Quiz Entry - updated: 2026.07.14
Across 2G, 3G, and 4G, which generations separated voice and data into circuit-switched and packet-switched domains, and which moved to an all-packet core?
2G and 3G kept voice and data in separate core-network domains: voice was circuit-switched via MSC/GMSC, while packet data used SGSN/GGSN. 4G LTE collapsed this into an all-packet EPC, with voice carried as VoLTE packets or, in early deployments, via fallback to 2G/3G.
| Generation | Voice path | Data path | Core model |
|---|---|---|---|
| 2G GSM/GPRS | Circuit-switched via MSC/GMSC | Packet-switched via SGSN/GGSN after GPRS | Split CS + PS domains |
| 3G UMTS | Circuit-switched via MSC/GMSC | Packet-switched via SGSN/GGSN | Split CS + PS domains |
| 4G LTE | VoLTE over IP, or CSFB to 2G/3G if VoLTE is unavailable | IP packets through EPC | All-packet core |
2G/3G approach — two parallel domains:
- CS domain (Circuit Switched): BSC/RNC → MSC → GMSC → PSTN — for voice calls
- PS domain (Packet Switched): BSC/RNC → SGSN → GGSN → Internet — for data
4G LTE approach — all packet-switched:
- No more MSC — voice is carried as VoIP packets (VoLTE = Voice over LTE)
- Everything goes through the Evolved Packet Core (EPC)
- CSFB (Circuit-Switched Fallback): early 4G networks without VoLTE would fall back to 3G/2G for voice calls — you might have noticed your phone briefly dropping to "3G" when receiving a call
Why this matters:
- Simpler architecture — one network instead of two parallel ones
- More efficient — no resources wasted on dedicated voice circuits
- Enables new services — video calls, VoIP, all treated as just data
- Lower latency for voice — surprisingly, VoLTE actually sounds better than traditional GSM calls because it uses wideband audio codecs (HD Voice)
Go deeper:
GPRS core network (Wikipedia) — the SGSN/GGSN packet domain that ran in parallel with the CS voice core through 2G/3G.- Voice over LTE (Wikipedia) — how the all-packet EPC carries voice as IMS/IP packets, and the CSFB fallback used before VoLTE.