Pallet-Level RTLS Accuracy: The 50 cm Bar for BLE 6.0 vs UWB

Why pallet-level tracking needs 50 cm, not 1 meter

A standard GMA pallet measures 48 by 40 inches, about 1.2 by 1.0 meters. In a floor staging lane pallets sit touching or a hand's width apart, so binding a position fix to the correct pallet means keeping error under half the short side, roughly 50 cm. Miss that and your WMS shows the right lane but the wrong pallet, and a forklift ships the neighbor's load.

Racking is tighter. A 3.7 m selective rack bay typically holds three pallet positions, and beam levels stack every 1.5 to 1.8 m vertically, so slot-level assignment needs about 0.5 m horizontal accuracy plus enough vertical accuracy to name the level. Compare that with aisle-level zone tracking, where 3 to 5 m of error is fine because aisles are wide and labeled.

• Which aisle or lane: 3 to 5 m error is acceptable
• Which pallet in the lane: under 50 cm
• Which slot on which beam: under 50 cm in three dimensions

Decide which question your operation actually asks before comparing radios. Geometry sets the spec, not the datasheet.

What does Bluetooth 6.0 Channel Sounding actually change?

Before the 6.0 core specification, published in September 2024, BLE location meant signal-strength fingerprinting at 3 to 5 m or Bluetooth 5.1 direction finding at meter level with careful antenna arrays. Channel Sounding adds true two-way ranging: phase-based ranging stepped across up to 72 channels in the 2.4 GHz band, combined with a round-trip time measurement that cross-checks the phase result and resists relay attacks. Distance comes from measured physics, not a propagation guess.

Silicon is real. Nordic's nRF54 series supports Channel Sounding today, and vendor line of sight results land between 30 and 50 cm. The part that matters for pallets: the tag is a single 2.4 GHz SoC with a coin cell, the same hardware class as the beacons warehouses already deploy by the thousand. No second radio, no second antenna, no separate UWB stack.

For a radio that was guessing distance from RSSI two years ago, that is a real jump. Ranging became a firmware-class feature.

Why steel racking punishes 2.4 GHz ranging

A receiver separates two propagation paths only when their lengths differ by roughly the speed of light divided by bandwidth. UWB under IEEE 802.15.4z uses 499.2 MHz channels, channel 5 near 6.5 GHz and channel 9 near 8 GHz, so paths 60 cm apart resolve cleanly and the chip locks onto the first arrival. Channel Sounding gets about 80 MHz of usable span in the 2.4 GHz ISM band, so any reflection within roughly 3.7 m of the direct path folds into the same estimate.

Now stand in a racked aisle. Steel uprights, beams, wire decking, forklift masts, and stretch-wrapped metal loads all sit inside that 3.7 m window. The clean 40 cm result from the office hallway degrades toward a meter or worse between racks. UWB deployments hold 10 to 30 cm in the same aisle because reflections arrive as separable pulses, not a smeared phase sum.

Loaded pallets add attenuation on top, since water-heavy product like beverages absorbs 2.4 GHz strongly. Steel aisles are a bandwidth test.

What does a tag cost per pallet in each radio?

A Channel Sounding tag is one SoC, a coin cell, and a case; volume hardware cost lands around 3 to 6 dollars. A UWB tag pairs a transceiver in the Qorvo DW3000 or NXP Trimension class with a BLE MCU for provisioning, and shipping tags typically price between 10 and 25 dollars. Across 20,000 tracked pallets that spread is 150,000 to 400,000 dollars in tags, before anchors.

Two things cut against the easy savings:

• Battery is a wash. Pallets move rarely. At one ranging event per minute, both radio types run for years on a coin cell, so power does not decide this.
• Error has a price too. Every mis-assigned pallet costs forklift minutes of searching and can hold an outbound trailer at the door. If your lanes are packed, a 1 m radio assigns the wrong pallet routinely, and the tag savings flow back out as labor.

Cheap tags only pay when they're right.

Should UWB go on the forklift instead of the pallet?

A pallet does not move itself. It changes position only when a forklift or hand jack moves it, which means you rarely need every pallet reporting high-accuracy position continuously. The architecture that pencils out at scale inverts the tag question: put UWB on the vehicles and cheap radios on the pallets.

• UWB tracks each forklift at 10 to 30 cm, including fork height from a mast-mounted tag, so the system knows exactly which slot received the pallet.
• A low-cost BLE or Channel Sounding tag on the pallet confirms identity at pick and drop.
• The location record comes from the put-away event, not from the pallet constantly ranging.

Fifty vehicles carrying UWB beats 20,000 pallets carrying it. Channel Sounding then covers what event-based tracking misses: floor lanes rearranged by hand, dock door assignment, staging overflow. You get slot-level truth in the racks and 50 cm class coverage on the floor without buying 20,000 premium tags. Track the vehicle, not twenty thousand pallets.

What this means if you're building the system

Whichever radio wins your bakeoff, the hard part is not the chip. It is the first-path detection, multipath rejection, and position calculation that keep accuracy honest inside steel, and those methods took the industry years of warehouse deployments to get right. Patent exposure compounds the problem: RF ranging and RTLS are crowded patent fields, and building a system from scratch does not mean you built it clear.

Position Imaging licenses hundreds of granted patents covering radio-frequency ranging, position calculation, and the computer vision and machine learning that fuse radio fixes with camera data. The portfolio is cited by Apple, Bosch, and other major firms in their own filings. Licensing means your team integrates proven methods and ships in months, with freedom to operate instead of a clearance headache. License the physics lessons someone already paid for.