Repeaters and Hubs

At the dawning of the ETHERNET, the repeaters had only two network connections and were used to connect two segments to each other in order to extend the segment length. Later, repeaters with more than two network connections were available. Those star distributors are called hubs. They are able to connect several segments. Apart from the number of network connections the functionality of hubs and repeaters is identical. This is why we only use the term “hub” in the following descriptions.

Hubs are operating on the lowest layer of the ISO/OSI model and are therefore independent of the protocols used on ETHERNET. The network connections of hubs are exclusively operated in half duplex mode. Due to this, collision domains can freely propagate beyond the hubs. A hub can only support one transmission rate for all connections. Therefore it is not possible to connect segments with different transmission rates via a simple hub. For this purpose a dual-speed hub has to be used. The fundamental functions of hubs are as follows:

  • Restoration of the signal magnitude
  • Regeneration of the signal timing
  • Propagation of a detected collision
  • Expansion of short fragments
  • Creation of a new preamble
  • Isolation of a faulty segment

When transmitted over the medium (e.g. a twisted pair cable) the data signal is attenuated. The task of a hub is to amplify an incoming signal in order to make the full signal magnitude available at the outputs again. Furthermore, a distortion of the binary signal’s on-off ratio (jitter) can occur during data transmission. When transmitted via a hub, the hub is able to restore the correct on-off ratio of the signal which avoids propagation of the signal jitter beyond the segment.

However, one of the most important tasks of a hub is to propagate occurring collisions within the entire collision domain so that the collision can be detected by all connected stations. If it detects a collision on one of his connections, the hub sends a so-called jam signal over all connections. If a hub receives a data fragment which, by its principle, could only be created by a collision, it first brings the fragment to a length of 96 bits and then forwards it via the ports. This shall guarantee that the data fragment can be received by all stations independent of their distance to the hub and removed from the network. The detected data fragments are removed by the terminal devices by not forwarding them to the higher layers.

By means of the data package preamble the beginning of a data package is detected so that the recipient can synchronize to the incoming data stream. However, during the data transmission it can occur that the first bits of a preamble are lost. The task of the hub is to restore a possibly incomplete preamble before forwarding it.

If collisions occur within one segment in large numbers in a short period of time or if e.g. a short circuit on a data line causes failures, the hub switches off the faulty segment to avoid interference to the entire collision domain.