5 RAN1 and the radio problem

5.1 The group that had to say whether it works

Everything in the other chapters of this course is somebody deciding what sensing should do. RAN1, the physical-layer group, had the one question that could have stopped all of it: can a 5G radio signal actually be read back well enough to find something?

RAN1 has 872 documents on this subject across 16 meetings, from RAN1#116 to RAN1#126 RAN1. That is more meetings than any other group spent on it, and it is the shape of a question that took a long time to settle.

The group led both radio studies — FS_Sensing_NR and FS_Sensing_NR_bis — and it did them in that order for a reason worth understanding, because it is the opposite order from the one a newcomer expects.

5.2 First the model, not the mechanism

The first study did not try to build sensing. It tried to agree how sensing would be simulated.

A channel model is the shared description of what radio does between a transmitter and a receiver — how it fades, reflects and scatters. In 3GPP that model is TR 38.901, and it is the assumption set every radio simulation in the organisation starts from TR 38.901.

Without a sensing version of it, two companies simulating the same idea would get different answers and neither could be checked. So the first study's product was not a sensing feature; it was 23 change requests into the channel model, out of 26 sensing documents naming it TR 38.901.

The objective goes on to list example objects the model must handle and distinguish from unintended ones. That list is not carried here — the register holds titles, scopes and objectives, and this objective's list runs on past where the register stops. It is in RP-242348 RP-242348.

The gap it was filling is stated plainly in the same document: positioning of a connected device already works, and what is missing is any built-in way to detect an object that never connected ^[4].

5.3 Why telling things apart is the hard part

The one phrase in that objective worth reading twice is "to enable them to be distinguished from unintended objects".

Detecting that something is there is not the problem. A world full of buildings, vehicles, trees and weather returns signal all the time. The problem is telling the thing you were asked about from everything else that also bounced the signal back, and that is a question about the model as much as about the equipment.

That is the whole reason the model had to be agreed before anything else, and it is why the first study's output is a set of changes to a document that nobody outside radio simulation ever reads.

5.4 The trail of the first study

• RP-234069 opened it at RAN#102 RP-234069.

• RP-240799 revised it at RAN#103 RP-240799.

• RP-242348 revised it again at RAN#105, and that is the revision the work plan still points at RP-242348.

The acronym FS_Sensing_NR carries 776 documents, more than any other acronym in this area ^[5]. Its target was a single document: modifications to TR 38.901 for the sensing aspects of the channel model FS_Sensing_NR.

TR 38.901 now stands at 19.4.0 TR 38.901 — a Release 19 document, and the sensing work in it is finished.

5.5 Then the measurement

With a model agreed, the second study could ask the real question. It is Release 20, RAN1 leads it and RAN3 is named second FS_Sensing_NR_bis.

Three narrowings are packed into that sentence, and all three survive into the specification being written today.

• gNB-based — the base station is the thing doing the sensing, not the phone ^[6].

• Mono-static — a single radio point where the transmitter and the receiver sit together, so the same equipment sends and listens. The other arrangement, where they sit apart, is called bistatic.

• UAV — drones, and only drones.

The reasoning the study gives for the whole idea is bare: the same radio signal used for sensing and for communication at the same time ^[7]. It names no benefit — no spectrum saving, no lower delay, no higher reliability — and neither should anybody quoting it.

The report it produced, TR 38.765, states one assumption in its own scope that is worth carrying away.

The baseline waveform is the one the network already transmits to carry data. That is the "integrated" in integrated sensing and communication, stated as a working assumption rather than as a slogan.

5.6 The trail of the second study

• RP-251861 opened it at RAN#108 RP-251861.

• RP-252819 revised it at RAN#109 RP-252819.

• RP-253246 revised it at RAN#110 — the revision the work plan now points at RP-253246.

• RP-261477 put the finished report to the plenary at RAN#112, where it was approved RP-261477.

FS_Sensing_NR_bis carries 407 documents ^[8] and TR 38.765 stands at 20.0.0, 168 documents having named it TR 38.765. The RAN plenary as a whole carries 53 sensing documents, which is where every one of those decisions was taken RAN.

5.7 Two studies, and why there are two

A reader meeting FS_Sensing_NR and FS_Sensing_NR_bis for the first time usually assumes the second is a revision of the first. It is not.

They are separate work items with separate report documents, separate acronyms, separate opening documents and separate approvals. The first is Release 19 and produced changes to TR 38.901; the second is Release 20 and produced TR 38.765 FS_Sensing_NR FS_Sensing_NR_bis.

The _bis in the second name marks a second piece of work on the same subject, not a second version of the first piece. Confusing the two makes the document counts nonsense: 776 documents belong to the channel-model study and 407 to the radio study, and they are not the same 776 and 407.

Those two counts stand against 3492 documents in the area altogether ^[9], and neither work item produced a specification anybody has to obey.

5.8 What this course cannot tell you about the radio

This is the chapter where the limits of the record bite hardest, so they are worth saying flatly.

The counts are floors as well: only documents carrying a sensing acronym were counted, and 440 documents whose titles say "integrated sensing" or "ISAC" carry none ^[10].

5.9 Where to read the real thing

Two documents, and this school carries the text of neither.

• TR 38.901, the channel model, for what the Rel-19 study actually changed TR 38.901.

• TR 38.765, the study report, for the evaluation and every number in it TR 38.765.

For the promises rather than the results, read the study item descriptions themselves: RP-242348 for the first RP-242348 and RP-253246 for the second RP-253246.

Next: why the rule-writing that follows this work sits in a different group — RAN3, and where the radio rule-writing went.