Review of LAN Topologies for industrial Ethernet, Part I

While the very foundation of legacy Ethernet was a non-deterministic protocol in a collision-based networking environment, a majority of factory and control automation applications adapted Ethernet to a more deterministic-like response time. This led to a plethora of proprietary implementations within the industrial automation and control industry. In 2002 the IEEE 1588 Precision Time Protocol Standard addressed the need for deterministic response by introducing a precision clock synchronization protocol for networked measurement and control systems. The adoption of this standard is implemented in various Real Time Industrial Ethernet networking protocols.

LAN topologies: From bus to star topologies
In the past two decades, Ethernet has evolved from a half-duplex protocol, supporting CSMA/CD(Carrier Sense, Multiple Access with Collision Detection) to a full-duplex Ethernet. As a Local Area Network (LAN) topology, it migrated from a shared medium bus topology to a star/hub topology. In the early 80's, the original commercial medium was based on a yellow coaxial cable, referred to as the thick cable with a characteristic impedance of 50 ohms. This was quickly replaced by the lower cost thin RG58 coaxial cable, commercially known as the Thinnet, Thin Ethernet or Cheapernet.

By the mid-80's, the lower cost cabling medium helped pushed Ethernet as the popular medium of choice for interconnecting PC's. Using BNC (Bayonet Neill Concelman, named after the designer) T-type coax connectors per each node connections, the 10Base-2 (IEEE802.3a Std.) network had a network diameter distance of 185 meters. The thin coaxial cable bus supported a multi-drop topology with its opposite ends terminated with a 50 ohm termination plugs. The physical cable minimum spacing between each multi-drop node must be greater than 50 centimeters (See Figure 1)

From a physical layer point of view, a collision circuitry is used to monitor the coaxial DC level. If the level is more negative than the collision threshold, the collision output is enabled by the coax PHY transceiver. The coax-based bus topology supported a half-duplex, shared medium at 10 Mbps which had enough performance for interconnecting NIC cards on AT or ISA busses on PC's whose bus bandwidth were adequate for the shared medium.

By the early 90's, the fundamental shared LAN topologies changed to what it is today. Two main inflection points of change for LAN topologies include:

• Implementation of low cost structured wiring based on unshielded twisted pair (UTP) cabling (e.g. CAT5 cabling), leading to centralized wiring hubs or star topology.
• Rapid development of Ethernet LAN switches overcame the barrier of limited system performance for repeater-based topologies.

Prior to LAN switches, hubs or repeaters were the interconnecting "concentrators" for supporting structured unshielded twisted pair, wiring connections. A typical 24 port repeater would support 24 ports in a single collision domain. The bundled structure telephone wiring located in a wiring closet usually comes in a bundle of 50 wires. Thus a typical 24 port LAN repeater at 10 Mbits/sec. would coexist in a structured wiring infrastructure with PABX's in an office environment, supporting both data and voice communications.

In 1995, 100Base-T or Fast Ethernet provided a smooth transition to existing 10 Mbps by introducing an innovative concept of auto-negotiation where the 10/100 Physical layer chips level would interchange status packets in the form of an embedded 16 bit word within the Fast Link Pulses (FLPs), determining the data rate and automatically adapting to the highest speed between the 2 Cat5 links. At the physical layer chip level, the same collision type circuitry would detect the simultaneous transmitting and receiving activity for both the TX and RX differential pairs. Thus it was easy for to have hybrid networks to support both existing 10 Mbits/ sec and 100 Mbits/sec. connections.

While 100 Mbps Ethernet is based upon the same principles of CSMA/CD as 10 Mbits/sec., the shared medium architecture did not provide an expected tenfold in performance. This is because the maximum bandwidth available to each user decreases proportionally to the number of network nodes and network delays are always variable due to CSMA/CD network access control method.

Single collision domain and repeater architectures
Previously, the meaning of collision domain was defined as a shared medium architecture, supporting a collection of nodes connected by any number of repeaters. With 100 Mbits/sec networks, the collision domain demands that the round trip delay must not exceed 512 bit times (one slot time) which amounts to 5.12 microseconds. This implies that it takes more time for a packet to traverse the network and collide with another packet, resulting in a collision, transmitting the status to the originating source. It also that ensures that if a collision arises in the network, it would be recognized at all node locations.

If this time is exceeded the 512 bit (one slot time) the originating node will not detect a collision prior to transmitting its 512 bit. This results in late collisions faults within the network. For 100 Base-T networks the network diameter distance is limited to 210 meters. With 100-Mbit repeaters, the timing budget constraints are calculated to ensure no late collisions. An example calculation is as follows:

With a limited bit budget, no more than 2 repeaters may co-exist in a collision domain. Between 2 repeaters, the inter-repeater link is generally limited up to 5 meters. The only exception is stackable repeaters that are interconnected via the internal backplanes of the inter-repeater bus. These repeaters are modular and expandable but all the interconnected ports are under one collision domain. Class 1 and Class 2 repeaters were defined in the IEEE802.3u standard, thus affecting topology differences. A Class 1 repeater will have a more relaxed timing budget, supporting only one repeater within its collision domain and permit further budget delays for stackable repeaters (See Figure 2).