Uncategorized
Ultra-wideband is the most accurate radio positioning technology you can buy for a building, and it is also the easiest one to under-budget. The accuracy figure in a datasheet is a line-of-sight number. The quote you sign is an anchor-count number. Those two things are connected by the walls in your building and by an emission limit written into federal regulation, and almost nobody explains that connection before the purchase order goes out.
This article does. Every number below comes from a regulator, a standards body or a named measurement study, with the link next to it.
What is a UWB indoor positioning system?
A UWB indoor positioning system locates tags or phones inside a building by measuring the flight time of very short radio pulses between a mobile device and fixed anchors. Position comes from distance and geometry, not from signal strength.
The regulatory definition is precise. Under 47 CFR §15.503(d) (2024 edition of Part 15, Subpart F), an ultra-wideband transmitter is one that “at any point in time, has a fractional bandwidth equal to or greater than 0.20 or has a UWB bandwidth equal to or greater than 500 MHz, regardless of the fractional bandwidth.” That bandwidth is the whole trick. A wide pulse in frequency is a narrow pulse in time, and a narrow pulse in time can be timestamped sharply enough to resolve a distance to centimetres.
The relevant air interface is IEEE 802.15.4z-2020, the amendment titled Enhanced Ultra Wideband (UWB) Physical Layers (PHYs) and Associated Ranging Techniques, approved by the IEEE SA Standards Board on 4 June 2020 and published on 25 August 2020. It has since been folded into the 802.15.4-2024 revision. When a supplier says “UWB”, this is the standard worth asking them to name.
How does UWB measure distance?
It times a round trip. In two-way ranging, a tag sends a pulse, an anchor replies, and the tag divides the elapsed time by two and multiplies by the speed of light. No clock synchronisation between the two radios is required, which is why two-way ranging is the common choice for tag-to-anchor systems.
The alternative is time difference of arrival, where several time-synchronised anchors compare when the same pulse reached each of them. That moves the hard engineering problem from the tag to your infrastructure: the anchors now need a shared time base, usually over wired backhaul. Two-way ranging costs airtime and battery. Time difference of arrival costs cabling and synchronisation. Both are described in NIST’s ranging comparison by Li, Shi, Moayeri and Benson (paper dated 2019, hosted on nist.gov).
Most deployments land on one of two channels. Channel 5 is centred at 6489.6 MHz and channel 9 at 7987.2 MHz, both with 499.2 MHz of bandwidth, per the band allocation table in An Overview of UWB Standards and Organizations (arXiv:2202.02190, 2022). Channel 9 matters for anyone shipping to Europe, for reasons that appear further down.
How accurate is UWB indoor positioning in a real building?
Sub-decimetre in line of sight, and metres once a few walls are in the way. The NIST study above tested two commercial UWB ranging systems and one Wi-Fi Round Trip Time system outdoors, in an office building and in a single-family house, at true distances from 1 m to 100 m.
| Metric, all tests pooled | Better UWB system | Wi-Fi RTT |
|---|---|---|
| Mean absolute error | 0.27 m | 3.04 m |
| Root mean square error | 0.41 m | 4.96 m |
| Standard deviation | 0.33 m | 3.90 m |
| Four or five drywalls between radios | no range estimate produced | estimate in every case |
Those are ranging errors, not position errors, and the UWB hardware tested implemented IEEE 802.15.4a rather than the later 802.15.4z. Treat the figures as the shape of the result rather than as your building’s specification.
A second, larger measurement tells the same story at site scale. A preprint from TUD Dresden University of Technology, posted November 2025 and revised April 2026 (arXiv:2511.10526), instrumented a 44 by 30 m indoor hall with 12 nodes and reports “0.28 m mean ranging error” in line-of-sight conditions “versus 3.78 m for NLOS.” Same technology, same site, a factor of thirteen between the two conditions.
For contrast, read the public-safety requirement NIST publishes for first-responder location: “a 3D accuracy requirement of better than 3 meters radius (95% probability), without beacons”, on its First Responder Location and Mapping Services page. UWB in line of sight beats that by an order of magnitude. UWB behind four walls does not produce a reading at all.
Why do walls decide how many anchors you buy?
Because the pulse is legally very quiet. The same regulation that defines UWB also caps its power far below Wi-Fi or Bluetooth levels, so the link budget runs out after a small number of obstructions, and coverage has to be bought in anchors instead.
The NIST authors state the consequence directly in their abstract: UWB’s signal “can penetrate a small number of obstacles due to typical UWB emission level restrictions. Consequently, one needs to deploy a large number of UWB radios in any moderately-sized or larger building.” Their floor test is the blunt version. With one radio on the first floor of an office building, no system produced any range estimate to the third floor; the better UWB system reached 3 of 10 test locations on the second floor and in the stairwell, while Wi-Fi RTT reached all 10.
Their conclusion puts a number on a small building: UWB “can provide accurate indoor localization for wooden structure single-family houses with drywalls equipped with about a dozen UWB anchor nodes.”
Three practical consequences follow. Anchor count scales with partition count, not with floor area, so a 2,000 m² open warehouse is cheaper to cover than a 2,000 m² clinic with forty exam rooms. Construction material changes the answer, as concrete, metal stud and metal shelving are not drywall. And positioning needs more than one reading, since trilateration in two dimensions needs three anchors in range and height needs four, which means “one anchor still hears the tag” is not coverage.
What do regulators allow?
Different bands in different regions, at the same power ceiling. In the United States, 47 CFR §15.517 governs indoor UWB systems. In the European Union, Commission Implementing Decision (EU) 2019/785 of 14 May 2019 harmonises the conditions.
| Condition | United States, §15.517 | European Union, 2019/785 |
|---|---|---|
| Band the UWB bandwidth must sit in | 3,100 to 10,600 MHz | 6 to 8.5 GHz generic row |
| Mean limit, 1 MHz resolution bandwidth | −41.3 dBm EIRP | −41.3 dBm/MHz |
| Peak limit in a 50 MHz bandwidth | 0 dBm EIRP | 0 dBm peak e.i.r.p. |
| Instrument | 47 CFR Part 15, Subpart F | Decision (EU) 2019/785 |
Channel 9, at 7987.2 MHz, sits inside both of those bands. Channels below 6 GHz do not sit inside the EU generic row, which is the practical reason a multi-region rollout standardises on the upper channels.
Three US rules catch buyers out, and all three are in the regulation rather than in anyone’s datasheet. First, §15.517(a) limits that section to transmitters “employed solely for indoor operation”, and §15.517(a)(2) says emissions “shall not be intentionally directed outside of the building”, so an anchor aimed through a window at a loading yard is not a configuration choice.
Second, devices under that section must carry this statement: “This equipment may only be operated indoors. Operation outdoors is in violation of 47 U.S.C. 301 and could subject the operator to serious legal penalties.”
Third, hand-held devices fall under §15.519 instead, which permits indoor or outdoor use but prohibits “any fixed outdoors infrastructure” and requires the transmitter to stop within 10 seconds without an acknowledgement from its receiver.
And one rule that surprises almost everyone: §15.521(a) states that UWB devices “may not be employed for the operation of toys” and that “operation onboard an aircraft, a ship or a satellite is prohibited.” If your asset-tracking plan includes a vessel or an aircraft cabin, that sentence is the first thing to resolve.
In Europe the conformity route runs through ETSI’s harmonised standards. Location tracking is Part 2 of the series, ETSI EN 302 065-2 (V2.1.1, November 2016 edition), covering devices “used for location tracking purposes”; later editions split it into sub-parts, so confirm the current version and its Official Journal status before you rely on a supplier’s declaration.
When is UWB the wrong choice?
When your accuracy requirement is metres rather than centimetres, or when the building has too many walls to anchor economically. Those cases are common, and pretending otherwise produces a system that is over-specified in the corridors and dark in the rooms.
| Property | UWB, 802.15.4 family | Wi-Fi RTT, 802.11mc |
|---|---|---|
| Ranging method | two-way ranging on short pulses | fine timing measurement round trip |
| Channel bandwidth | from 499.2 MHz upward | 5 to 160 MHz |
| Behind five drywalls, NIST test | no estimate produced | estimate in every case |
| Pooled mean absolute error, NIST | 0.27 m | 3.04 m |
| Infrastructure needed | dedicated anchors and backhaul | existing access points |
The NIST authors reach the same balanced verdict: “depending on the application, Wi-Fi RTT can be a viable alternative to UWB for RF ranging and indoor localization.” Bluetooth Low Energy sits in a third place again, cheap to deploy and battery-friendly, with accuracy from signal strength or angle of arrival rather than flight time.
A realistic large-site design is usually mixed. UWB where centimetres pay for themselves, such as forklift safety zones, tool cribs, sterile-instrument trays and robot docking. A broader technology for room-level and floor-level context everywhere else. One map, one coordinate system, one set of IDs underneath both.
How should you write the specification?
Specify the error at a stated confidence, in named test conditions, against a published test method. “Up to 10 cm” is a marketing claim, not an acceptance criterion, and it is satisfied by a single reading in an empty corridor.
ISO/IEC 18305:2016, Information technology — Real time locating systems — Test and evaluation of localization and tracking systems, exists for this. It treats the system as a black box, defines building categories and entity and mobility classes, and combines them into a fixed set of test scenarios. ISO lists the 2016 edition as current after its 2022 review, with a draft revision in progress.
| Clause to put in the specification | What it prevents |
|---|---|
| Error at 95%, not “up to” | a best-case reading becoming the spec |
| Scenarios named from ISO/IEC 18305 | each side inventing its own test |
| Separate line-of-sight and NLOS targets | one average hiding the failures |
| Anchor count frozen after an on-site survey | a re-quote halfway through installation |
| Tag battery life quoted at your update rate | life measured at 1 Hz, deployed at 10 Hz |
Add one more line that costs nothing and saves arguments: require a pilot in the worst room, not the best one. The room with metal shelving and a wet wall decides your anchor count, and it is cheaper to discover that in a two-week pilot than in a change order.
If you are scoping ultra-wideband for a specific building, the useful next step is a look at your actual floor plan, construction materials and accuracy requirement, which is what sets the anchor count and the price. Mapsted works on indoor positioning, indoor mapping, wayfinding and asset tracking for sites like these. Book a scoping call and bring your floor plans.
Frequently asked questions (FAQ)
How accurate is UWB indoor positioning?
In pooled line-of-sight and mild non-line-of-sight testing, NIST measured a mean absolute ranging error of 0.27 m for the better of two UWB systems, against 3.04 m for Wi-Fi RTT. A TU Dresden preprint in a 44 by 30 m hall reports 0.28 m in line of sight and 3.78 m in non-line-of-sight conditions. Obstruction, not the radio, dominates the result.
How many UWB anchors does a building need?
It depends on partitions and materials rather than floor area. NIST concluded that a wooden single-family house with drywalls needs “about a dozen UWB anchor nodes”, and noted that any moderately sized or larger building requires a large number of radios because the permitted emission level only penetrates a few obstacles. Trilateration also needs three anchors in range for two dimensions, four for height.
Can UWB anchors be installed outdoors?
Not under 47 CFR §15.517, which limits that section to transmitters employed solely for indoor operation and forbids intentionally directing emissions outside the building. Hand-held UWB devices fall under §15.519 and may be used outdoors, but that section prohibits any fixed outdoor infrastructure and allows antennas only on the hand-held device itself.
Which UWB channel should a multi-region deployment use?
Channel 9, centred at 7987.2 MHz, because it falls inside both the US 3,100 to 10,600 MHz range in Part 15 Subpart F and the 6 to 8.5 GHz generic row in Commission Implementing Decision (EU) 2019/785. Channel 5 at 6489.6 MHz is also inside both. Channels below 6 GHz are not covered by the EU generic row.
Is UWB better than Bluetooth Low Energy for indoor positioning?
For centimetre-level distance it is, because flight time on a 499.2 MHz pulse resolves multipath that signal strength cannot. For room-level context across a whole site, Bluetooth Low Energy or Wi-Fi is usually cheaper to cover with, since UWB coverage is bought in anchors. Most large sites end up mixing both under one map.
