Cellular Networks & Mobility
WiFi covers a room or a building. Cellular networks cover a country. They solve the same core problem - carrying data over the air - but at a scale that demands a completely different architecture built around handing devices off between towers as they move.
Cellular architecture
The name "cellular" comes from how coverage is organized:
- Cells - the coverage area is divided into many small regions, each served by one base station. Adjacent cells reuse frequencies carefully so they do not interfere, which lets the same limited spectrum serve a huge area by using it over and over in cells that are far enough apart.
- Base stations - the tower and radio equipment for a cell. A phone always talks to one base station at a time (the one with the best signal) and is handed to a neighbor as it moves. In LTE the base station is called an eNodeB; in 5G, a gNodeB.
- Core network - the wired backbone behind the towers. It authenticates subscribers, tracks which cell each device is in, enforces billing and data limits, and connects the whole system to the public internet. In 4G this is the Evolved Packet Core (EPC).
The base stations handle the radio; the core network handles identity, mobility, and the gateway to the internet.
How a phone attaches
When a phone powers on, it does not simply "connect to WiFi." It runs an attach procedure:
- Scan and select a cell by listening for the strongest base-station signal.
- Authenticate using the credentials stored on the SIM card, which identifies the subscriber to the carrier's core network.
- Register with the core network, which records the phone's location (which cell it is in) so incoming calls and data can be routed to it.
- Establish a bearer - a data channel between the phone and the internet gateway - and the phone receives an IP address and can exchange traffic.
From then on the core network always knows roughly where the phone is, so it can deliver traffic and switch cells as the user moves.
4G LTE vs 5G
4G LTE made mobile data fast and all-IP - voice, video, and web all ride the same packet network. 5G pushes throughput far higher and latency much lower, and adds high-frequency millimeter-wave (mmWave) spectrum for enormous capacity in dense areas. mmWave is fast but short-range and easily blocked, so 5G also runs on lower bands for coverage.
| Property | 4G LTE | 5G |
|---|---|---|
| Peak throughput | ~1 Gbps | ~10-20 Gbps |
| Typical latency | 30-50 ms | 1-10 ms |
| Spectrum | Sub-6 GHz | Sub-6 GHz + mmWave (24 GHz+) |
| Device density | Thousands per km² | ~1 million per km² (IoT scale) |
| Core network | EPC (4G core) | 5G core (service-based, cloud-native) |
The low latency and huge device density are what make 5G interesting beyond fast downloads - they open the door to real-time applications and to connecting vast numbers of low-power IoT sensors that 4G could not economically support.
Mobility: staying reachable while moving
The defining feature of wireless networks - cellular most of all - is that endpoints move. A phone crosses dozens of cells on a train; a laptop roams between WiFi access points. Keeping a device reachable and its connections alive while it physically relocates is mobility management, and it forces the network to answer one hard question: how do you deliver traffic to something whose location keeps changing?
The mental model is a home network (where a device permanently belongs) and a foreign network (wherever it currently is). The home network always knows how to reach the device: when it roams, the home network tracks which foreign network it is currently visiting and helps forward traffic there. This home/foreign split is the backbone of mobile IP and exactly what the cellular core does for every phone.
A handoff (or handover) is what makes that tracking work in real time. As a device moves out of one coverage area and into the next, the network:
- Continuously measures signal strength to the current and neighboring base stations.
- Decides to switch when a neighbor becomes reliably stronger (with hysteresis, so it does not flip back and forth at the boundary).
- Transfers the device's session - its authentication state and active data bearers - to the new station.
A good handoff is seamless: packets keep flowing and the user notices nothing.
To make that hysteresis concrete: a common LTE handover trigger (the "A3 event") fires only when a neighboring cell's measured signal (RSRP) exceeds the serving cell's by a margin of roughly 2-3 dB, and only if that condition holds continuously for a time-to-trigger of around 320-640 milliseconds. So a phone sitting almost exactly at a cell boundary, where signal strength jitters above and below the crossover every second, still will not hand off - the combination of the margin and the sustained timer filters out that jitter and prevents the "ping-pong" of repeatedly re-registering with two towers a few meters apart.
The catch is that a device's IP address is usually tied to the network it is attached to, and every open TCP connection is keyed on that address. If a moving device got a new IP every time it changed cells, every connection would break instantly. Mobility schemes hide this: the device keeps a stable identity (a home address, or a session anchor in the cellular core) while the network reroutes traffic to wherever it currently is - the same indirection that lets a video call survive walking from one room to another, or from WiFi to cellular.
TCP assumes packet loss means congestion - a router queue overflowed - so its response to loss is to slow down. On a wireless link that assumption is often wrong: packets are lost to fading, interference, or a handoff gap, not congestion. TCP cannot tell the difference, so it needlessly throttles a connection whose path was never actually congested. Real deployments add link-layer retransmission and forward error correction below TCP to hide wireless loss before TCP ever sees it - a case where a lower-layer reality (unreliable, moving radio links) leaks up and breaks an assumption baked into a higher layer.