Why RTK Needs a Reference
Left alone, a single GNSS receiver computes a position that is honest but imprecise. Atmospheric delay stretches the apparent distance to every satellite by a slightly different, constantly shifting amount. Orbit and clock data broadcast in real time carry small errors of their own. None of this is a malfunction. It is simply the accuracy ceiling of one receiver working in isolation, and that ceiling sits at a meter or more.
RTK exists to break through that ceiling, and it does not do it by measuring more precisely. It does it by comparing. This is the traditional RTK mechanism specifically; PPP reaches similar precision through a very different route, using precise orbit and clock products instead of a nearby reference, and it will get its own companion later in this series.
RTK does not improve the raw satellite measurement. It removes error that is shared between two nearby receivers by comparing their observations directly, leaving only the small residual that actually describes the distance between them.
Errors That Travel Together
A patch of atmosphere overhead does not know or care which receiver is looking through it. If a base station and a rover are within a reasonable distance of each other, a satellite signal passing through the atmosphere on its way to one receiver experiences almost exactly the same delay on its way to the other. The same is true of small errors in the satellite's broadcast orbit and clock. These errors are common to both receivers, which means they can be subtracted away, rather than lived with.
A base station, occupying a fixed point, broadcasts both its own coordinate and its raw satellite observations. The rover receives that broadcast and differences the observations directly against its own, for the same satellites, canceling the atmospheric and orbital error the two receivers share, in real time.
The closer a rover is to its base or network reference, the more completely the shared error cancels, because the same atmosphere and the same small orbit errors are affecting both receivers almost identically.
What Is Actually Left Over
Once the shared error is removed, what remains is dominated by something GNSS is extremely good at measuring: the precise, whole-and-fractional-cycle difference in carrier phase between the base and the rover. Resolving that difference correctly, down to the exact whole number of wavelengths involved, is what an RTK FIX solution represents. It is why RTK positioning can reach centimeter precision even though a single autonomous receiver typically can't do much better than a meter or so on its own.
If the base and rover are both experiencing nearly the same atmospheric delay, does the atmosphere actually need to be measured accurately in an RTK solution? Not directly. It only needs to be common enough to cancel.
The Trade the Reader Should Notice
This design has a consequence worth flagging early, because it will return throughout this companion. RTK's precision depends entirely on having a trustworthy reference point nearby. If the base coordinate is wrong, or the correction source is describing a different reference frame than the rover expects, RTK will still resolve FIX and still report a precise number. Precise and correct are not the same claim, and this companion is going to spend a great deal of time keeping them separate.
RTK reaches high precision by canceling shared error between a rover and a reference, not by making any single measurement inherently better. That is powerful, but it also means RTK is only as trustworthy as the reference it is compared against.
That distinction, between a precise measurement and a correct one, comes down to what RTK is actually measuring against. The next chapter explains why carrier phase makes centimeter precision possible at all, and the one complication that comes with it.