4 The radio side and the architecture side

4.1 Two questions, two halves of 3GPP

Sensing splits cleanly into two questions, and 3GPP answered them in two different buildings.

The radio question is whether the signal can see the thing at all, and how well. The system question is who may ask for a sensing result, who is allowed to be told, and which part of the network does what.

Neither half is useful without the other, which is why the two sides ended up writing letters to each other.

4.2 The radio side started with the assumptions

The first radio study did not measure anything. It fixed the assumptions everyone else would measure against.

A channel model is the shared set of assumptions about how a radio signal travels, bounces and fades. Every radio simulation in 3GPP starts from the same one, so that two companies arguing about a result are at least arguing about the same world.

That model is TR 38.901, and it already existed. The Rel-19 study did not write a new report — it changed the old one: 26 documents name it and 23 of them are change requests TR 38.901.

Notice the last clause of that objective: distinguishing the object you want from the ones you do not. A network that reports every passing bird is not a sensing network.

4.3 Then it asked whether it works

The second radio study is the one that evaluates.

The parenthesis does the explaining: transmitter and receiver sitting in the same place. That is a monostatic arrangement, and it is the simplest one — the base station shouts and listens for its own echo.

Its report is TR 38.765, and 168 documents name it TR 38.765. The report states one assumption up front that a reader should not skip.

A waveform is the shape the radio signal is given. Naming a baseline waveform early is how a study stops itself re-arguing that choice in every contribution.

4.4 And now the radio rules

The rule-writing that followed sits with RAN3, the group that owns the interfaces between the parts of a radio network.

Read the second sentence slowly. A limit on impact, stated in the objective, is one of the things that decides how large a work item turns out to be.

Its target documents are TS 38.300, the overall description of the 5G radio network, TS 38.401, its architecture, and TS 38.473, the protocol between the two halves of a split base station TS 38.300 TS 38.401 TS 38.473.

4.5 The architecture side started with a question about the system

SA2 owns stage 2 — how the system is put together. Its study said what it was looking at.

Its report is TR 23.700-14, and 657 documents name it TR 23.700-14. It also introduced the two words that everyone now uses.

A Sensing Function is the logical function that supports the sensing service ^[7]. gNB-based sensing means the base station is the sensing entity ^[8].

The rule-writing that follows is Sensing-ARC, and it points straight back at the study.

"KI#1" is a numbered key issue of the study. This school's register holds no key issue text, so what KI#1 actually asks is in TR 23.700-14 and nowhere here. The specification being written from those conclusions is TS 23.137, and 262 documents name it already TS 23.137.

4.6 The two sides had to write to each other

38 documents in this area are liaisons ^[9]. Three of them the data marks as actually sent, and those three show the seam between the two halves.

SA2 sent S2-2511045, on aspects related to RAN coordination S2-2511045. RAN3 replied with R3-261541 R3-261541. And SA6 sent S6-254616 to announce that it had started a study of its own S6-254616.

Those letters are the cheapest thing to read in the whole area, and they are where a reader can see what one group was worried the other would get wrong.

4.7 Where to read this yourself

The radio half is TR 38.901 for the assumptions and TR 38.765 for the evaluation, with TS 38.300, TS 38.401 and TS 38.473 as the documents the rules will land in. The system half is TR 23.700-14 for the study and TS 23.137 for the specification.

This school does not carry the text of any of them. It says which one answers which question.