Full-duplex Ethernet
While the early versions operated with coaxial cable were designed for half- duplex , all of today's common versions of twisted pair ( 10BASE-T , 100BASE-TX and 1000BASE-T ) and fiber optic Ethernet have separate channels or lines for transmission and receiving. The exception is the Automotive Ethernet (100BASE-T1, 1000BASE-T1).
In order to enable the use of hubs and for compatibility with other Ethernet variants, they were originally implemented as half-duplex, with a transceiver for collision detection and with loopback switching of the data so that the host itself also receives this, as if the transmission had taken place on a shared medium.
In a switched network, the participants do not communicate directly, but via a switch . If the hardware supports the full duplex mode, the collision detection can be dispensed with, since only one participant addresses the send and receive lines. So if there are no hubs at either end of the connection and hardware support is available, the two channels can be split up to create a true full duplex connection.
However, if the autonegotiation mode is activated at one end and full duplex operation is enforced at the other end, the participant with autonegotiation recognizes the link as half duplex, which causes a large number of transmission errors.
Individual evidence
- ↑ Mehmet Tazebay: Deployed 100Mb / s One Pair OABR PHY . Ed .: IEEE Standards Association. Norfolk December 2014.