Current data path

The wand computes distances from anchor replies and uploads them. The gateway filters position; gyro-based orientation supplies aim. The wand samples IMU at 200 Hz, decimates tracking samples to 100 Hz and batches them into 30 UDP datagrams/s. These rates are distinct from UWB cycles.

Future TDoA

Wands would send blinks; synchronized anchors would timestamp reception and report those records to a new gateway correlation/solver path. Current anchors do not do this. PoE splitters provide power; management data remains Wi-Fi.

What the current map means

The checked-in map is 200 ms, eight 25 ms cycle-start slots, with nominal 5 Hz UWB updates. The current scheduler explicitly pauses surplus assignments instead of sharing start slots. Its two-wand deployment does not validate physical capacity or over-capacity behavior with a larger fleet. The scheduler cannot stop a radio exchange at a slot boundary, and Wi-Fi delivery jitter can overlap starts.

The old 100 ms / 16 × 6 ms / 8–10 Hz model assumed an exchange time that later hardware measurements did not support. It is not a measured capacity curve.

Installed versus ranged anchors

The latest test uses five anchors per wand. Geometry, redundancy and NLOS resilience must be balanced against airtime; automatic selection uses a geometry heuristic. More physical anchors are a coverage option, not a promise of free capacity.

Measure 1/2/4/8/12/16 physical wands: distinct cycle rate, reply success, position error, p95/p99 gaps, latency and moving behavior. Twelve to sixteen players remains a product target.

TDMA: approximate start times

Gateway assigns 8 × 25 ms slots in a 200 ms superframe and sends a Wi-Fi time reference every 1 s. Installed wand firmware atomically claims at most one cycle per assigned superframe and pauses new ranging without fresh Wi-Fi sync. With the current 1 s sync cadence, the freshness limit is 5 s. Two-wand starts remain 100 ms apart. An in-flight cycle may continue beyond its start slot; this is not a hard airtime reservation.

CSMA: listen before the poll

DWM v1.6.0 added channel listening with bounded random backoff, up to six attempts. In v1.7.0 it guards the single multi-poll. It can reduce some collisions but adds variable delay and cannot eliminate hidden terminals or nearly simultaneous starts.

Multi-poll: one poll, ordered replies

DWM v1.7.0 lists the anchors in one poll; each responds in its assigned response slot. The wand still calculates a separate SS-TWR distance for every reply. These anchor response slots are distinct from gateway-assigned wand start slots.

IMU and aim

One 200Hz BMI270 sampler, 100Hz tracking samples, 30Hz UDP batches. Gyro bias learning, rest detection/ZUPT and explicit arena alignment help, but six-axis sensing has no absolute yaw reference. Position accuracy and aim drift are separate problems.

UWB and position

DWM v1.7 multi-poll SS-TWR, up to five anchors per wand. Current configured start schedule is 200 ms / eight 25ms slots: nominal 5 cycles/s, not a guaranteed radio/frame rate. CSMA can additionally guard transmission. Neither is TDoA or precise anchor clock synchronization.

Current estimation

The gateway already uses range-domain Kalman prediction/correction, range gating, uncertainty and recovery handling. It is not merely a confidence-weighted average waiting for its first estimator. NLOS, correlated range bias, survey geometry and sparse corrections remain material limits.

TDoA is a separate migration

Tags would blink; anchors would timestamp reception and forward observations in a shared radio-time model. Current anchors reply to TWR polls and do not stream that timestamp data. New synchronization, backhaul/firmware and a TDoA solver need design and validation. A PoE power splitter supplies none of those automatically.

Optics and effects

Camera-side IR needs its own installation assessment. Dense haze can reduce optical visibility; UWB/IMU operation and validated degradation behavior matter independently. The finished visible laser assembly has no established Class 1 classification from a seller label or attenuation cover alone.