A Coordinate Needs a Reference Frame
It would be convenient if there were exactly one ECEF coordinate system, permanently agreed upon, so that a set of XYZ values meant the same physical point everywhere, forever. That is not quite how it works. Several distinct realizations of an Earth-centered, Earth-fixed system exist at once, and a coordinate is not complete until it names which one it belongs to.
A reference frame is not something added to a coordinate partway through processing. It is present from the very first instant a position is solved, inherited from the satellite orbit data and the base station's own known coordinate, whether or not anyone ever writes its name down.
The Frame Was Never Missing, Only Unlabeled
It is tempting to imagine a coordinate starting out generic and later being stamped with a reference frame, the way a package might be stamped with a shipping label somewhere along its route. That picture is backwards. The broadcast ephemeris a satellite transmits is already expressed in a specific reference frame. The base station's coordinate, whatever it is being compared against, was already surveyed in a specific frame long before the rover ever turned on. The rover's solved position inherits that same frame automatically, the moment RTK resolves. Nothing is added later. What sometimes happens later is a deliberate transformation, converting a coordinate that already had a frame into a different frame entirely, and that transformation is the only step that is genuinely optional.
Ask not when a reference frame gets attached to a coordinate. Ask instead which frame was already present from the start, and whether anything downstream converted it into a different one.
Why More Than One Frame Exists at All
A reference frame is a practical, physical realization of a reference system, built from the measured coordinates of a network of real tracking stations at a particular moment. Because the Earth's crust genuinely moves, because measurement techniques keep improving, and because different regions have different practical needs, more than one realization has been built over time. WGS84 underlies GPS broadcasts. ITRF, refined periodically as tracking improves, serves as an international scientific standard. NAD83, in its various realizations, underlies most everyday United States surveying and mapping work. Each is a genuine, carefully constructed reference frame. None of them is simply wrong. They differ because they were built to serve different purposes, at different times, from different networks.
WGS84, ITRF, and NAD83 are not the same reference frame under different names, and none of them is the single correct one. They are separate, deliberately constructed realizations, close to one another but not identical, and a coordinate is only meaningful once its frame is known.
A Small Gap With a Real Cause
Two receivers occupying the same physical monument, each reporting a coordinate in a different reference frame, will disagree by a small but very real amount, often on the order of a meter or more depending on the frames involved. That gap is not measurement error in the usual sense. It exists because the two frames themselves are not perfectly aligned with each other, and no amount of additional precision from either receiver will close it. Only a documented transformation between the two frames will.
Two RTK rovers occupy the same monument. One reports a coordinate in NAD83. The other reports a coordinate in WGS84. If both receivers are working perfectly, should the numbers match exactly? No. The frames themselves are close but not identical, and the small gap between them is expected, not a sign that either receiver failed.
A reference frame is present in a coordinate from the moment it is solved, inherited from the satellite data and the base coordinate, not attached afterward. Several valid frames exist simultaneously, and comparing coordinates from two different frames without a transformation is comparing two different, if closely related, descriptions of the same point.
A reference frame answers which physical realization a coordinate belongs to. It does not yet answer a question this series has not raised before: when that realization was measured, because the ground it was built from does not hold still either.