Channel Sounding Chipsets vs UWB Silicon: 2026 Field Accuracy

What channel sounding silicon actually ships in 2026?

The Bluetooth SIG adopted Channel Sounding in September 2024 as part of Bluetooth 6.0, and the first production parts are now in volume. Nordic's nRF54L15 supports it in the stock SDK, and NXP's KW47 and MCX W72 families were built around it. The feature combines two measurements: phase-based ranging (comparing carrier phase across hopped channels) and round-trip time, with the RTT leg doubling as a spoofing check for applications like car keys and door access.

Vendor demos quote 10 to 30 cm of ranging error. Read the fine print: those numbers come from open-air, line-of-sight tests with fixed antennas. Field results in ordinary indoor spaces run closer to 0.5 to 1 m, and 1 to 3 m once walls, racking, or people sit between tag and anchor. That is still a large improvement over RSSI-based BLE, which struggles to beat 3 to 5 m indoors. But it is not UWB parity, and no firmware update will make it so. Datasheet accuracy is a lab number.

Why UWB still wins raw accuracy: 500 MHz beats stitched megahertz

Ranging accuracy is set by how sharply a radio can resolve arrival time, and that is set by bandwidth. An 802.15.4z UWB channel is 499.2 MHz wide, so a pulse lasts about 2 ns. Light travels 30 cm per nanosecond, which means silicon like the Qorvo DW3000 family or NXP's Trimension SR150 can separate the direct path from a reflection arriving a few nanoseconds later. That is the whole trick behind 5 to 10 cm line-of-sight accuracy, and it is why Apple has shipped UWB in every flagship iPhone since 2019.

Channel sounding works with 1 MHz BLE channels and rebuilds distance from phase measurements stitched across roughly 72 MHz of the 2.4 GHz band. In clean air the math works well. In multipath, reflected copies of the carrier corrupt the phase at each hop, and there is no sharp pulse edge to fall back on. The RTT leg helps sanity-check the result but has meter-class resolution on its own. Bandwidth decides, everything else negotiates.

Datasheet accuracy vs warehouse accuracy: what the floor does to both

Every buyer should test in their own building, because steel is the great equalizer. Metal racking, forklifts, and mezzanines create dense multipath that punishes both radios, just not equally.

• UWB typically slips from 10 cm to 30 cm through light obstruction, and can reach 1 m through a fully loaded rack. Leading-edge detection keeps the direct path findable as long as some signal gets through.
• Channel sounding slips from 0.5 m to somewhere between 1 and 3 m in the same conditions, because phase corruption compounds across every hopped channel.

Airtime matters as much as error. A full channel sounding procedure steps through dozens of channels and takes tens of milliseconds per tag pair, while a UWB double-sided two-way ranging exchange completes in about a millisecond. At 50 tags per zone that gap decides whether you get one position per second or one every ten. Model your update rate before you fall in love with either BOM. The floor always taxes the datasheet.

When channel sounding is the right call, and when it is not

Choose channel sounding when meter-level answers close the workflow and cost rules. It reuses the 2.4 GHz radio already in your tag, so there is no second chip, no second antenna, and no UWB certification pass. That fits pallet-level location, finding an infusion pump on the correct wing, presence detection at a door, or theft alerts at a portal. It also inherits BLE's install base: every phone on earth speaks Bluetooth today, while channel sounding handset support is only starting to arrive in 2026.

Choose UWB when the tolerance is under 30 cm and the workflow breaks without it: bin-level put-away verification, robot docking, tool cribs, or keep-out zones around moving equipment. The strongest 2026 designs use both, with BLE handling discovery and coarse zoning while UWB fires only when a precise fix is needed, which also rescues battery life. Match the radio to the tolerance.

The radio is swappable. The positioning math is not.

Here is what chipset marketing skips: the receiver hands you a noisy range estimate, and everything that turns ranges into a trustworthy position is algorithm work. Multipath rejection, anchor geometry, motion models, fusing RF with inertial or vision data, all of it behaves the same whether the range came from a 499.2 MHz UWB pulse or a stitched 2.4 GHz phase sweep. Teams that rebuild this from scratch typically spend 12 to 24 months rediscovering failure modes that are already solved, and much of that solution space is patented.

Position Imaging licenses hundreds of granted patents in radio-frequency ranging and real-time positioning, including US 11,774,249, US 12,079,006, US 12,066,561, and US 12,000,947, from a portfolio cited by Apple and Bosch. Because the IP sits upstream of waveform choice, a license holds whether you ship channel sounding tags this year and add UWB next year, or the reverse. That is freedom to operate plus a shortcut through the hard math. License the math, then pick the radio.