UWB vs BLE Channel Sounding: Accuracy in Steel-Racked Warehouses

What does steel racking actually do to radio ranging?

Datasheet accuracy numbers come from open labs. A distribution center is the opposite: rows of steel uprights and crossbeams in bays roughly 2.7 meters wide, product stacked 10 meters up, forklift masts moving through every aisle. At 2.4 GHz, where Bluetooth operates, the wavelength is about 12.5 cm and steel is a near-perfect reflector. Every ranging measurement arrives as a bundle of copies: the direct path, plus reflections off racking that land nanoseconds behind it.

Bluetooth 6.0 channel sounding estimates distance from phase measurements stepped across roughly 72 channels spanning about 80 MHz of spectrum. That math works when the direct path dominates. When a loaded rack blocks line of sight, the strongest arriving energy is a reflection, the phase slope bends, and the estimate quietly locks onto a path that bounced off a beam two aisles over. The resulting error is not noise you can average away. It is a bias, and it almost always reads long, because every reflected path is longer than the straight one.

Racking turns clean phase math into guesswork.

Why does UWB hold accuracy where phase ranging drifts?

IEEE 802.15.4z UWB channels are 499.2 MHz wide, roughly six times the spectrum Bluetooth channel sounding can stitch together. Bandwidth buys time resolution: a 2 nanosecond pulse lets a receiver separate arrival paths about 60 cm apart. Instead of inferring distance from a phase slope, a UWB receiver reconstructs the channel impulse response and picks the leading edge, the first energy to arrive. That first path is the direct one even when a stronger reflection follows a few nanoseconds later, which is exactly the situation between loaded racks.

The practical result: production silicon such as Qorvo's DW3000 family and NXP's Trimension SR150 delivers 10 to 30 cm ranging in warehouse conditions, and it degrades gradually rather than jumping meters when line of sight closes. Time-difference-of-arrival deployments keep the tag side simple too. The tag transmits a short blink, ceiling anchors timestamp it, and a central solver does the geometry, so tag firmware stays dumb and cheap to certify.

Bandwidth pays for accuracy, and nothing else does.

How accurate is Bluetooth 6.0 channel sounding on a live floor?

Channel sounding shipped in the Bluetooth Core 6.0 specification in September 2024, combining phase-based ranging with round-trip time as a spoofing check. Silicon is real: Nordic's nRF54 series and NXP's KW47 both support it, and vendor line-of-sight demos show 10 to 50 cm. On a working floor with racking, our field experience and published chipset guidance both point to 1 to 3 meters once paths get obstructed.

That is genuinely useful for a large class of warehouse jobs:

• Dock doors: which door a pallet left through, and in which direction
• Staging: presence in a lane, not position within it
• Tools and MHE: which aisle the scanner or pallet jack sits in
• Cycle counting: confirming a tagged asset is in the expected zone

It is not enough for slot-level putaway verification. A standard GMA pallet footprint is 1.2 by 1.0 meters, so adjacent slots sit well inside a 2 meter error bar, and picking the right pallet among four in a staging lane becomes a coin flip.

Buy zone accuracy and plan for zone accuracy.

What does each radio cost per tracked pallet?

Tag hardware sets the headline number. BLE asset tags run roughly $3 to $10 at volume; UWB tags run roughly $15 to $40 depending on battery and housing. Across 10,000 tracked pallets, a $12 per-tag gap is $120,000 before infrastructure, which is why nobody puts UWB on assets that only need zone answers.

But cost per pallet is tag price plus amortized infrastructure plus battery service, and the other two terms cut the other way. A UWB TDoA blink is a sub-millisecond transmission heard by every anchor in range, so one tag transmission yields one position fix and batteries stretch to multiple years at reasonable update rates. Bluetooth channel sounding runs as a connected two-device procedure, so each fix costs a ranging session per anchor pair, which taxes both tag batteries and anchor capacity as tag counts climb. Model your real update rate, per aisle and per shift, before trusting any per-tag comparison. Then price the failure mode: one truckload loaded from the wrong staging position usually costs more than a year of tag price deltas.

Count the errors, not just the tags.

Which radio should you ship in 2026, and whose patents are you standing on?

The decision rule is short. If the workflow decision depends on which pallet or which slot, use UWB. If it depends on which zone, dock, or aisle, Bluetooth 6.0 channel sounding is accurate enough and far cheaper per tag. Large floors should run both: BLE on the long tail of returnable containers and tools, UWB on the assets where a wrong pick has a dollar figure attached, with one location engine fusing the two.

Either way, you are shipping into a dense patent field. Time-of-flight ranging, multilateration solvers, first-path detection, and hybrid RF-plus-vision tracking are covered by granted patents held by companies that got there years ago, and a working prototype is not freedom to operate. Position Imaging licenses hundreds of granted patents in RF ranging, real-time positioning, and computer vision, work cited by Apple and Bosch in their own filings; the portfolio includes US 11,774,249, US 12,079,006, US 12,066,561, and US 12,000,947. Licensing proven methods means your engineers spend their months on the product, not on re-deriving solver math that is already claimed.

License the ranging, ship the product.