BLE Channel Sounding vs UWB: Price the Accuracy Gap, Not the Tag

What accuracy does each radio actually deliver on a warehouse floor?

Bluetooth 6.0, published by the Bluetooth SIG in September 2024, added Channel Sounding: two-way ranging that combines phase-based measurements across the 2.4 GHz band with round-trip timing. A sounding exchange hops up to 72 channels at 1 MHz spacing, roughly 80 MHz of usable spectrum. That bandwidth sets the floor, because raw range resolution scales as the speed of light over twice the bandwidth: 80 MHz resolves about 1.9 meters before any clever processing. Phase unwrapping recovers much finer estimates on open benches, which is why vendor demos show 10 to 30 cm, but between steel racking with blocked direct paths the honest planning number is 0.5 to 1 meter, with occasional multi-meter outliers. UWB under IEEE 802.15.4z times a nanosecond-scale pulse across a 499.2 MHz channel. The same math gives 30 cm of raw resolution, and real deployments on Qorvo DW3000 or NXP Trimension silicon hold 10 to 30 cm through typical warehouse multipath. UWB has six times the bandwidth.

What does the accuracy gap cost per tag?

A channel sounding tag is a single BLE SoC. Nordic's nRF54 series supports the feature, and parts in that class price around $1 to $2 at 10,000-unit volume. A UWB tag needs a ranging chip, typically from the Qorvo DW3000 family or NXP Trimension line, at roughly $3 to $6, and almost always carries a BLE companion radio anyway for provisioning and wake-up, plus a second antenna, extra matching components, and a larger battery to feed transmit bursts. Built and tested, expect a $4 to $8 delta per tag. Across 20,000 tags that is $80,000 to $160,000 before infrastructure. Anchors widen the gap: UWB anchors commonly land at $100 to $300 in hardware before cabling and lift time, while channel sounding rides on cheaper BLE gateway hardware. Spread over a five-year tag life, though, the delta shrinks to pennies per asset per month. The BOM alone decides nothing.

How do you price the error instead of the radio?

List every workflow where a person or machine acts on a reported position, then ask two questions: how often does the system answer, and what does a wrong-by-one-meter answer cost when it does?

Error cost per day = location-driven events per day x share that go wrong at this accuracy x minutes lost per miss x loaded labor rate.

A worked hypothetical with placeholder numbers: pick faces 1.4 meters wide mean a 1-meter error regularly points at the neighboring bin. If 2 percent of 5,000 daily picks trigger a 30-second search at $25 per hour loaded labor, that is about $21 per day, roughly $5,400 per year, for one error mode in one building. Run the same arithmetic for lost pallets, misdirected forklifts, and cycle-count rework, with your own rates. Then compare the total against the $80,000 to $160,000 UWB tag premium from the section above. Sometimes the error bill dwarfs the radio bill, and sometimes it is the reverse. Price the wrong answer, then pick the radio.

Where channel sounding's meter wins the math

Standard pallet bays run 2.6 to 3 meters wide, so a position good to 1 meter still names the correct bay almost every time. That covers a surprising share of warehouse work: finding returnable containers, staging pallets, locating carts and tools at zone level, and confirming a load reached the right dock door. These workflows want tag count, not precision. Ten thousand $2 radios beat one thousand $8 radios when the job is knowing roughly where everything is. Channel sounding adds one more lever: Android's Ranging API exposes it on supported Bluetooth 6.0 handsets, so a worker's own phone handles the last 10 meters of a search with no extra reader hardware. UWB phones exist too, but only channel sounding shares its silicon with the dollar-class beacon economy, which is what makes tagging every tote and cart affordable in the first place. At bay level, the extra meter costs nothing.

Where UWB's 30 centimeters still pays the premium

Some jobs break at a meter:

• Bin-level pick and put verification, where slots sit 30 to 40 cm apart and a one-bin miss is a mispick.
• Forklift proximity warning, which must tell a worker at 1 meter from one at 3, reliably, past a steel mast.
• Robot and AGV docking, which wants repeatable sub-10 cm guidance at the final approach.
• Fast-moving assets, where UWB's short pulses support higher update rates with less motion blur in the estimate.

There is also a floor effect that averages hide. Channel sounding's outliers, the multi-meter misses when the direct path is blocked, land exactly where racking is densest and the answer matters most. A verification or safety system is judged on its worst reading, not its median, and 2.4 GHz phase measurements degrade in precisely the cluttered aisles those systems exist to protect. Below half a meter, buy UWB.

Why the positioning stack matters more than the radio you pick

Whichever radio wins your math, the engineering months go into what sits between raw ranges and a position someone trusts: multipath rejection, anchor calibration, outlier filtering, and fusing RF with inertial or vision data. That ground is heavily patented, and it does not care whether your ranges came from an 80 MHz phase sweep or a 499.2 MHz pulse. Position Imaging licenses hundreds of granted US patents in radio-frequency ranging and position estimation, including US 11,774,249 and US 12,066,561, in a portfolio cited by Apple, Bosch, and other major filers. Licensing buys two things a from-scratch build cannot: a proven path from noisy ranges to stable positions that ships in months instead of quarters, and freedom to operate once your tracking product starts winning deals and drawing attention. The radio choice is a BOM line that you can revisit next hardware rev. The positioning stack is the product. License the math, spend engineers on product.