Connection Types and Power over Ethernet
Not every link is built the same way. Some connect exactly two devices and give them the whole wire; others let a crowd share the medium and take turns. This lesson draws that distinction, tours the four physical connection types you have now met, and then shows how the same Ethernet cable that carries data can also carry electricity — Power over Ethernet, the reason a ceiling access point or desk phone needs no mains socket.
Half duplex and full duplex
Duplex is the direction rule for a link.
- Half duplex — a device can send or receive, but not both at the same instant. This is the natural mode for a shared medium, because two devices must take turns. Half-duplex links rely on collision detection, and they can never exceed the medium’s single base rate no matter how many devices are present.
- Full duplex — a device sends and receives at the same time, because the two directions use separate transmit and receive paths. Full duplex requires a dedicated point-to-point link, and it is the default on modern Ethernet switch ports.
Because full-duplex links do not share the wire, collisions are impossible on them and CSMA/CD is switched off. That is why a healthy modern Ethernet port should report zero collisions — a fact you will lean on hard in Lesson 8, when collisions and late collisions become the signature of a duplex mismatch.
The four connection types
You met the media individually earlier in the week; here they are as connection types, with the trade-offs that decide which one a designer reaches for.
| Type | Medium | Typical reach | Best for |
|---|---|---|---|
| Ethernet (IEEE 802.3) | Copper twisted pair (UTP/STP), RJ45 | Up to 100 m | LAN edge: PCs, phones, APs, cameras; the default access link |
| Fibre optic | Glass/plastic strand, light | Hundreds of metres to many kilometres | Backbones, building-to-building, high EMI areas, long runs |
| Coaxial | Single copper core in a shield | Varies; often hundreds of metres | Cable broadband (DOCSIS), older Ethernet (10BASE2/5) |
| Wi-Fi (IEEE 802.11) | Radio (2.4/5/6 GHz) | Tens of metres indoors | Mobility: laptops, phones, tablets, IoT |
Ethernet over twisted pair is the workhorse of the wired LAN. Its 100 metre limit, its cheapness, and its support for Power over Ethernet make it the natural access connection. Fibre trades cost and complexity for enormous distance and total immunity to electromagnetic interference; it is the standard choice for the links between switches and buildings. Coaxial survives today mainly as the medium your cable modem uses. Wi-Fi is the only connection type that is fully shared at the physical layer, and it is chosen whenever users must be free to move.
What Power over Ethernet is
Power over Ethernet (PoE) delivers direct-current electrical power and Ethernet data over the same twisted-pair cable. It exists because the devices we hang in awkward places — a wireless access point on a ceiling, a camera on an eave, a phone on a desk — are exactly the devices for which running a separate electrical socket is expensive or impossible. PoE means the installer runs one cable and both jobs are done.
Two roles define every PoE link:
- The PSE (Power Sourcing Equipment) is the side that supplies power. Usually this is a PoE-capable switch, though a standalone midspan injector placed between a plain switch and the device can add power too.
- The PD (Powered Device) is the side that consumes power — the access point, IP phone, IP camera, or sensor.
Power is injected using what engineers call phantom power: the DC voltage rides on the same wires as the data without disturbing the signal. There are two wiring schemes. Alternative A uses the same pairs that carry the data (pins 1–2 and 3–6), so the power and data share conductors via centre taps. Alternative B uses the unused spare pairs (pins 4–5 and 7–8). Devices negotiate which scheme to use automatically. In endspan deployment the switch itself is the PSE; in midspan deployment a separate injector sits mid-cable.
PoE standards: af, at, and bt
The IEEE has published three generations of PoE, each raising the power ceiling. Learn the tiers and their headline numbers — they are fair game on the exam.
| Standard | Marketing name | PSE port power | Power at the PD | Pairs used |
|---|---|---|---|---|
| IEEE 802.3af | PoE | up to 15.4 W | about 12.95 W | 2 pairs |
| IEEE 802.3at | PoE+ | up to 30 W | about 25.5 W | 2 pairs |
| IEEE 802.3bt (Type 3) | PoE++ / 4PPoE | up to 60 W | about 51 W | 4 pairs |
| IEEE 802.3bt (Type 4) | PoE++ / 4PPoE | up to 90 W | about 71.3 W | 4 pairs |
Notice that the power at the PD is always lower than the power at the PSE port: some energy is lost as heat in the cable itself, and the longer the cable, the greater the loss. That gap is why the standards quote two figures. 802.3af comfortably runs a basic IP phone or a legacy access point. 802.3at (PoE+) was created for devices such as dual-radio wireless access points and pan-tilt-zoom cameras. 802.3bt uses all four pairs to push even harder, powering high-end APs, video displays, PoE lighting, and thin clients.
Power budgets, classes, and where PoE is used
A PoE switch cannot give every port its maximum indefinitely; it has a finite power budget set by its power supply. A 24-port switch with a 370 W budget, for example, cannot run 24 ports at 30 W (that would need 720 W). The switch therefore reserves power per port according to the device’s class and refuses to power up a port if the budget is exhausted. This is a real design constraint: an under-budgeted switch may simply refuse to power the last few access points.
| Class | Typical device | Reserved at port |
|---|---|---|
| Class 0 | Default / unclassified | 15.4 W |
| Class 1 | Very low power | 4 W |
| Class 2 | Low power | 7 W |
| Class 3 | Typical phone, basic AP | 15.4 W |
| Class 4 | PoE+ device | 30 W |
| Class 5–8 | 802.3bt high-power devices | Up to 90 W |
Negotiation happens in two steps. Hardware classification uses the electrical signature and a class resistor, working down to the device even before it boots. Finer-grained negotiation — down to fractions of a watt — uses LLDP or CDP, the Layer 2 discovery protocols, to carry a power request and grant. The 802.3bt amendment adds Autoclass, which measures actual consumption so the switch can hand unused power back to the budget.
Where PoE is used
PoE is everywhere a small, network-connected device would otherwise need a power outlet in a difficult spot:
- Wireless access points — usually ceiling-mounted, powered straight from the access switch. A dual-radio Wi-Fi 6 AP often needs 802.3at or 802.3bt.
- IP telephones — one cable to the desk carries both voice data and power; the phone may even pass a switched port through to a PC.
- IP cameras — mounted on ceilings and eaves; pan-tilt-zoom models need PoE+.
- Access control and security — badge readers, door controllers, intercoms.
- IoT and building systems — environmental sensors, digital signage, and PoE lighting.
The design payoff is practical: one cable per device, one place to manage power (the switch), and the ability to remotely power-cycle a locked-up access point by bouncing its port. The design cost is that the switch’s power budget, not just its port count, becomes a planning number.
Review
Check what you have learned with these review questions.
Sources & references
- IEEE. IEEE 802.3: Ethernet, including Clause 33 (802.3af), 802.3at, and 802.3bt Power over Ethernet amendments. standards.ieee.org.
- Cisco. Power over Ethernet (PoE) Configuration Guide (PSE/PD roles, classes, budgets, and negotiation). cisco.com.
- Cisco. CCNA (200-301) Exam Topics, Network Fundamentals — physical interfaces and cabling. cisco.com.
- Cisco. CCNA: Introduction to Networks — “Network Media” and “Physical Layer” (connection types and PoE). netacad.com.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.