BLE 6.0 Channel Sounding vs UWB: Anchor Density for Sub-Meter

Why doesn't the ranging spec predict shipped accuracy?

Chipset marketing quotes one number: ranging error on a single link, line of sight, on a bench. Asset tracking needs a position, and a position solver combines three or more ranges into an x, y estimate. Geometry multiplies per-link error through dilution of precision, the same DOP term GPS engineers watch.

Concrete case: mount your anchors along one wall of a warehouse because that is where power is. Tags in the far aisles see all anchors from nearly the same direction, DOP climbs to 3 or 4, and a 40 cm ranging error becomes 1.2 to 1.6 m of position error. Surround the same tag with anchors at wide angles and DOP drops near 1.3, so the same radio delivers roughly 50 cm.

This applies identically to Bluetooth 6.0 Channel Sounding and to UWB. The spec sheet tells you the floor of your error, and the anchor layout tells you the multiplier. Geometry eats spec sheets.

How dense an anchor grid does sub-meter channel sounding need?

A working rule for both radios: every tag position should see at least three anchors with usable signal paths and more than 60 degrees of angular spread between them.

UWB TDoA deployments commonly run 20 to 40 m anchor spacing under open ceilings and still hold 10 to 30 cm, because each anchor's timestamp survives clutter well. Channel sounding at 2.4 GHz propagates through racking, but what arrives is a phase estimate corrupted by every reflection along the way. In racked aisles, plan more like 10 to 20 m between channel sounding anchors so each fix has enough clean-geometry links to reject the bad ones. Budget roughly 1.5 to 2 times the anchor count of a UWB design for the same floor if sub-meter P95 is the target.

The saving grace: a channel sounding anchor is a BLE SoC, hardware like Nordic's nRF54 series, so the denser grid can still cost less per installed point than UWB infrastructure. Grid for geometry, not for coverage.

What does 80 MHz of bandwidth do to multipath indoors?

Time resolution scales inversely with bandwidth, and this is the physics neither vendor roadmaps nor firmware updates can change.

UWB per 802.15.4z HRP uses 499.2 MHz channels and roughly 2 ns pulses. The receiver can separate two paths whose lengths differ by about 60 cm, so it picks the direct path out of the clutter in the time domain and ignores the bounce off the racking.

Channel Sounding hops across up to 72 channels at 1 MHz spacing in the 2.4 GHz band, an effective span near 80 MHz. That resolves paths about 3.7 m apart. A reflection whose path is 2 m longer than the direct one merges into a single corrupted phase estimate, and the error lands on your range. The standard pairs phase-based ranging with a round-trip time check, but RTT at this bandwidth is a meters-grade sanity test, not a fix.

Line of sight, channel sounding demos hold 30 to 50 cm. Add metal racking and moving forklifts and 1 to 2 m is the honest expectation. Bandwidth is the one thing you cannot tune.

Does connection-based ranging cap your tag count?

Channel sounding runs over a BLE connection. Each ranging procedure is a scheduled two-way exchange between the tag and one anchor, costing tens of milliseconds of radio time, and a position fix needs that repeated against three or more anchors. Multiply tags by anchors by update rate and you are spending shared 2.4 GHz airtime, in the same band as every WiFi access point on the floor.

UWB TDoA inverts the whole model. The tag transmits one blink, under a millisecond of airtime, and never connects to anything. Anchors timestamp the blink and the infrastructure solves the position. That is how a single cell tracks thousands of tags at once-per-second updates while coin-cell tags last for years.

So run the capacity math before the accuracy math:

• Hundreds of assets, updates every 10 to 60 seconds: channel sounding airtime fits comfortably.
• Thousands of pallets, one-second updates: two-way ranging schedules collapse, and TDoA wins regardless of accuracy.

Airtime is the hidden anchor budget.

How do you prove the grid, and what should you build versus license?

Pilot both radios on the same floor section before committing hardware at scale. Measure the full error distribution, judge P95 under real forklift traffic rather than mean error in an empty aisle, and deliberately include your worst geometry, the corner where anchors cluster on one side.

Then look hard at the software between raw ranges and a trustworthy position: multipath rejection, DOP-weighted solvers, fusion of RF with inertial or vision inputs. That layer takes positioning teams years to get right, and it sits in heavily patented territory on both the BLE and UWB sides.

The radio standard gives you access to the PHY. It gives you nothing on the tracking methods built on top of it.

Position Imaging licenses hundreds of granted US patents in RF ranging, real-time positioning, computer vision, and machine learning, IP cited by Apple, Bosch, and other major firms. Licensing the proven methods means your team integrates instead of re-deriving, with freedom to operate documented up front, and products that ship in months. License the ranging math, ship the product.