An indoor ultra-wideband (UWB) location system combines tags, fixed reference devices called anchors, and software that estimates positions. A useful procurement requirement describes where, when and how reliably a position must reach the application. A single accuracy number leaves too much of that task unspecified.
Start with a floor plan, the objects and routes to be tracked, the intended business decision and an independent way to establish reference positions. This guide concentrates on UWB implementation and acceptance testing; the broader choice between last-seen, zone and coordinate information belongs in the location-evidence guide.
Choose the architecture before counting anchors
Two-way ranging exchanges timed messages between devices. Time-difference-of-arrival systems compare reception timing across anchors and require a suitable synchronization design. Angle-of-arrival systems use antenna arrangements to estimate direction. Qorvo’s UWB portfolio illustrates these different architectures; support varies by device and software package.
Ask the supplier to mark tag, anchor, synchronization, network and location-engine responsibilities on one drawing. Record which component calculates coordinates and which delivers them to your application. A range measurement between two devices is an input to positioning. Anchor coordinates, geometry and the chosen solver determine how those inputs become a useful position.
Require a proposed layout for the actual floor, including anchor heights, installation access, power and network connections. Test edge areas, corners and racks at working height. A successful center-of-room demonstration leaves these locations unresolved. Request the manufacturer’s calibration procedure for the exact hardware; Qorvo’s resource list includes separate DW1000 notes for antenna-delay calibration and ranging errors.
Write five separate acceptance requirements
On a small screen, scroll sideways to read all columns.
| Measure | Definition to agree | What to retain |
|---|---|---|
| Position error | Horizontal or 3D distance from independent truth, with units and a percentile or limit. | Coordinate frame, truth method and uncertainty. |
| Availability | Expected observation windows with a usable fresh fix. | Missing and rejected fixes, plus the denominator. |
| Latency | Time from the defined measurement event to application receipt. | Clock basis, measurement and receipt timestamps. |
| Update rate | Fresh delivered fixes per second for each tag. | Per-tag intervals at the intended tag population. |
| Battery service | Operating duration under a stated movement and reporting profile. | Battery type, settings, temperature and service assumptions. |
Choose thresholds from the task. Finding a cart in an aisle and guiding a precision operation require different evidence. Define a maximum age for a usable position and what the application shows after that age. Repeated delivery of the same old coordinate should remain distinguishable from a fresh measurement.
Count missing fixes in the result
The example CSV includes reference and estimated coordinates in metres, latency in milliseconds and a missing-fix row. Its data instructions define the fictional setup and reproduce the calculation. Horizontal error is the square root of the sum of squared x and y differences. Empty coordinates mean missing data. Keep them empty during analysis.
Ten invented observations demonstrate arithmetic, not statistical confidence or a product result. For a real pilot, agree the number and distribution of observations before testing. Report results by area and operating condition so a large easy area cannot hide a critical failure zone.
Run the site through its working conditions
- Survey reference points and anchor coordinates in the same frame; retain the survey method and uncertainty.
- Record hardware, firmware, antenna orientation, tag mounting, calibration and software settings before each run.
- Repeat static points with representative racks, loads and people present. Include obstructed routes and different tag orientations.
- Test movement using time-aligned independent truth. A delayed coordinate can be geometrically plausible but wrong for the current instant.
- Load the intended tag population and application pipeline. Measure delivered timing and missing fixes at the receiving application.
- Exercise an anchor or network outage through the supplier-supported test process and verify stale-position behavior.
When a route fails, keep the raw evidence, change one documented factor and repeat the affected conditions. Recheck adjacent areas after moving an anchor. Finish with an approved layout, configuration export, acceptance report and a process for retesting after racking, equipment or software changes.
Keep a working record
Download the editable working record (plain text). It includes purpose, instructions, evidence fields and this guide's source links. Save a separate copy for each evaluation and record unresolved issues with their owner.
Related reading
Researched and updated September 20, 2026. Recheck source documents when equipment, software or applicable requirements change.