BLE Channel Sounding Chipsets in 2026: Can They Replace UWB Yet?

Which Bluetooth 6.0 channel sounding chipsets can you actually buy in 2026?

The Bluetooth SIG finalized Channel Sounding in the Bluetooth 6.0 spec in late 2024, and the silicon gap has mostly closed since. Nordic Semiconductor's nRF54L15 is the part most tag designers start with: channel sounding runs in production silicon with support in the nRF Connect SDK, and it sits in the same price class as the nRF52 tags already bolted to warehouse assets. NXP announced the KW47 and MCX W72 families with channel sounding at the start of 2025, aimed at industrial and automotive designs. Other vendors have announced parts, but before you commit a board, check three things that marketing pages skip:

• Does the stack support both ranging methods, phase-based ranging and round-trip time, or only one?
• What is the real ranging exchange time, since it directly caps update rate and battery life?
• Is reflector mode supported, so your cheap tag can stay dumb while anchors do the math?

The spec is two years old. The errata are younger.

What accuracy do channel sounding chipsets deliver next to steel racking?

Channel sounding measures phase across up to 72 narrow channels stepped through the 2.4 GHz ISM band, roughly 80 MHz of usable spectrum, and fuses that with round-trip timing. Bandwidth decides multipath behavior. A radio can only separate two signal paths whose lengths differ by about the speed of light divided by the bandwidth: 80 MHz gives roughly 3.7 m of path resolution, while UWB's 499.2 MHz channel gives about 0.6 m. In a warehouse, that difference is the whole story. A reflection off a steel upright that arrives 2 m of path length after the direct signal merges into one blurred estimate on BLE, while UWB still picks the first path cleanly.

In practice that means channel sounding delivers around 0.5 m in open line of sight, the number the Bluetooth SIG itself targets, and 1 to 3 m in dense racking where the direct path is attenuated or blocked. That is bay-level and aisle-level truth, not slot-level truth. Good estimators narrow the gap. Physics does not close it.

Has UWB silicon stood still while BLE caught up? No.

The comparison is not against 2020 UWB. Qorvo's DW3000 family and NXP's Trimension line (SR150 class parts) implement 802.15.4z HRP, operate on 499.2 MHz wide channels around 6.5 and 8 GHz, and hold 10 to 30 cm ranging accuracy through the multipath that pushes BLE to meters. FiRa Consortium certification now covers interoperability, so mixing anchor and tag vendors is a procurement decision instead of a lab project. And Apple's second-generation UWB chip, shipping since the iPhone 15 in 2023, extended Nearby Interaction range to roughly three times the first generation, which matters if handhelds or phones serve as mobile readers on your floor.

UWB's costs are the familiar ones: the radio adds dollars to a tag that BLE builds for cents of incremental silicon, each ranging burst draws more current than a BLE exchange, and anchors need their own power and backhaul. None of that changed in 2026. Neither did the accuracy lead.

How should you choose for a warehouse tag design starting now?

Decide from the error budget backward, not from the datasheet forward.

• Bay or zone level (1 to 3 m acceptable): channel sounding wins. One SoC handles telemetry, firmware update, and ranging; a CR2032 tag reporting every few minutes runs for years.
• Slot, pallet-position, or person-safety level (under 0.5 m, enforced): UWB. No BLE estimator survives a contractual 30 cm requirement next to racking.
• Phones as infrastructure: Android 16's Ranging API exposes channel sounding, UWB, and Wi-Fi RTT behind one interface. If workers' devices are your readers, write to that API and treat the radio as a deployment detail.
• Mixed floors: run channel sounding everywhere and add UWB anchors only in high-value zones, cold storage, pharma cages, outbound staging. You pay for precision only where an error has a dollar value.

The trap is designing one tag for both jobs and shipping a compromise that satisfies neither. Pick the error budget first. The radio follows.

Both stacks sit on patent estates. The spec does not clear you.

Adopting Bluetooth 6.0 or 802.15.4z licenses the declared standard-essential claims on the radio layer, and nothing above it. The methods that make ranging data useful in a building, fusing phase and time-of-flight estimates, fingerprinting multipath, fixing position across moving assets, scheduling exchanges across hundreds of tags, are covered by system-level patents that no SIG membership touches. Teams discover this at diligence, usually after the architecture is frozen and a rewrite costs two quarters.

There is a faster route than inventing around twenty years of filings. Position Imaging licenses hundreds of granted US patents in RF ranging, real-time positioning, and computer vision, including US 11,774,249 and US 12,079,006, a portfolio cited in filings by Apple, Bosch, and other major firms. Licensing proven positioning IP does two things at once: it gives your team working methods instead of a cleanroom rebuild, and it converts an open-ended freedom-to-operate question into a line item. Build the product. License the physics-adjacent math.