Ultra-wideband ranging estimates the separation between compatible radios by timing their signal exchange. To judge a UWB application, ask which measurement it returns, under what conditions it works, and what happens when a usable measurement is unavailable.
A distance estimate can help find a tagged object or establish proximity. A direction arrow or position on a map requires additional capabilities and geometry. Start with the result your application actually needs before comparing accuracy claims.
Understand the measurement
Radio propagation time is extremely short: light travels approximately 30 centimeters in a nanosecond. Practical ranging protocols coordinate transmissions and responses so that propagation delay can be distinguished from device processing delay. UWB's wide bandwidth helps separate signal paths in time. The FiRa secure-ranging white paper, August 2022, explains time-of-flight ranging and the security mechanisms needed when a distance estimate is used to make an access decision.
Secure ranging is an implementation property involving the protocol and its protections. A product still needs suitable credential handling, authorization and application behavior. For example, a proximity result can be one input to a decision; the application must define how it handles stale, missing or low-confidence results.
Scroll the table horizontally; keyboard users can focus it and use the arrow keys.
| Result | What it tells you | Evidence to request |
|---|---|---|
| Distance | Estimated separation between two ranging endpoints. | Units, error definition, operating distance and obstructed-path behavior. |
| Direction | An estimated bearing relative to the receiving device. | Angle support on the exact hardware, orientation conventions and usable field of view. |
| Position | Coordinates in a defined frame, derived from suitable measurements and geometry. | Anchor layout or other positioning inputs, calibration and coordinate reference. |
| Proximity event | The application's interpretation of a measurement crossing a threshold. | Freshness limit, uncertainty handling and behavior near the boundary. |
Check the entire pair of endpoints
Record both device models, their software versions, the application and the supported ranging profile. Radios with the same broad technology label can expose different features. Check whether the intended initiator/responder roles work together and whether the application receives distance alone or also angle information.
Android's UWB developer guide describes session setup, capability checks and out-of-band exchange of ranging parameters. That surrounding exchange can use a separate channel such as Bluetooth. Platform support, permissions and device capabilities belong in the compatibility test alongside the UWB hardware. Follow the current documentation for the exact Android deployment rather than assuming every Android phone supports ranging.
Check the manufacturer's supported countries and operating conditions for both endpoints. UWB spectrum use depends on the jurisdiction and application; FiRa's regulatory guide provides context for those differences. Use current local requirements and product documentation for an actual deployment.
Read an accuracy claim as a test description
Ask how error was calculated: mean absolute error, a percentile, maximum observed error, or another measure. Then ask how many trials failed to return a value. A distribution computed only from successful measurements can hide long periods of unavailability.
Body blockage, metal furniture, antenna orientation and reflected paths are useful test conditions because they can change which signal path the receiver resolves. Compare a clear line of sight with the actual pocket, bag, wall or room arrangement the reader will use. Keep endpoint positions and orientation documented.
Fictional interpretation exercise: at a reference separation of 3.00 m, five reported distances are 3.02, 3.05, 2.98, 3.10 and 3.35 m. Their absolute errors are 0.02, 0.05, 0.02, 0.10 and 0.35 m. The mean absolute error is 0.54 รท 5 = 0.108 m; the largest observed error is 0.35 m. If these five readings came from six attempts, the return rate was 5/6, about 83%. These invented observations illustrate reporting; they establish no product specification or reliable percentile.
Run a trial that answers your application question
- Define a reference distance or marked position and the application's acceptable error and delay.
- Collect repeated attempts at several distances and orientations, recording missing results as well as returned values.
- Repeat the same sequence with representative obstructions and motion.
- Record device/software versions, update interval, result age and how the application displayed uncertainty.
- Test near any action threshold. Confirm that small fluctuations and stale data produce the intended behavior.
For an object finder, useful success might be a person reaching the correct shelf using the display. For a position system, it might be the required share of estimates falling within a defined area with acceptable delay. Choose the test from that outcome; an attractive single distance reading is only one observation.