How Ethernet Evolved from Shared Coax to Modern Gigabit Networks

Early Ethernet Architecture

Ethernet originated at Xerox PARC as a shared‑medium, half‑duplex LAN. All stations were attached to a single coaxial cable and contended for the medium using the CSMA/CD protocol.

CSMA/CD Mechanics

Carrier Sense (CS) : a NIC monitors the line voltage; transmission starts only if the line is idle.

Multiple Access (MA) : any station may attempt to transmit once the line is sensed idle.

Collision Detection (CD) : while transmitting, the NIC continuously samples the line. If the sensed voltage deviates from the transmitted pattern, a collision is declared.

On collision detection the NIC:

Sends a jam signal (32 µs) to ensure all stations see a corrupted frame.

Stops the current transmission.

Waits a random back‑off interval chosen from the range 0 … (2^k – 1) * slot_time, where k is the collision count (capped at 10) and slot_time is 512 bit‑times (≈51.2 µs at 10 Mbps).

Retries transmission.

The protocol imposes a minimum frame size of 64 bytes (512 bits) so that a transmitting station can still detect a collision before the frame ends. It also defines a maximum cable length for a given data rate to keep the round‑trip propagation delay below one slot time.

First Commercial Ethernet – 10BASE5 (Thicknet)

10BASE5 was the first widely deployed Ethernet standard. Its key parameters were:

Speed : 10 Mbps

Maximum segment length : 500 m

Cable type : 50 Ω RG‑8/U coax, bright yellow for easy identification

Connector : N‑type BNC at each end, terminated with 50 Ω resistors

Attachment method : vampire tap – a metal clamp that pierces the cable insulation without cutting the conductor.

Vampire Tap Installation

To add a station without interrupting the network:

Mark the precise tap location (black‑painted reference points are spaced according to the Ethernet timing budget).

Use a specialized drilling tool to make a clean hole through the cable jacket.

Clamp the transceiver onto the exposed conductor; the tap provides both signal coupling and a 50 Ω termination for the segment.

Advantages: no network outage and flexible placement. Drawbacks: requires skilled technicians; a mis‑drilled tap can cause intermittent faults or total segment failure.

Alternative – Cut‑and‑Splice

The more disruptive method involves cutting the cable, installing an N‑type connector, mounting the transceiver, and then restoring the link. This forces a full network outage and is rarely used in practice.

Attachment Unit Interface (AUI)

Early Ethernet NICs did not contain a transceiver. Instead they connected to an external transceiver via a 15‑pin AUI cable (approximately 50 cm long). The AUI cable is passive; power for the transceiver is supplied from the host. This architecture added bulk and complexity compared with modern integrated PHYs.

Transition to Switched, Full‑Duplex Ethernet

Shared‑medium Ethernet suffered from:

Collision domains that limited aggregate throughput.

Difficulty scaling beyond a single 500 m segment.

High installation and maintenance cost due to thick coax and precise tap locations.

The introduction of Ethernet switches in the late 1980s provided a dedicated point‑to‑point link for each port, eliminating collisions. Full‑duplex operation (simultaneous transmit and receive) became standard, allowing the CSMA/CD algorithm to be retired.

Speed Evolution Timeline

1973 – 3 Mbps (≈2.94 Mbps) over RG‑11 coax.

1980 – 10 Mbps (10BASE5 thicknet).

1990 – 100 Mbps Fast Ethernet (100BASE‑TX over Category 5 twisted pair).

1998 – 1 Gbps (1000BASE‑SX/LX over fiber, 1000BASE‑T over Cat5e).

2001 – 10 Gbps (10GBASE‑SR/LR over fiber, 10GBASE‑T over Cat6a).

2010 – 40 Gbps (40GBASE‑SR4).

2015 – 100 Gbps (100GBASE‑SR10, 100GBASE‑LR4).

Key Takeaways

Early Ethernet introduced several concepts that remain fundamental:

Distributed media access without a central controller.

Deterministic timing constraints (minimum frame size, maximum segment length) derived from the physics of signal propagation.

The elegant CSMA/CD state machine, still taught as a classic example of contention‑based networking.

Understanding the constraints of the original coaxial, shared‑medium design explains why modern Ethernet migrated to switched, full‑duplex, twisted‑pair and fiber media, achieving orders of magnitude higher bandwidth while simplifying installation and troubleshooting.