Skip to content
Documentation Background

Bluetooth, satellite, IoT, and positioning

Wi-Fi and cellular cover most traffic. Other radio systems fill gaps that those cannot: very short range at very low power, places with no ground infrastructure, and devices that must run for years on a battery. This page covers them, plus the positioning systems phones use to find themselves.


Bluetooth (IEEE 802.15.1) is a personal area network technology: tens of meters, low power, low cost. It is a cable replacement for keyboards, headsets, wearables, and car audio. It works without any infrastructure, so it is an ad hoc network.

Bluetooth shares the crowded 2.4 GHz unlicensed band with Wi-Fi, Zigbee, and microwave ovens. It copes with frequency hopping spread spectrum (FHSS):

  • The band holds 79 one-megahertz channels, and time is divided into 625 microsecond slots, which is 1,600 slots per second.
  • In each slot, the transmitter uses a different channel, following a pseudo-random sequence shared by the network.
  • Interference on one frequency therefore hits only a small share of slots. Adaptive hopping removes channels with persistent interference from the sequence.
  • Different piconets use different hopping patterns, so many can share a room.

Classic Bluetooth reaches roughly 2 to 3 Mbit/s.

A piconet has one central device and up to seven active peripherals. The central controls everything: it sets the clock and hopping sequence, admits devices, sets the peripherals’ transmit power, and grants transmission turns by polling, as described in multiple access protocols. Peripherals do not talk to each other directly.

  1. Inquiry. The central broadcasts inquiry messages across 32 channels. A listening device answers with its ID after a random delay of up to 0.3 seconds, so replies do not collide.
  2. Paging. The central pages the chosen device on its own frequencies. The device acknowledges, and the central sends the hopping pattern, clock, and an active-member address, then starts polling.

A piconet is one hop, so Bluetooth has no network or transport layer. A layer called L2CAP sits above the link layer. It multiplexes applications, segments and reassembles data, offers reliable and unreliable channels, and can reserve periodic slots for streaming audio.

BLE targets sensors and wearables. A device sleeps by default and wakes to advertise or report, then sleeps again. See the energy models below. Pairing security is a recurring weak point, so treat BLE devices as untrusted on sensitive networks and see Wireless security for the wider picture.

Terminal window
# Linux: scan for nearby devices and inspect one
bluetoothctl
# [bluetooth]# scan on
# [NEW] Device AA:BB:CC:DD:EE:01 Headset
# [bluetooth]# info AA:BB:CC:DD:EE:01

Geostationary (GEO)Low Earth orbit (LEO)
AltitudeAbout 35,000 km500 to 1,200 km
PositionFixed over the equatorMoving, crosses the sky in minutes
Round tripRoughly 500 to 800 msRoughly 30 ms
Coverage per satelliteA third of the planetA small footprint
Constellation sizeA handfulHundreds to thousands

LEO constellations now account for most new launches because the latency is close to a terrestrial path. A rough physical minimum for each orbit is derived in the wireless link math deep dive.

  • Each satellite covers a footprint on the ground. Its downlink is a broadcast to all terminals in the footprint, and the uplink from terminals is a shared channel that needs a multiple access scheme.
  • Ground stations include user terminals (fixed dishes, or ones on ships, aircraft, and vehicles) and gateways that connect to the terrestrial internet.
  • Bent pipe. The satellite simply relays: traffic goes up from a terminal, is reflected down to a gateway, and enters the internet there. It needs a gateway within view.
  • Network in the sky. Satellites connect to each other with inter-satellite links, usually optical. Packets can cross several hops in orbit before coming down near the destination. The topology changes constantly, so routing must adapt continuously.
  • The infrastructure moves. On a cellular network the user moves past fixed towers. Here the towers move at tens of thousands of kilometers per hour. A terminal stays under one satellite for only minutes, so handover between satellites is continuous and the network keeps changing paths.
  • Latency changes with each handover. Applications see jitter, and RTT-sensitive TCP behaves worse than on a steady path. Satellite operators sometimes use performance-enhancing proxies, which break end-to-end semantics.
  • No single standards body covers LEO the way 3GPP covers cellular, so implementations differ by operator.

For GEO links, the long RTT hurts interactive protocols and slow start, so larger initial windows and window scaling matter. See TCP congestion control.


IoT devices are small and often battery-powered with a life of years. No radio maximizes range, battery life, and data rate at once, so the technologies fall into three families:

FamilyEmphasisTechnologiesRange
Short-range, low-powerLow energy, moderate rateWi-Fi HaLow (802.11ah), BLE, ZigbeeTens of meters to about 1 km
Low-power wide-area (LPWAN)Coverage and battery life, low rateLoRaWAN, NB-IoTKilometers
Cellular IoTHigher rate, mobility, reliabilityLTE-M, 5G for massive machine communications and ultra-reliable low latencyCellular coverage
  • Wi-Fi HaLow (802.11ah) uses the sub-1 GHz band with stripped-down headers to reach a few hundred meters up to 1 km, at hundreds of kbit/s to several Mbit/s, and supports target wake time.
  • Zigbee builds on the IEEE 802.15.4 physical and MAC layers for low-rate personal networks, often in mesh form (home automation, lighting).
  • LoRaWAN works in unlicensed sub-1 GHz bands (915 MHz in the Americas) at a few hundred bit/s to tens of kbit/s per channel. Gateways in a star-of-stars topology relay to a network server.
  • NB-IoT is a cellular standard using a 200 kHz channel in licensed spectrum, about 250 kbit/s at most.
  • LTE-M has a wider channel than NB-IoT and supports handover and mobility, for things like asset trackers.
  • 5G mMTC targets very dense deployments (up to about a million devices per square kilometer). URLLC targets industrial control and similar jobs needing very low latency and high reliability.

Communication takes a large share of a device’s energy, often 20 to over 50 percent. A cellular radio can draw over a watt when active and a few milliwatts when asleep, so every IoT design is about sleeping as much as possible. There are two models.

Coordinated sleep and wakeUncoordinated wake-and-send
HowDevice and base station agree on a schedule. The base station buffers downlink dataThe device sleeps until it has data, then sends and briefly listens for a reply
ExamplesWi-Fi power save and target wake time, 5G DRXLoRaWAN Class A, BLE, NB-IoT
StrengthThe network can reach the device on a scheduleSimple, very low energy
CostWaking at agreed timesThe network cannot send to the device until the device next transmits, so downlink latency is high

Design consequence: for uncoordinated devices, commands from the cloud sit in a queue until the device’s next uplink. Design the control plane for that. An operation that needs an answer in a second cannot target a LoRaWAN Class A sensor.

In a typical deployment the gateway or base station forwards data over IP to a cloud ingestion service, often using a lightweight protocol such as MQTT over TLS. The radio choice decides battery life and maintenance costs more than the cloud side does.


Location services combine two systems.

A constellation of satellites with atomic clocks continuously broadcasts its position and the exact time. A receiver measures how long each signal took to arrive, which gives its distance to that satellite. With four satellites in view it can solve for latitude, longitude, altitude, and its own clock error (trilateration). Three would suffice if the receiver’s clock were perfect, but the fourth fixes the clock error that a cheap receiver clock introduces.

GPS needs a line of sight to the sky. Indoors, and in dense city streets where signals are blocked or reflected, it degrades or fails.

When GPS is weak, phones use the access points around them:

  1. The phone scans and records the beacon signal strength of each visible AP, identified by SSID and MAC.
  2. It sends that list to a cloud location service, which has a database of AP locations.
  3. The service estimates distance from each signal strength and trilaterates a position, combining it with any GPS data.

The AP database is itself built by crowdsourcing. Phones with a good GPS fix report the APs they hear along with their coordinates, and the service solves the same equations in reverse to place each AP. APs help phones find themselves, and phones help the service map APs.

GPSWi-Fi positioning
InfrastructureSatellites with atomic clocksMillions of access points
MethodTime of flight, trilaterationSignal strength, trilateration
Best inOpen skyCities and indoors
Depends onNothing but the receiverA cloud database