How Do You Implement BLE Direction Finding (AoA and AoD)?

Detailed Explanation

Direction finding is a Bluetooth 5.1 feature that estimates the angle a BLE signal arrived from (or was sent toward) using an antenna array and phase-based signal processing, rather than the signal-strength-based estimation (RSSI ranging) BLE products relied on before. It's the feature behind indoor asset-tracking systems that report a tag's position to within a metre or so, and behind some wayfinding and proximity-beacon products. For the broader BLE version history this feature belongs to, see Bluetooth versions 3.0 through 6.0 compared; for BLE protocol fundamentals, see What Is Bluetooth Low Energy?

Direction finding is a link-layer feature, not a GATT-level one — it works on specially-flagged advertising or connection packets and requires hardware support in the BLE radio itself, not just a firmware or stack update. This is the detail that most catches teams out mid-project: a BLE SoC can be fully Bluetooth 5.x compliant for every other feature and still lack the specific radio hardware direction finding needs.

Angle of Arrival vs Angle of Departure

Both modes use the same underlying mechanism — a Constant Tone Extension combined with antenna switching and I/Q sampling — but differ in which side of the link carries the antenna array:

The choice between them is primarily a cost-and-complexity placement decision: whichever side of the link needs to stay cheap and simple gets the single antenna, and whichever side can absorb the array and switch hardware cost does the switching. Asset tracking puts the array on the fixed infrastructure (AoA) because tags need to be cheap and small in large numbers; a beacon reporting its position to ordinary phones puts the array on the beacon (AoD) because a phone's BLE radio has no antenna array and never will.

How CTE and IQ Sampling Work

The Constant Tone Extension is an unmodulated RF tone appended to the end of a specific advertising PDU (connectionless CTE, used for AoA/AoD beacons) or a Link Layer control PDU (connection CTE, used within an active connection). Because the tone carries no data, the receiver can sample it purely to characterise the RF signal's phase, unaffected by the data modulation that would otherwise complicate phase measurement.

During the CTE, the antenna array side switches between its elements at fixed intervals (1 µs or 2 µs slots, depending on the configured sampling mode) while the radio samples the tone's in-phase (I) and quadrature (Q) components at each switch position. Because each antenna element is at a slightly different physical position, the same incoming (or outgoing) wavefront arrives at — or departs from — each element with a slightly different phase, and that phase difference is what encodes the angle.

The underlying relationship for a simple two-element pair is the standard phase-interferometry expression: the phase difference between two elements spaced a known distance apart is proportional to the sine of the angle of arrival, scaled by the wavelength. In practice, production direction-finding implementations use more than two elements and combine multiple phase measurements (rather than a single two-antenna phase difference) to resolve angle more robustly against multipath reflections and measurement noise — the two-element relationship is the concept to understand, not the literal algorithm a real locator runs.

Antenna Array Hardware Requirements

The array itself is a set of antenna elements in a known, fixed geometric arrangement, connected to the radio through a fast RF switch:

• Element spacing must be kept below half the RF wavelength to avoid angle ambiguity (spatial aliasing, sometimes called grating lobes) — at 2.4 GHz, half-wavelength is approximately 6.2 cm, so a uniform linear array's elements are typically spaced somewhat closer than that. Spacing wider than half-wavelength can produce more than one angle solution that fits the same measured phase difference, which the algorithm cannot distinguish between.
• Array geometry determines what the system can resolve: a uniform linear array (elements in a single row) resolves azimuth angle only, in one plane; a two-dimensional array (a rectangular or U-shaped arrangement) resolves both azimuth and elevation, at the cost of more elements and a more complex switch.
• RF switching speed and timing must hit the CTE's 1 µs or 2 µs sampling slots precisely, which is why switching is handled by dedicated radio hardware (via PPI/DPPI-style hardware event routing on Nordic parts — see PPI/DPPI on the nRF52 and nRF53) toggling a fast RF switch IC, rather than by firmware bit-banging a GPIO, which cannot reliably meet the required timing.
• Phase-matched RF paths from the switch to every antenna element are required — an RF trace length mismatch between elements introduces a phase offset that reads as a false angle, which is why array PCB layout needs the same trace-length discipline used for controlled-impedance or differential routing elsewhere in high-speed PCB design.

Nordic Silicon Support and Firmware

Nordic's nRF52833 and nRF52811 include the Radio peripheral's Direction Finding hardware extension — CTE generation and reception, and hardware-timed antenna switching — as does the nRF5340's network core. Despite otherwise being capable, widely-used Bluetooth 5.x parts, the nRF52832 and nRF52840 do not include this hardware; a team that has already committed to either of those parts for a product and later adds a direction-finding requirement will need to change silicon, not just firmware, which is worth checking against the current nRF52833 Product Specification at the start of any direction-finding project rather than discovering it mid-development.

In the nRF Connect SDK, direction finding is enabled via CONFIG_BT_DF and its connectionless/connection-specific sub-options, and Nordic ships sample applications (direction_finding_connectionless_tx, direction_finding_connectionless_rx, and connection-based equivalents) that configure the CTE parameters, antenna switch pattern, and IQ sample reporting. See setting up the nRF Connect SDK and Zephyr RTOS for the base toolchain these samples build on, and BLE peripheral with Zephyr GATT on nRF52 for the surrounding BLE peripheral role these samples extend with direction-finding-specific configuration.

Design Considerations

• Confirm hardware DF support before finalising silicon — this is a radio hardware capability, not a firmware feature, and the specific nRF52 variant matters (see above).
• Decide AoA vs AoD from the cost distribution of your deployment, not accuracy alone — both modes achieve comparable angle accuracy for a comparable array; the real decision driver is which side of the link can economically host the array.
• Budget PCB area and layout complexity for the array early — a phase-matched multi-element array with a fast RF switch is a materially more involved RF layout than a single-antenna BLE design, and needs to be planned into the enclosure and board outline from the start rather than retrofitted.
• Plan a calibration step for production units — RF path phase offsets vary slightly between manufactured units, and most practical direction-finding systems apply a per-unit or per-batch calibration correction to the raw IQ data rather than assuming ideal, identical hardware across every unit. Zeus Design develops BLE direction-finding firmware and the antenna-array hardware it depends on for asset-tracking and positioning products.

Common Mistakes

• Choosing a direction-finding-incapable nRF52 variant (nRF52832 or nRF52840) based on general BLE 5.x feature marketing, without checking the specific part's Direction Finding hardware support.
• Underestimating the RF layout discipline an antenna array requires — treating it as "just more antennas" rather than a phase-matched array where trace-length mismatches directly corrupt the angle measurement.
• Spacing array elements wider than half-wavelength to physically fit a mechanical constraint, introducing angle ambiguity that isn't caught until angle accuracy is unexpectedly poor or inconsistent in testing.
• Skipping per-unit calibration and assuming every manufactured locator or tag has identical RF phase characteristics, then seeing accuracy that varies unexpectedly between physically identical units.
• Choosing BLE direction finding for a use case that actually needs UWB-grade accuracy — see what is Ultra-Wideband (UWB) for when centimetre-scale positioning genuinely justifies UWB's added cost and power draw instead.