How the Rover Actually Uses It
A base station does not compute anything on the rover's behalf, and it certainly does not hand the rover a finished answer. Everything the base sends is raw material. The actual position gets solved entirely on the rover's side, by comparison.
The rover differences its own satellite observations against the base's, on the assumption that both receivers are experiencing nearly identical atmospheric, orbital, and clock error. That differencing cancels the shared error, leaving mostly the genuine geometric relationship between the two receivers to be solved.
Comparison, Not Computation Handed Over
A satellite signal passing through the atmosphere on its way to a base a few kilometers from the rover experiences almost exactly the same delay reaching both receivers. The same is true of small errors in that satellite's broadcast orbit and clock. None of this needs to be measured or corrected directly. It only needs to be common to both receivers, so that comparing the two sets of observations makes it cancel out on its own.
The base's assumed coordinate is not something the differencing process touches or absorbs, but it is also not a separate number added on afterward. It is built into the estimation from the very start, which is exactly why differencing has no way to notice when that one number is wrong.
Why This Matters for a Bad Base Coordinate
Because differencing never actually checks the base's assumed position against anything, a wrong base coordinate passes straight through the entire process untouched. The rover's carrier-phase solution can be flawless, every shared error correctly cancelled, integer ambiguities correctly resolved, and the resulting position will still carry whatever error the base's own coordinate carried, precisely and uniformly.
A precise, repeatable RTK solution does not verify that the base's assumed coordinate was correct. Differencing cancels shared error between the two receivers. It has no mechanism at all for catching an error in the one number the entire result depends on: the base coordinate itself.
What This Means for Everything Ahead
This is the exact mechanism that makes the rest of this companion make sense. Because a base coordinate error passes through uniformly rather than being absorbed into the differencing, a base established quickly, and imperfectly, can still be refined later, and that refinement can be applied cleanly to whatever a rover already recorded, a subject this companion returns to directly once every method of establishing that base coordinate has been covered.
Two rovers, using the same base, report centimeter-level repeatable positions all day. Does that repeatability confirm the base's own coordinate was correct? No. It confirms the shared-error cancellation worked well. The base coordinate itself was never something either rover was in a position to check.
A rover solves its own position by differencing its observations against the base's, canceling shared atmospheric, orbital, and clock error. The base's assumed coordinate is never verified by this process, only inherited by it, which is exactly why a base coordinate's accuracy deserves careful attention on its own terms.
With the mechanism itself settled, this companion can turn to the physical side of building a base: the hardware, the antenna, and where it actually sits.