Hub Construction
Part of Networking
How Ethernet hubs work and how to build or repair basic hub-equivalent devices.
Why This Matters
An Ethernet hub is the simplest multi-port network device: it receives a signal on any port and immediately retransmits it on all other ports. This simple behavior creates a shared medium where all connected devices can communicate with each other without needing a switch’s intelligence. While modern networks use switches instead of hubs, understanding hub construction provides foundational knowledge for understanding more complex devices and for working with legacy equipment.
In a reconstruction context, hubs are valuable because they are simpler to repair and understand than switches, and they enable basic networking when switching equipment is unavailable. A hub can also be built from discrete components — unlike a modern switch with its application-specific integrated circuits, a basic Ethernet hub’s core function can be understood and potentially replicated with sufficient electronics knowledge.
Even if you only ever use commercial hubs rather than building your own, understanding their internal workings helps you diagnose problems, understand failure modes, and make better decisions about when a hub is adequate versus when a switch is needed.
How Hubs Repeat Signals
The core function of a hub is signal repetition. When a device connects via twisted pair (10BaseT or 100BaseTX), it sends a differential signal on one pair and receives on another. The hub must:
- Receive the incoming differential signal and reconstruct the digital bit stream
- Re-time the bits using the hub’s clock (to prevent jitter accumulation)
- Retransmit the reconstructed signal on all other ports simultaneously
The retransmission must maintain the correct timing for CSMA/CD to work. If the hub simply buffered the incoming data and retransmitted after a delay, the collision detection timing would be violated. Hubs retransmit with minimal delay — one or two bit times at most — so that the overall propagation delay in the collision domain stays within Ethernet specifications.
A hub also performs collision detection itself. If two ports transmit simultaneously, the hub detects the collision (signals received on multiple ports at once), transmits a jam signal on all ports, and waits for the backoff period. This happens transparently to all connected devices.
Signal Handling and Regeneration
On the incoming side, the hub’s repeater circuitry must handle several signal quality issues. The incoming signal arrives attenuated (weaker than transmitted), distorted (edges rounded by cable capacitance), and with timing jitter (slight variations in bit timing). The hub must amplify, reshape, and retime the signal before rebroadcasting.
Equalization circuits compensate for the frequency-dependent attenuation of twisted pair cable. Higher frequencies are attenuated more than lower frequencies, causing sharp digital edges to be rounded. The equalizer applies the inverse of this attenuation curve — boosting high frequencies — to restore a sharper signal.
The data recovery circuit extracts timing information from the incoming signal and uses it to drive a phase-locked loop (PLL) that generates a clean local clock. The bits are then re-clocked using this local clock, removing timing jitter accumulated over the cable run.
The retransmit path takes the re-clocked bits and drives the output line drivers on all other ports with appropriate voltage levels and slew rates for the specific Ethernet standard.
10BaseT vs. 100BaseTX Hubs
10BaseT hubs operate at 10 Mbps using Manchester encoding, where each bit is encoded as a transition in the middle of the bit period. Manchester encoding makes clock recovery straightforward because every bit contains at least one transition. The signaling voltage is ±2.5V differential.
100BaseTX hubs operate at 100 Mbps using MLT-3 encoding (three-level encoding that reduces the highest frequency component) and 4B5B data encoding (groups of 4 bits encoded as 5-bit symbols). The faster signaling requires better signal integrity, which is why 100BaseT requires Cat5 cable rather than the Cat3 adequate for 10BaseT.
Most commercial hubs are either 10 Mbps or 10/100 Mbps autosensing. True autosensing hubs detect the speed of each incoming device and connect it to the appropriate internal repeater segment. A 10/100 hub does not bridge between the two speed segments — devices at different speeds cannot communicate through a plain hub.
Building a Simple Hub Equivalent
A minimal working hub can be constructed using integrated circuits designed for this purpose. In the era of 10Base2 coaxial Ethernet, multiport repeater chips (such as the Intel 82501 series) were designed specifically for this function. For twisted-pair 10BaseT, chips like the SEEQ 8004 or equivalent provided multi-port repeater functions.
A practical minimal 4-port 10BaseT hub requires:
- A multi-port Ethernet repeater IC
- Four transformer-isolated line interface circuits (one per port), each containing a pulse transformer for isolation, a line receiver, and a line driver
- Clock generation circuit (10 MHz crystal oscillator)
- Power supply (typically 5V, 100-500 mA depending on port count)
- RJ-45 jacks with integrated magnetics (transformers built into the connector housing)
The RJ-45 connectors with integrated magnetics simplify construction considerably — instead of separate transformers for each port, the isolation and impedance transformation are built into the connector housing. These are available as off-the-shelf components.
Configuring the repeater IC involves setting its operating mode (generally through logic-level pins for basic functions), connecting the transmit and receive pairs from the line interfaces, and providing the clock signal. The IC handles all the Manchester decoding, bit retiming, and retransmission internally.
Commercial Hub Repair
More practical than building from scratch is repairing failed commercial hubs. Common failure modes include:
Power supply failure: The most common failure in older hub equipment. Hub power supplies use electrolytic capacitors that fail after 10-20 years, causing either complete failure or erratic behavior. Identifying and replacing the failed capacitors restores the hub. Visual inspection often reveals failed capacitors: domed tops (they should be flat), or electrolyte leakage around the base.
Port failure: Individual ports fail more often than the hub’s core circuitry. Typically caused by electrostatic discharge or voltage spikes on the connected cable. The transformer isolation protects the core logic, but the line driver/receiver circuits on individual ports can fail. A port tester or cable tester confirms which port has failed.
Cooling failure: Hubs with fans fail when the fan stops. The repeater ICs generate heat and will overheat without airflow, causing intermittent operation or permanent damage. Replacing the fan (a standard DC brushless fan) restores normal operation.
Testing a repaired hub requires at minimum two devices that can communicate through it and a cable tester to verify all ports. A simple ping test between two connected computers confirms basic hub operation. Network monitoring software can verify that traffic from each port appears on all other ports (hub behavior) and measure collision rates and error rates.