Technical detail · UWB Ranging Techniques
Clock drift error budget
30.6 Clock Drift Error Budget
In SS-TWR, device A sends a poll, device B replies after a known processing delay, and A computes distance from the round-trip time minus B’s reply delay. The catch is that the reply delay is measured on B’s clock while the round trip is measured on A’s clock. Those crystals never run at exactly the same frequency. That frequency offset multiplies the reply delay into a ranging error, and reply delays are long enough that small parts-per-million offsets become centimeters.
DS-TWR reduces this by adding a third message so each device measures a round trip. Combining the two round-trip measurements cancels much of the first-order clock-rate mismatch error. The cost is one more message and more energy per range. The choice is concrete: SS-TWR when the fewest messages matter and more error is acceptable; DS-TWR when accuracy must hold despite clock drift; TDoA when tag scale and battery life matter more than avoiding anchor synchronization.
Put numbers on the drift. Suppose the responder waits 100 microseconds before sending its reply and its clock differs by 20 ppm from the initiator. The apparent reply-delay error is about 100 microseconds x 20 / 1,000,000 = 2 ns. In a single-sided range, that becomes about 0.30 m/ns x 2 ns / 2 = 0.30 m of distance bias. If the reply delay stretches to 1 ms, the same 20 ppm offset becomes 20 ns, or roughly 3 m of distance bias before multipath, antenna-delay, or timestamp-noise effects.
DS-TWR still leaves the rest of the error budget. Antenna delay can shift timestamps, non-line-of-sight paths can arrive late, and filtering can publish a fresh-looking range from stale evidence. TDoA has a different budget: if anchor synchronization is off by 0.5 ns, the distance-difference evidence is biased by about 0.15 m. With a geometry factor of 2, that can become about 0.30 m of position uncertainty before environment bias.
Engineering review should attach method, timing age, calibration, channel evidence, geometry, and confidence to every accepted range or coordinate. The distance is only trustworthy when the evidence trail matches the application consequence.