Detecting a tag is trivial; placing it to sub-metre inside a steel-framed building is not. The angle-of-arrival maths is textbook — the engineering is everything that corrupts the angle: multipath off machinery, antenna-array phase error, and a site geometry that never matches the drawing.
A real-time location system built on Bluetooth angle-of-arrival reads the phase of a signal across an antenna array to compute the bearing to a tag. Two or more locators intersect those bearings into a position. On paper the tag is located to sub-metre; in a real building it often is not.
The reason is that the array does not measure the direct path — it measures the sum of every path. A reflection off a metal rack, a moving forklift or a mesh wall arrives with its own phase and drags the estimated angle away from the truth. The radio link is strong the whole time; the position is simply wrong.
So the system is engineered around the corruption, not the link: antenna calibration, phase-coherent IQ capture, multipath-aware angle estimation and a site survey that treats geometry as a first-class input rather than an afterthought.
Stated before any of them had an answer.
The array sums direct and reflected paths. A strong reflection off metal shifts the estimated bearing with no drop in link quality to warn you.
Element phase and gain vary with manufacturing and temperature; an uncalibrated array reports a confident, wrong angle.
Position is the intersection of bearings. Small angle errors at range become large position errors, and geometry decides how large.
Racks move, mezzanines appear, walls are steel where the plan says plasterboard. Positioning that ignores the real geometry drifts.
The angle is easy to compute and easy to corrupt. Accuracy comes from calibration, multipath rejection and honest geometry — not from a stronger radio.
The scope of the work, rather than a capability list.
Rarely the subsystem that sounds difficult.
The direct path and a strong reflection can look equally valid to a naive estimator; separating them is the core of the accuracy.
An array calibrated on the bench drifts with temperature and mounting; keeping calibration valid in situ is harder than doing it once.
Where the bearings cross at a shallow angle, position error explodes; locator placement is an optimisation, not a convenience.
A demo in an empty room proves nothing; the acceptance test has to be on the real, cluttered, moving floor.
Tags located to a stated, tested accuracy on the real floor rather than in an empty room.
Reflections discriminated and down-weighted, so a metal-rich environment degrades gracefully rather than lying.
An array calibration that survives temperature and time, not a one-off bench figure.
A repeatable way to place locators and model geometry that the customer can apply to the next site.
Customer projects are presented at property, capability, outcome and integration level. Customer names, internal architecture, confidential deliverables and commercial terms are not disclosed. Where a detail would identify a customer it is omitted rather than approximated. More is available under a non-disclosure agreement, within the limits the customer has agreed.
Every item links to its own page.
Angle-of-arrival positioning.
PLATFORMThe wireless platform.
TECHNOLOGYLink budget to antenna.
SOLUTIONLocating things that move.
PLATFORMEdge compute and gateways.
TECHNOLOGYWhere the phase is won.
Because angle-of-arrival measures phase across an antenna array, and the array sums every path the signal took — direct and reflected. A reflection off metal arrives with its own phase and pulls the estimated bearing off, while the link stays strong the whole time. Accuracy comes from rejecting multipath and calibrating the array, not from link margin.
Three things dominate: how well the antenna array is calibrated, how well the estimator separates the direct path from reflections, and the site geometry — where the locators sit relative to each other and the tag. Shallow bearing intersections turn small angle errors into large position errors, so locator placement is part of the design.
Yes, but only if it is engineered for one. An empty-room demo is meaningless; the system has to be tuned and tested against real reflections, moving equipment and the actual floor plan, which is why the site survey and multipath rejection are treated as core work rather than setup.
Written for engineers. Share it with one.
Tell us the specification, the constraint and the deadline. Programmes that cross silicon, radio, embedded and AI are where Faststream is strongest.