Transmission Modes: Half-Duplex and Full-Duplex
Full-duplex Ethernet allows independent transmission in both directions. Telecommunications distinguishes three transmission modes by the direction of data flow. The mode describes the transmitter and receiver relationship on a link, not an application-layer protocol.
Simplex
Data flows in only one direction. Examples include a radio broadcast or a one-way link. One device remains the transmitter and the other the receiver.
Half-Duplex
Data can flow in both directions, but not at the same time. One transmitter must finish before the other direction can use the medium. A walkie-talkie is a typical half-duplex example.
Full-Duplex
Data can flow in both directions simultaneously. The two interfaces can transmit and receive at the same time. A telephone call is a typical full-duplex example.
Early Ethernet using shared media, such as coaxial buses and hubs, operated in half-duplex mode. Stations had to coordinate access to one channel, which limited link utilization. Switched point-to-point links made full-duplex operation possible, with independent physical-layer resources for the two directions.
Half-Duplex and Collision Domains
In a half-duplex Ethernet environment, all devices share a . One channel carries sending and receiving, so devices use CSMA/CD to coordinate access. The main properties of this mode are:
- Sense Before Transmitting: A device waits until the medium is idle before it begins sending data.
- Monitor During Transmission: The device monitors the medium while sending and detects a collision if another station starts transmitting at the same time.
- Shared Bandwidth:The link's total bandwidth, for example 100 Mbps, is shared between sending and receiving. At a given moment the medium is transmitting, receiving, or idle. Useful throughput is lower than the nominal rate because of waiting and collision handling. The available capacity is therefore divided between stations contending for the same segment.
Technologies Enabling Full-Duplex
The move from half-duplex to full-duplex required Ethernet switches and cabling that supports independent transmission directions.
- Ethernet Switches: A hub repeats a signal to all ports, while a switch forwards a frame to the port selected from its MAC table. The connection between a PC and the switch is point-to-point and is not a shared medium for other hosts. Each port has its own collision domain, and both endpoints must agree on full-duplex operation.
- Multi-Pair Cabling: 100BASE-TX and 1000BASE-T use twisted-pair cable with multiple pairs, usually four. In 100BASE-TX, one pair carries transmit data (the TX pair) and another carries receive data (the RX pair). In 1000BASE-T, all four pairs carry signals in both directions, while the PHY separates the local transmitted signal from the received signal.
A switch together with suitable PHYs and cabling creates a point-to-point link. In 100BASE-TX the directions use separate pairs; in 1000BASE-T signal processing provides the separation. CSMA/CD collision handling is disabled on such a full-duplex link.
Effects of Full-Duplex Operation
Enabling full-duplex mode on a switched, point-to-point Ethernet link removes contention for a shared medium and changes how transmission is handled. The mode affects access behavior, while the physical link rate remains the configured port rate.
1. No collisions
Collisions as defined by CSMA/CD do not occur on a full-duplex point-to-point link, so Collision Detection is not needed. A full-duplex device disables its CSMA/CD logic, does not perform random backoff, and can transmit once the required interframe spacing is met. Autonegotiation can select speed and duplex, while a mismatch produces errors and frame loss.
No waiting for the shared medium
The device does not wait for a shared medium, send a Jam signal, or back off after a collision. Latency still depends on processing, queueing, and link properties.
2. Throughput in both directions
A device can send and receive at the same time, so aggregate throughput across both directions is the sum of the transmit and receive rates. This does not increase the rate of one direction: a 100 Mbps port still sends at up to 100 Mbps, but can reach that rate in both directions simultaneously.
Half-Duplex 100 Mbps Link
Total shared capacity: 100 Mbps.
(At a given time, sending or receiving at up to 100 Mbps)
Full-Duplex 100 Mbps Link
Total aggregate throughput: 200 Mbps.
(100 Mbps sending AND 100 Mbps receiving simultaneously)
For applications with symmetric traffic, such as a file server receiving requests and sending data, simultaneous directions increase possible aggregate throughput. With asymmetric traffic, the benefit depends on whether both directions are loaded.
Full-Duplex in Gigabit and 10-Gigabit Ethernet
Full-duplex operation also applies to faster standards. In 1000BASE-T (Gigabit Ethernet), all four pairs carry transmit and receive signals at the same time. DSPs use hybrid circuits and echo cancellation to distinguish the local signal from the received signal on each pair. The result is 1000 Mbps in each direction, or 2000 Mbps aggregate throughput. 10GBASE-T also operates as a full-duplex link using signaling specified by IEEE 802.3. Full-duplex does not remove receiver congestion; PAUSE frames can provide flow control, but they are not a collision-detection mechanism.