5G Evolution &
Network Vision
From the all-IP world of LTE to a service-driven, cloud-native 5G System — and why the road through 5G-Advanced inevitably leaves the ground and goes non-terrestrial.
Four decades to reach one millisecond
Each generation solved one bottleneck: analog noise, digital capacity, mobile data, all-IP throughput. 5G NR is the first radio designed not for a single service — but as a configurable toolbox where latency, rate and density are dialed per use case.
Dial μ and watch the radio reshape itself
Why scalable numerology exists — and why μ is a compromise
Subcarrier spacing is a three-way fight. The cyclic prefix must outlast the channel's delay spread — rural macro echoes arrive microseconds late, so wide-area coverage wants 15 kHz with its generous 4.69 μs CP. Phase noise grows brutally with carrier frequency — mmWave oscillators smear narrow subcarriers into each other, so FR2 demands 120 kHz. And the slot is the scheduler's heartbeat: it shrinks as 1/2μ, so URLLC pushes μ up to cut latency. NR refuses to crown one winner: μ is signalled per Bandwidth Part, so one gNB can serve 30 kHz eMBB and 120 kHz URLLC side by side on the same carrier.
The NTN twist (Day 3): at 600 km altitude even the longest CP is irrelevant against ±25 ms of propagation — satellite geometry is absorbed by timing advance, not by the cyclic prefix. Rel-17 NTN in S-band runs μ = 0/1, exactly the numerologies you just dialled.
One network, three personalities
ITU-R IMT-2020 defines three usage scenarios pulling the design in opposite directions. Click a corner of the triangle — every 5G feature you'll meet in this course exists to serve one of these corners.
From triangle corners to network slices
The triangle isn't marketing — it's encoded in the protocol. Every PDU session carries an S-NSSAI identifying its slice: SST 1 = eMBB, SST 2 = URLLC, SST 3 = MIoT (the mMTC corner), SST 4 = V2X, SST 5 = HMTC. The NSSF selects the slice, the PCF attaches policy, and each QoS flow is stamped with a 5QI whose standardized table row literally writes the corner's KPIs into every scheduler on the path — 5QI 1 for voice, 5QI 9 for best-effort eMBB, 5QI 82–85 for deterministic URLLC. One physical network, N logical ones.
Where this course goes (Day 4): slice continuity across TN and NTN — keeping an SST-2 SLA alive while the user plane rides a satellite. That's why the gauges above matter: a slice is a contract, and orbit changes what you can promise.
NSA or SA — who anchors the control plane?
The single most important question of any 5G rollout. In NSA Option 3x the LTE eNB stays master and 5G NR rides as a secondary cell on the old EPC. In SA Option 2 the gNB talks natively to the 5G Core — unlocking slicing, sub-ms QoS flows and, later, NTN. Flip the switch and watch the control plane (dashed red) move.
NSA · Option 3x (EN-DC) FAST TO MARKET
- LTE eNB is master — RRC anchor and mobility stay on 4G; NR adds a data-boost leg.
- Reuses the deployed EPC; UE attaches with legacy NAS — no 5G registration.
- In 3x, S1-U lands on the gNB, which splits the bearer over X2 toward LTE when needed.
- No network slicing, no 5QI flows, no NTN — capabilities are capped by the EPC.
- How most operators launched "5G" in 2019–2021.
SA · Option 2 FULL 5G SYSTEM
- gNB connects natively to 5GC via N2 (NGAP control) and N3 (GTP-U data).
- 5G NAS over N1: true 5G registration, PDU sessions, 5QI QoS flows.
- Unlocks network slicing (NSSF/S-NSSAI), edge breakout, RedCap, precise positioning.
- NTN requires SA — Rel-17 satellite access is specified against the 5G System, not EPC.
- This course's labs (Open5GS + UERANSIM) run pure SA.
Why SA is the ticket to space
Separating control from user plane is 3GPP's longest-running refactor: CUPS in Rel-14 split the S/P-GW, and the 5GC finished the job — the AMF owns NAS signaling, the UPF owns packets, and the two scale independently. NTN leans on that separation hard. Rel-17 gives NAS its satellite patch: extended timers in TS 24.501, because a GEO round trip of ~541 ms would blow through terrestrial retransmission timers; new satellite RAT types — NR(LEO), NR(GEO) — visible to the AMF so the core knows it is serving an orbit, not a street; and per-orbit policy so an operator can treat a LEO PDU session differently from a fiber-fed one.
None of this exists in EPC/NSA. That single fact decides the architecture of every NTN deployment you will build in this course: it's SA, always.
Rel-15 → Rel-20: watch NTN grow up
3GPP ships 5G in releases, each frozen roughly every 18–24 months. Track the violet 🛰 chips — NTN enters as a study item, becomes a Rel-17 work item, and by Rel-19/20 satellites are first-class citizens on the road to 6G.
Why 5G-Advanced leaves the ground
Terrestrial networks chase population, not area. Towers need power, backhaul and economics — oceans, deserts, airspace and disaster zones have none of the three. The only radio site that covers them all is in orbit.
Fly the satellite, feel the physics
Same question, four paths — who answers first?
The physics floor no release can patch
Light covers 299,792 km every second — and that is the only spec 3GPP cannot rewrite. The lab above runs the true geometry: the slant range at 10° elevation follows d = RE(√((RE+h)²/RE² − cos²ε) − sin ε), and a transparent payload pays that distance four times per round trip (UE→satellite→gateway, then back). At 600 km that lands on 25.77 ms — matching TR 38.821 Table 4.2-2 to the hundredth of a millisecond. Every Day-3 mechanism — Koffset, extended HARQ timers, stretched RACH response windows — is engineering wrapped around these few milliseconds.
And the reason TCP hates satellites (Day 4): the bandwidth×delay product balloons while slow-start still creeps up one RTT at a time — a GEO link spends most of a short transfer just waiting for ACKs.
Space layer
LEO mega-constellations, MEO and GEO. Rel-17 transparent payloads today; Rel-19 puts the gNB on board with inter-satellite links.
Aerial layer
HAPS at ~20 km and UAV platforms — quasi-stationary, low delay, rapid deployment over disasters and events.
Ground layer
Classic gNBs and small cells where density pays. In 6G's IMT-2030 vision all three layers fuse into one 3D network with seamless TN↔NTN mobility.
Prove it — 5 questions
Instant feedback with 3GPP-grounded explanations. Score 4+ to consider Lesson 1.1 mastered.
Carry these into Lesson 1.2
5G NR is a configurable radio toolbox — scalable numerology, lean carrier and flexible slots let one standard serve eMBB, URLLC and mMTC.
NSA Option 3x anchors RRC on LTE over EPC; SA Option 2 connects gNB↔5GC via N2/N3 and unlocks slicing, 5QI flows — and NTN.
The NTN thread runs Rel-15 (study) → Rel-17 (normative NR-NTN + IoT-NTN) → Rel-18/19 (Ka-band, regenerative, store-and-forward) → Rel-20/6G.
Terrestrial networks cover people, not the planet: most land and all oceans are dark. LEO at ~600 km brings RTT down to ~26 ms — good enough to merge with 5G.