread), TI CC2652R (Zigbee), Silicon Labs EFR32ZG23 (Z-Wave), Semtech SX1262 (LoRa), HopeRF RFM69HCW and TI CC1101 (sub-GHz FSK), and nRF9160 / Quectel modules (cellular). The figures for HaLow, Wi-Fi 5 GHz, Sigfox modules, cellular modules and the sleep current of Wi-Fi 6 with TWT are approximate and marked so in the sources. Each technology’s details card names its source.

  • Link budget. Maximum path loss = TX power + antenna gains − cable loss − receiver sensitivity at the chosen data rate. LoRa sensitivities use the same estimate as the LoRa airtime tool (−174 dBm/Hz + 10·log₁₀ BW + 6 dB noise figure + the SF’s SNR limit). For NB-IoT, LTE-M and Cat-1 the sensitivity is the base station’s, derived from the 3GPP maximum coupling loss targets (164, 156 and about 141 dB with a 23 dBm device). The Okumura-Hata check: 900 MHz, 30 m base, 1.5 m mobile, 1 km in a medium city gives 126.4 dB.

  • Wall losses are rounded from NIST IR 6055 (Stone, 1997), ITU-R P.2040 and P.1238 at 900 MHz, 2.4 GHz and 5.8 GHz, interpolated in log-frequency. Real walls spread by half again either way.

  • Battery life uses the model of the battery-life estimator: each state’s current × duration over the cycle, sleep for the rest, plus self-discharge and usable capacity, with no regulator. The LoRaWAN airtime and receive windows come from the LoRa airtime tool. The durations of connection set-up (Wi-Fi association, DHCP, TLS; cellular RRC) are rules of thumb for typical firmware and networks: measure yours.

  • The recommender is a set of plain rules with visible points: the distance against the technology’s typical range (a mesh of mains routers counts several hops), whether the data fits (Sigfox’s 140 × 12 bytes a day, the EU 1 % duty cycle, the US 400 ms dwell time, throughput), the battery estimate, latency (a sleepy device receives commands at its poll), hub support, indoor/outdoor, the number of devices, cost per device over five years and network status. It picks a slow, robust PHY for long distances (LoRaWAN SF11–12 at 10 km, BLE Coded past 50 m).

  • Sigfox: Sigfox SA went into receivership in 2022 and was bought by UnaBiz. The network still runs in many countries, but coverage has been withdrawn in some; check before relying on it.

Limits

  • Real range depends on the terrain, the antennas' placement, interference and the receiver; the models give medians and the fade margin covers some of the spread. For a terrain profile use the Meshtastic planner or the propagation tool.

  • Regulations change and differ by country; the summaries here are for orientation, not compliance.

  • Prices are rough single-unit figures in US dollars (2026) and change quickly.

Privacy

Everything runs in your browser; settings are kept in this browser only.

Runs entirely in your browser; nothing you type leaves this page.