2009LTEOFDMASC-FDMAEPCeNodeBMMES-GWP-GWPCRFX2SAE

3GPP Release 8: 4G LTE — A Clean-Sheet Redesign

Release 8 is the most significant mobile standard since Rel-99. It discarded CDMA and TDM entirely and built 4G — LTE and the Evolved Packet Core — from a blank sheet. Nothing from the UMTS radio survived intact; even the flat network architecture was a deliberate break from the hierarchical Node B / RNC / SGSN stack.

Overview

Where Rel-4 through Rel-7 added speed to UMTS incrementally — HSDPA, HSUPA, HSPA+ — Rel-8 started over. The ITU had issued requirements for what it called IMT-Advanced, the successor to IMT-2000 (3G), and UMTS could not meet them regardless of how many enhancements were layered on. The 3GPP response was a two-track effort: LTE (Long Term Evolution) for the radio access side, and SAE (System Architecture Evolution) for the core network. Both were frozen in Rel-8 and together constitute what the world knows as 4G.

The radical choice was to abandon W-CDMA completely. LTE uses OFDMA on the downlink and SC-FDMA on the uplink — a different modulation family entirely. The core network also changed fundamentally: the SGSN / GGSN pair and the MSC are gone, replaced by four new all-IP nodes in a flat topology.

OFDMA — The New Radio (Downlink)

Orthogonal Frequency Division Multiple Access divides a 20 MHz channel into 1,200 active subcarriers (plus guard bands) spaced exactly 15 kHz apart. The defining property of OFDM: when one subcarrier is at its peak amplitude, every other subcarrier is precisely at zero — they are mathematically orthogonal. This eliminates inter-carrier interference, allowing subcarriers to be packed together without guard bands between them.

Resources are allocated in Physical Resource Blocks (PRBs): 12 subcarriers in the frequency domain multiplied by one slot (0.5 ms) in the time domain. The scheduler assigns PRBs to users every 1 ms Transmission Time Interval (TTI), selecting the best PRBs for each user based on their reported Channel Quality Indicator (CQI). This is called frequency-selective scheduling — exploiting the fact that different users experience deep fades at different frequencies at any given moment — and it is the fundamental reason LTE achieves significantly better spectral efficiency than HSPA.

Channel bandwidths: 1.4, 3, 5, 10, 15, and 20 MHz — the same OFDM numerology scales to all of them
Modulation: QPSK, 16-QAM, 64-QAM selected per PRB per TTI based on channel conditions
Peak DL rate: ~150 Mbps with 2×2 MIMO in a 20 MHz channel; ~300 Mbps with 4×4 MIMO

SC-FDMA — Uplink (Power-Efficient OFDMA)

SC-FDMA (Single Carrier FDMA) uses the same 15 kHz subcarrier grid as OFDMA, but transmits data using a pre-coding step (a DFT spread) that produces a single-carrier waveform before it is mapped onto OFDM subcarriers. The practical result is a dramatically lower Peak-to-Average Power Ratio (PAPR): where OFDMA might reach a PAPR of 10–12 dB, SC-FDMA stays around 2–4 dB.

This matters enormously in mobile devices. A high PAPR forces the power amplifier to operate well below its saturation point most of the time, which wastes battery power and drives up the cost of the amplifier component. SC-FDMA achieves the same spectral efficiency as OFDMA while keeping uplink power amplifiers affordable and extending device battery life. Peak uplink rate: approximately 75 Mbps with 1×2 MIMO in a 20 MHz channel.

EPC — The Evolved Packet Core

SAE (System Architecture Evolution) defined a completely flat, all-IP core with four key nodes — and no Circuit-Switched domain at all. Every packet, whether it eventually carries voice or data, travels as IP from the radio all the way to the internet.

MME — Mobility Management Entity

The MME is a signalling-plane-only node — it never touches user data. It handles UE attach and detach, authentication (querying the HSS), tracking area updates, paging idle UEs, and handover signalling between eNodeBs and between LTE and other RATs. Separating signalling from the data plane means the MME can be located anywhere in the network and scaled independently of data throughput.

S-GW — Serving Gateway

The Serving Gateway is the user-plane anchor within the PLMN. It routes user data packets between the eNodeB and the P-GW and maintains the data path during intra-LTE handovers — when a UE moves to a new eNodeB, the data tunnel to the S-GW stays constant and only the eNodeB-side endpoint changes. It also buffers downlink packets for UEs that are paging.

P-GW — PDN Gateway

The PDN Gateway is the internet gateway and UE IP address anchor. It assigns the UE an IP address (via DHCP/DHCPv6), enforces per-flow QoS rules received from the PCRF, performs NAT where required, and acts as the mobility anchor point when the UE moves between LTE and non-3GPP accesses such as Wi-Fi or cdma2000.

PCRF — Policy and Charging Rules Function

The PCRF is the real-time policy brain. When a session begins, the P-GW sends session information to the PCRF over the Gx interface. The PCRF responds with dynamic policy and charging rules: which traffic flows get what QoS class, what bandwidth limits apply, and how to charge for each flow. Rules can be updated mid-session — for example, upgrading a video stream's QoS when a user purchases a premium add-on.

LTE EPC architecture: two eNodeBs coordinate via the X2 interface. Each eNodeB connects to the MME (signalling) via S1-MME and to the S-GW (user data) via S1-U. The S-GW connects to the P-GW via S5; the P-GW is the internet gateway. The PCRF pushes QoS and charging policies to the P-GW and S-GW in real time.

eNodeB and the X2 Interface

In LTE, the base station is called eNodeB (evolved NodeB). The contrast with UMTS is stark. In UMTS, Node Bs were relatively dumb RF front-ends with all intelligence concentrated in the RNC. In LTE, eNodeBs are intelligent nodes that run the radio scheduler, manage HARQ retransmissions, handle inter-cell interference coordination, and control handovers — without any centralised controller above them.

eNodeBs connect directly to each other via the X2 interface — a direct peer-to-peer link that carries:

Fast handover coordination: the source eNodeB prepares the target eNodeB directly over X2 rather than routing signalling via the core network; handover latency drops to approximately 50 ms versus ~300 ms in UMTS
Inter-Cell Interference Coordination (ICIC): eNodeBs exchange information about their resource usage and load, allowing them to negotiate which PRBs to avoid at cell edges to reduce inter-cell interference
SON (Self-Organising Network) functions: Automatic Neighbour Relation (ANR), load balancing, and coverage optimisation all use X2 as a data channel

Each eNodeB connects to the core using the S1 interface — split into S1-MME (signalling, toward the MME) and S1-U (user data, toward the S-GW). A single eNodeB can connect to multiple MMEs and S-GWs simultaneously for resilience.

No CS Domain — The Voice Challenge

Rel-8 LTE has no Circuit-Switched domain at all. Every connection is a packet data session. This created an immediate practical problem: how do LTE subscribers make and receive ordinary phone calls? Three approaches were defined or emerged in practice:

CS Fallback (CSFB): the UE temporarily drops from LTE to 3G or 2G to make or receive a call, then returns to LTE when the call ends. Widely deployed for early LTE networks (2010–2014) because it reused existing 3G CS infrastructure
VoLTE (Voice over LTE): voice carried as an IMS (IP Multimedia Subsystem) session entirely over LTE, using the EPS Bearer QoS framework to guarantee the QoS class needed for voice. Standardised in Rel-10/11 deployment profiles; became mainstream from approximately 2015
OTT VoIP: applications such as Skype, FaceTime, or WhatsApp carrying voice as best-effort IP traffic, with no network-level QoS guarantee — acceptable for data but not operationally reliable in all conditions

Why Rel-8 Mattered

Spectral efficiency 3–5× better than HSPA: OFDMA frequency-selective scheduling and MIMO together deliver far more bits per Hz than W-CDMA-based systems could achieve
Latency under 10 ms: the flat architecture and removal of the RNC eliminated multiple processing hops; round-trip latency in LTE is typically 5–15 ms versus 50–100 ms in UMTS
Flat architecture reduced cost per bit dramatically: removing the RNC tier cut both capital expenditure and the operational complexity of managing a three-layer radio access network
eNodeB intelligence enabled self-organising networks: X2-based coordination and SON functions allowed LTE networks to partly configure and optimise themselves, reducing ongoing operating costs
TD-LTE opened China and unpaired spectrum: the same Rel-8 specification supports TDD mode; China Mobile built the world's largest LTE network on TD-LTE, demonstrating that the standard scaled to hundreds of millions of subscribers
LTE became the universal 4G standard: by 2020, LTE operated in virtually every country and served the majority of mobile data traffic globally — all of it tracing directly back to the specifications frozen in Rel-8