9 The architecture, phase by phase — and backhaul, which runs the other way

9.1 What an architecture group actually decides

What the system was asked to do ended with a requirement: the system shall be able to do such and such. Somebody then has to say which box does it, what each box has to know, and what messages pass between them.

That is SA2. It writes the model of the 5G system, the procedures that run over it, and the policy rules beside them — three documents that everything else in the core network is built against.

SA2 carries 3588 documents of this subject over 50 meetings. That is the largest group outside the radio groups, and it has sat on more separate meetings of this subject than any other group in it.

9.2 The pattern, once, so the phases read quickly

Every phase of SA2's satellite work has the same two steps.

• S1. A study item, short name beginning FS_, which produces a technical report numbered in the 23.7 range. Nobody has to obey a report.

• S2. A work item after it, which changes TS 23.501, the system architecture, TS 23.502, the procedures, and usually TS 23.503, the policy framework.

Those three carry 486, 214 and 69 documents of this subject TS 23.501 TS 23.502 TS 23.503. Almost every one is a change request rather than a discussion paper.

9.3 Phase 1: getting a satellite into the model at all

FS_5GSAT_ARCH is the Release 17 study, planned from 2018-06-06 to 2020-06-25, carrying 359 documents over 27 meetings FS_5GSAT_ARCH. Its report is TR 23.737, and its register entry records no parsed text, so this school cannot quote it.

5GSAT_ARCH is the work item after it: Release 17, sitting on two work plan rows, planned from 2018-06-06 to 2022-03-11, carrying 508 documents over 81 meetings 5GSAT_ARCH.

Its change table names three documents and says what it wanted from each — a functional description in the system model, updates to procedures, and an update of the policy framework.

Eighty-one meetings for one work item is the number to hold on to. Its documents run from meeting RP-88-e to meeting R3-133, so the conversation reached the radio plenary and the radio interface group and was not SA2's alone 5GSAT_ARCH.

9.4 Phase 2: coverage that comes and goes

Phase 2 is the one phase with no requirements item beside it, and it is about a problem the first phase could ignore: a satellite that is not always there.

FS_5GSAT_Ph2 is the Release 18 study, planned from 2021-12-20 to 2022-12-12, carrying 193 documents over 18 meetings, producing TR 23.700-28 FS_5GSAT_Ph2.

5GSAT_Ph2 is the work item after it, and it is much bigger than its study: 810 documents over 57 meetings ^[10], sitting on four work plan rows, with CT1, CT6 and SA2 all named in its groups 5GSAT_Ph2. Its objective says what it was for:

Read the two named systems in that sentence. Not only the 5G core but the older packet system as well, and its change table names TS 23.401, the older packet core, beside the 5G documents 5GSAT_Ph2. Why the older system was taken in too is not recorded anywhere this register reads.

9.5 Phase 3: a base station in orbit, and messages that wait

FS_5GSAT_Ph3_ARCH is the Release 19 study, planned from 2023-09-20 to 2024-06-06, carrying 442 documents over 20 meetings, producing TR 23.700-29 FS_5GSAT_Ph3_ARCH. That report's scope names what it went after first:

A regenerative payload is one that does real work in orbit rather than only relaying: it transforms and amplifies the signal coming up before sending it down again, and the transforming may include demodulating, decoding, re-encoding, re-modulating or filtering ^[11].

Put plainly, part of the base station moves off the ground.

5GSAT_Ph3-ARC is the work item after it: Release 19, planned from 2024-06-12 to 2024-12-12, carrying 533 documents over 24 meetings 5GSAT_Ph3-ARC. Its objective names the second of the study's key issues outright:

Its change table names six documents 5GSAT_Ph3-ARC: the three core ones, the older packet core TS 23.401, the telephony subsystem TS 23.228 and the document that governs how the network tells outside applications what is happening TS 23.682.

9.6 Phase 4: a phone call through a satellite that stands still

FS_5GSAT_Ph4_ARC is the Release 20 study, planned from 2025-03-26 to 2026-03-03, carrying 579 documents over 28 meetings ^[12]. Its justification says where the pressure came from and does not pretend the ground was fresh:

Four releases of study behind it, and the requirements group writing new use cases while the architecture study was being proposed. Its objective names its first work task:

Unpacked: a voice call, carried by the telephony subsystem, over a geostationary satellite, on the narrowband internet-of-things radio, into the older packet core. Every one of those choices is the awkward one, which is why it is a study and not an assumption.

The report it produced, TR 23.700-19, opens its scope on exactly the same task:

5GSAT_Ph4-ARC is the work item turning that into rules: Release 20, planned from 2025-12-03 to 2026-09-09, carrying 204 documents over six meetings, with a completion value of 0.8 5GSAT_Ph4-ARC. Its change table names seven documents, more than any other SA2 item here.

9.6.1 Two words phase 4 defines for itself

TR 23.700-19 defines two phrases in its own terms clause, and both are worth learning because they turn up in the CT work as well.

The second phrase is not only a definition.

Phase 4's change table names it against three documents at once — the system model, the procedures and the policy framework — each entry saying the change is for UE-Satellite-UE communication with a user plane function carried on board the satellite 5GSAT_Ph4-ARC.

9.7 Every SA2 item on one page

Four phases, each a study and a work item, plus backhaul and one late addition.

Item	Release	Documents	Completion	What it produced
FS_5GSAT_ARCH	Rel-17	359	1	TR 23.737
5GSAT_ARCH	Rel-17	508	1	edits to three documents
FS_5GSAT_Ph2	Rel-18	193	1	TR 23.700-28
5GSAT_Ph2	Rel-18	810	1	edits to four documents
FS_5GSAT_Ph3_ARCH	Rel-19	442	1	TR 23.700-29
5GSAT_Ph3-ARC	Rel-19	533	1	edits to six documents
FS_5GSAT_Ph4_ARC	Rel-20	579	0.84	TR 23.700-19
5GSAT_Ph4-ARC	Rel-20	204	0.8	edits to seven documents
FS_5GSATB	Rel-18	111	1	TR 23.700-27
5GSATB	Rel-18	208	1	edits to three documents
TEI20_QoS_MWAB_SAT	Rel-20	0	0.28	edits to one document

The completion column cuts each of the work plan's long decimals at two places and changes nothing else. Three rows stand below 1, and all three are Release 20.

Two shapes are worth pulling out of that table.

The work item is heavier than its study in three of the four phases — 359 against 508, 193 against 810, 442 against 533. That is the opposite of what What the system was asked to do found on the requirements side, where phase 3 and phase 4 both had the heavier study.

Both sets of numbers were counted the same way, so the contrast is real rather than an artefact.

Phase 4 is the exception so far, at 579 against 204 — and its work item is the one row still standing below 1, so the comparison is between a finished count and a running one.

9.8 Where each phase's edits land

The three system documents take the weight, but not evenly.

• TS 23.501, the system model, carries 486 documents of this subject and 483 change requests. Five work items name it in their own change table, and the register also lists eight work items whose documents carry it.

• TS 23.502, the procedures, carries 214 documents, every one a change request. Seven work items name it — every phase from 5GSAT_ARCH to 5GSAT_Ph4-ARC, plus backhaul.

• TS 23.503, the policy framework, carries 69, again all change requests, and is named by four items.

• TS 23.401, the older packet core, carries 570 documents. Four of these work items name it, and that is where the internet-of-things satellite work lives rather than in the 5G core.

The last of those is the one people miss. A subject that presents itself as 5G work has one of its heaviest documents in the previous generation's core, and three of the four work items naming it are the satellite phases themselves.

9.9 What SA2 hands to somebody else

Most of these work items carry a field naming the other groups they expect to depend on, and reading those fields down the list shows where the boundaries of an architecture group actually are.

• Phase 2's study says security goes to SA3, charging to SA5, and radio issues to RAN FS_5GSAT_Ph2. Phase 3's study says the same three things in the same order FS_5GSAT_Ph3_ARCH.

• Phase 3's work item goes further and names lawful interception alongside security, management alongside charging, and says the support of regenerative satellite access is due to be developed in RAN3, with SA2 aligning to it if needed 5GSAT_Ph3-ARC.

• Phase 4's work item says the voice and emergency call over the narrowband internet-of-things radio is due to be developed in the RAN groups and low data rate in SA4, keeping security for SA3 and charging and management for SA5 5GSAT_Ph4-ARC.

Read together, those fields are a description of who does what — including who does the parts the architecture group does not.

The meeting ranges say the same thing from the other side. Phase 1's work item has documents from meeting RP-88-e to meeting R3-133 — a radio plenary at one end and a radio interface group at the other 5GSAT_ARCH.

Phase 4's study runs from S2-168 to C1-162, so its newest papers sit at a protocol working group's meeting while the study row still stands below 1 FS_5GSAT_Ph4_ARC.

9.10 Backhaul, which points the other way

Everything above assumes the satellite reaches the phone. Satellite backhaul is the same technology turned around: an ordinary mast on the ground serves the phone, and the satellite carries the mast's connection back to the network.

FS_5GSATB is the Release 18 study, planned from 2021-12-20 to 2023-03-03, carrying 111 documents over 11 meetings, producing TR 23.700-27 FS_5GSATB. That report's register entry records no parsed text either, so it can be named and not quoted.

5GSATB is the work item: Release 18, sitting on five work plan rows, carrying 208 documents over 32 meetings, with CT3, CT4, SA1 and SA2 all in its groups 5GSATB. Its objective is the shortest of the SA2 set:

Its change table names the three core documents, and its wider list of documents touched runs to twelve — the protocol documents of the core network among them 5GSATB.

Release 20 adds one more piece.

TEI20_QoS_MWAB_SAT is about quality of service where a mobile wireless access backhaul runs over satellite: planned from 2026-03-24 to 2026-12-12, no document yet, naming TS 23.502 for one change about identifying which kind of satellite backhaul is in use TEI20_QoS_MWAB_SAT.

Its protocol half is TEI20_QoS_MWAB_SAT-CT, which is in What the phone says to the network.

9.11 Reading the four reports in order

If you have to work with this half of the subject, the four architecture study reports in order are the shortest route in. Two of them this school can quote and two it cannot.

• A1. TR 23.737 — phase 1, getting satellites into the model. No parsed text here.

• A2. TR 23.700-28 — phase 2, coverage that comes and goes. No parsed text here either.

• A3. TR 23.700-29 — phase 3, the regenerative payload and store and forward. Quotable, and quoted above.

• A4. TR 23.700-19 — phase 4, the voice call over a geostationary satellite. Quotable, and the source of both definitions above.

Then TS 23.501 and TS 23.502 for what actually became binding.

9.12 Where this sits in the rest of the course

• R1. What the phone says to the network — the CT groups, which turn every one of these decisions into messages a device sends.

• R2. What the system was asked to do — the requirements each of these phases was answering.

• R3. The overview course covers the four phases in one chapter, and the quick-start course mentions only that the architecture work exists.