UWB vs BLE 6.0 Ranging: The Three Numbers That Pick Your Radio

How accurate is BLE channel sounding against UWB indoors?

Start with bandwidth, because it decides everything downstream. Bluetooth 6.0 channel sounding, released by the Bluetooth SIG in September 2024, measures phase across up to 72 hops in the 2.4 GHz band, about 80 MHz of usable spectrum. UWB under IEEE 802.15.4z uses a single 500 MHz channel near 6.5 or 8 GHz. That ratio matters because bandwidth sets multipath resolution: a 500 MHz signal can separate reflections arriving more than roughly 60 cm apart in path length, while 80 MHz blurs everything within about 3.7 meters into one arrival.

In the field that cashes out as:

• UWB: 10 to 30 cm in production deployments, sub-10 cm with good anchor geometry and TDoA.
• BLE channel sounding: 30 to 50 cm on the bench, 0.5 to 1 m line of sight in real buildings, 1 to 2 m or worse around metal racking.

Nordic's nRF54L15 and NXP's KW47 both ship channel sounding today, and their application material quotes tens of centimeters under clean conditions. Treat that as the best case, not the plan. Physics sets the floor, not firmware.

What does each tag cost at 10,000 units?

Price both tags at a realistic volume, say 10,000 units. A channel sounding tag is one chip: a BLE 6.0 SoC in the nRF54 or KW47 class runs $1.50 to $2.50 at that volume, and with antenna, CR2032 cell, and enclosure you land near $4 to $8 per tag. A UWB tag is two radios, because nearly every UWB design pairs the ranging chip (Qorvo DW3110 or NXP Trimension class, $3 to $5) with a BLE MCU for wake-up and control, plus a larger battery to feed UWB transmit bursts. Landed cost: $12 to $20.

Call the delta $8 to $12 per tag. At 10,000 tags that is $80,000 to $120,000. At 50,000 tags it clears half a million dollars, before counting batteries replaced over a five-year life. Infrastructure tilts the same way: a channel sounding anchor is a commodity BLE node in the $30 to $100 range, while installed UWB anchors commonly run $150 to $400 each. The delta compounds with every tag you ship.

What does switching radios cost beyond the tag?

The spreadsheet above misses the third number: what the migration itself costs. Budget these line items before deciding.

• A new site survey. Channel sounding error behaves differently around metal and glass than UWB first-path detection, so anchor placement gets redone, not copied.
• Firmware and air-time scheduling. A CS procedure and an 802.15.4z two-way ranging session have different exchange patterns, collision behavior, and duty cycles.
• Filter retuning. Your Kalman or particle filters were tuned to UWB's error distribution. CS errors are larger and skew differently in multipath, so the fusion layer needs rework, not a config change.
• Certification. A new tag SKU means FCC and CE runs again, plus Bluetooth SIG qualification in place of FiRa interoperability testing.
• Parallel operation. Fleets never flip overnight. Plan two to three quarters of running both radios, both dashboards, and both spare pools.

Switching costs more than the spec sheet shows.

When should you switch, and in which direction?

Direction matters, and the two common moves in 2026 are not symmetric.

Switch from UWB to channel sounding when all of these hold: 1 meter of accuracy genuinely covers the use case (pallet-level location, room-level equipment finding, zone entry and exit), the fleet is 10,000 tags or larger so the BOM delta repays the migration inside roughly 18 months, update rates stay at or below about 1 Hz, and the space is offices, retail, or light warehousing rather than dense steel racking.

Stay on UWB, or move up to it, when any of these hold: you need 50 cm or better (bin-level picking confirmation, forklift and pedestrian separation, robot docking), you need fast updates on moving assets, or the environment is heavy multipath where 80 MHz cannot isolate the first path.

One more case: if you are on BLE RSSI or AoA today, channel sounding is close to a free upgrade, same band, same class of node, meaningfully better ranging. Below half a meter, UWB keeps the job.

How do you keep the choice reversible?

This decision will come up again. Channel sounding silicon is one generation old, UWB keeps shipping in phones through Apple's second-generation chip, and accuracy requirements tighten the first time operations sees what sub-meter data shows them.

The hedge is architectural: keep the positioning engine radio-agnostic. Ranges and angles go in, positions come out, and the filtering, fusion, and tracking math in the middle is where accuracy actually gets made. It is also the slowest layer to build well and the layer where most patent exposure sits, because the methods predate whichever radio feeds them.

Pick the radio on this year's BOM. Build the tracking layer so next year's radio is a swap, not a rewrite.

Position Imaging licenses hundreds of granted patents in RF ranging, real-time positioning, computer vision, and machine learning, IP cited by Apple and Bosch. Patents such as US 11,774,249 and US 12,066,561 sit in that ranging and tracking layer rather than in any single radio's PHY, which is what makes a radio switch survivable. License the math, keep the radio flexible.