A common misconception about RTK is treating it as a switch: on means centimetre-level, off means metre-level. In practice the fix state changes repeatedly, and the most damaging periods are those that look fixed but are no longer trustworthy.

Baseline length is the first constraint

RTK cancels common errors through correlation between base and rover observations. The further apart the receivers, the weaker the correlation of ionospheric and tropospheric errors, and the harder it is to resolve integer ambiguity.

For that reason a "maximum baseline" figure is meaningless without conditions. Size against your actual operating radius rather than a datasheet limit.

Data-link latency is routinely underestimated

Corrections must reach the rover in real time. As latency grows, corrections no longer align with current observations, accuracy degrades and the fix may be lost.

For unmanned platforms, the data link often becomes the bottleneck before GNSS itself does.

Obstructions and multipath do not disappear with RTK

RTK removes common errors, not local ones. Buildings, trees, metal structures and ground reflections directly degrade pseudorange and carrier observations.

The typical symptom is a decent satellite count, a displayed fixed solution, and a position that drifts or jumps slowly. Increasing transmit power does not help; the remedy is antenna placement, elevation masking and environmental avoidance.

Criteria for integrators

Acceptance testing

Test in the real operating environment rather than an open field, covering the worst time of day and obstruction conditions. Run continuously and measure fix availability rather than quoting the best minutes. Require suppliers to state baseline length, data-link method and environment for their figures — without that, schemes cannot be compared.

Conclusion

RTK stability is jointly determined by baseline, data link and environment. Treating "fixed or not" as the only metric misses the most dangerous case: fixed status with an untrustworthy position.