Bob Metcalfe’s original design for Ethernet consisted of DTE devices connected to a common coaxial cable through the use of “tee” connectors, like this:
This cabling arrangement suffered several problems. First, it was inconvenient to run through an office building, since each DTE device needed to be coupled rather closely to the main “trunk.” Short cable segments (called stubs, spurs, or drops) joining the main trunk line to each DTE device could not be too long, or else they would could cause multiple signal reflections to occur in the main line. Secondly, the signal strength decreased with each “tee” connector: every time the signal branched, it would lose power. Thirdly, the need for termination resistors at the far ends of the “ether” cable opened up the possibility that those terminators might fail, fall off, or be forgotten during installation or maintenance .
As Ethernet evolved as a practical networking standard, one of the many improvements added to its design was the concept of a repeating hub. A “repeater” is an active device designed to rebroadcast a signal, usually to overcome inevitable power losses incurred as that signal propagates along a cable. Repeaters are common in the telecommunications industry, where telephone, television, and computer signals must travel hundreds or thousands of miles between points of transmission and reception. A “repeating hub” is a repeater with multiple ports for many cables to plug into, where any signal entering on any cable is repeated to all ports on the device. Thus, a repeating hub (or simply “hub”) allows multiple Ethernet devices to interconnect with no degradation in signal quality:
Not only do hubs improve system performance by boosting signals’ voltage levels, but they also eliminate the need for termination resistors in the network. With a hub-based system, each and every cable terminates at either a DTE or DCE device, which is (now) designed with the proper termination resistance built-in to their internal transceiver circuitry. This means each and every Ethernet cable is automatically terminated by the proper impedance simply by plugging it in to the Ethernet port of any device. “Stub” or “spur” cables with their length restrictions are also a thing of the past, since no cable ever splits or branches in a hub-based network system.
Hubs are considered “layer 1” devices, because they operate purely on the physical layer of Ethernet: all they do is receive Ethernet signals and re-broadcast those signals in boosted form to all other devices plugged into the hub. As a piece of interconnecting hardware, a hub is considered a DCE (Data Communications Equipment), as opposed to the end-of-cable devices such as computers and printers which are DTEs (Data Terminal Equipment).
Repeating hubs may be connected together to form larger networks.
Since hubs are merely “layer 1” devices, mindlessly boosting and re-broadcasting signals received to their ports, their presence does not mitigate collisions between transmitting devices. As far as collisions between those devices is concerned, they might as well be directly connected together on a single piece of coaxial cable. One way to express this concept is to say that all portions of the network are part of the same collision domain. In other words, a collision happening in one portion of the network happens in all portions of the network.