Collision Detection in Ethernet

Every Ethernet cable has a maximum length. Exceed it, and your network fails—sometimes obviously with no link, sometimes subtly with intermittent errors and late collisions. But these limits aren't arbitrary. They emerge from the intersection of physics and protocol design.

Why is 10BASE-T limited to 100 meters? Why could 10BASE5 span 500 meters per segment? And how does fiber reach kilometers while copper struggles past 100 meters? The answers lie in understanding the factors that constrain segment length.

Three distinct factors can limit how far Ethernet signals can travel. Understanding which factor dominates in each scenario helps explain the variety of distance limits across different standards.

Which Factor Dominates?

The dominant factor varies by standard and media type:

• 10BASE5 (Thick Coax): Timing dominates. Could go 500m per segment due to low attenuation but needed repeaters for CSMA/CD timing.

• 10BASE-T (UTP): Attenuation dominates. 100 meters is where Cat3/5 cable attenuates 10 MHz signals below reliable thresholds.

• 100BASE-TX (UTP): Attenuation AND timing. 100 meters for attenuation; shorter hub-to-hub for timing.

• 1000BASE-SX (Multimode Fiber): Modal dispersion dominates. Limits range to 220-550m depending on fiber grade.

• 1000BASE-LX (Single-mode Fiber): Attenuation dominates. Can reach 5km+ with low-loss single-mode fiber.

• 10GBASE-LR (Single-mode): Dispersion becomes significant again. 10km limit involves both attenuation and chromatic dispersion.

Classic coaxial Ethernet (10BASE5 and 10BASE2) had segment limits primarily driven by timing requirements, with attenuation as a secondary concern.

10BASE5 (Thick Coaxial) — 500 Meters:

The 500-meter segment limit derives from several factors:

1. Propagation delay budget: At 0.77c, 500m contributes ~2.16 μs one-way delay. Five segments (2500m) contribute ~10.8 μs, leaving room for repeater delays within the 25.6 μs one-way slot time budget.

2. Signal attenuation: Thick coax has approximately 8.5 dB loss per 500m at 10 MHz. This is within acceptable receiver sensitivity.

3. Station loading: Each transceiver presents impedance to the cable. More than 100 transceivers degrades signal quality unacceptably.

10BASE2 (Thin Coaxial) — 185 Meters:

Thin coax has higher attenuation (~20 dB/500m) and lower velocity factor:

1. Higher attenuation limits segment to 185m before signal becomes unreliable
2. Lower velocity factor (0.65c) means more delay per meter
3. Simpler connectors (BNC) are cheaper but less reliable over long runs
4. Fewer stations (30) due to higher per-tap loading

Twisted pair Ethernet dominates modern installations. The universal 100-meter limit seems arbitrary but emerges from careful engineering considering attenuation, crosstalk, and return loss at the signaling frequencies used.

The 100-Meter Convention:

The 100-meter limit consists of three components:

• Horizontal cable: Maximum 90 meters (fixed infrastructure)
• Patch cords: 5 meters at each end (equipment room + work area)
• Total channel: 100 meters (90 + 5 + 5)

This "permanent link + patch cord" model reflects real-world cabling practice where building infrastructure is permanent but equipment connections use patch cables.

Why 100 Meters?

For 10BASE-T operating at 10 MHz:

• Signal attenuation in Cat3: ~13 dB/100m at 10 MHz
• Receiver sensitivity threshold: typically -40 dBm
• Transmitter output: ~0 dBm
• Budget: 40 dB minus connector losses and margin ≈ 13 dB available ≈ 100m

For faster standards (100BASE-TX, 1000BASE-T), higher frequencies mean higher attenuation per meter. However, improvements in cable quality (Cat5, 5e, 6, 6a) compensate, maintaining the 100m limit.

10GBASE-T Exception:

10 Gigabit Ethernet over copper uses signaling up to 500 MHz. At these frequencies:

• Cat6 cable reaches its limits at ~55 meters
• Cat6a (augmented Cat6) with alien crosstalk shielding reaches 100m
• Careful installation practices are required

Optical fiber offers far greater distances than copper, with limits determined by optical power budget (attenuation) and pulse dispersion rather than electrical constraints.

Multimode vs. Single-mode Fiber:

Multimode fiber (typically 50/125 μm or 62.5/125 μm core) carries multiple light paths (modes) simultaneously:

• Pros: Cheaper transceivers (LED sources), easier alignment
• Cons: Modal dispersion limits distance; different modes travel at different speeds, causing pulse spreading
• Typical use: Data center interconnects, building backbones (<500m)

Single-mode fiber (9/125 μm core) carries only one light mode:

• Pros: No modal dispersion; much longer distances possible
• Cons: Laser sources required, more precise alignment needed, more expensive
• Typical use: Campus backbones, metro networks, telecommunications

Fiber Grade (OM) Ratings:

Multimode fiber is classified by its bandwidth capacity:

• OM1 (FDDI-grade): 200 MHz·km at 850nm, 500 MHz·km at 1300nm
• OM2 (50μm OM2): 500 MHz·km at 850nm, 500 MHz·km at 1300nm
• OM3 (laser-optimized): 2000 MHz·km at 850nm
• OM4 (extended laser-optimized): 4700 MHz·km at 850nm
• OM5 (wideband): supports SWDM (shortwave wavelength division multiplexing)

Higher OM grades support longer distances at faster speeds due to reduced modal dispersion.

For half-duplex networks, the maximum network diameter is constrained by slot time. Let's derive these limits mathematically.

The Fundamental Timing Equation:

$$D_{max} = \frac{(t_{slot} - t_{overhead}) \times v}{2}$$

Where:

• D_max = maximum network diameter (meters)
• t_slot = slot time (seconds)
• t_overhead = repeater + interface delays (seconds)
• v = signal velocity (m/s)
• Factor of 2 accounts for round-trip

10 Mbps Ethernet Example:

• Slot time: 51.2 μs
• Overhead budget: ~20 μs (for 4 repeaters + interfaces)
• Available for propagation: 51.2 - 20 = 31.2 μs (for round-trip)
• One-way propagation: 15.6 μs
• Signal velocity: 0.77c = 2.31 × 10⁸ m/s

$$D_{max} = 15.6 \times 10^{-6} \times 2.31 \times 10^8 = 3600\text{ m (theoretical)}$$

The IEEE standard specifies 2500m to provide safety margin and account for component variation.

100 Mbps Fast Ethernet Example:

• Slot time: 5.12 μs
• Overhead budget: ~3 μs (fewer repeaters allowed)
• Available for propagation: 2.12 μs (for round-trip)
• One-way propagation: 1.06 μs
• Signal velocity: 0.65c = 1.95 × 10⁸ m/s (Cat5 UTP)

$$D_{max} = 1.06 \times 10^{-6} \times 1.95 \times 10^8 = 207\text{ m}$$

This theoretical maximum of ~200m explains why Fast Ethernet specifies ~100m station-to-hub distances with limited cascading.

In full-duplex networks (or where timing isn't the limit), signal attenuation determines maximum distance. Let's examine how to calculate these limits.

The Link Budget Equation:

$$\text{Received Power (dBm)} = \text{Transmitted Power (dBm)} - \text{Total Loss (dB)}$$

For successful communication:

$$\text{Received Power} \geq \text{Receiver Sensitivity}$$

Therefore:

$$\text{Maximum Loss} = \text{Transmitted Power} - \text{Receiver Sensitivity}$$

$$D_{max} = \frac{\text{Maximum Loss} - \text{Connector/Splice Loss}}{\text{Attenuation per km}}$$

Twisted Pair Example (1000BASE-T):

• Transmitter output: ~1.0 V peak-to-peak
• Receiver sensitivity: requires ~100 mV for reliable detection
• Maximum attenuation: about 20 dB (10:1 voltage ratio)
• Cat5e attenuation at 100 MHz: ~22 dB/100m
• Connector losses: ~2 dB (4 connectors × 0.5 dB each)

$$D_{max} = \frac{20 - 2}{22} \times 100 = 82\text{ m (theoretical)}$$

The 100m standard includes margin for crosstalk, return loss, and cable variation.

Single-Mode Fiber Example (10GBASE-LR):

• Transmitter power: typically -3 to -1 dBm
• Receiver sensitivity: typically -14.4 dBm or better
• Maximum loss budget: approximately 11 dB
• Fiber attenuation: 0.35 dB/km at 1310 nm
• Connector losses: 1.5 dB (2 connectors × 0.75 dB)
• Splice losses: 0.1 dB/km average (for long runs)

$$D_{max} = \frac{11 - 1.5}{0.35 + 0.1} = \frac{9.5}{0.45} = 21\text{ km (theoretical)}$$

The 10GBASE-LR specification of 10 km includes significant margin and accounts for real-world installation quality.

Multimode Fiber Example (10GBASE-SR on OM4):

• Transmitter power: typically 0 dBm
• Receiver sensitivity: typically -10.3 dBm
• Maximum loss budget: approximately 10 dB
• OM4 attenuation at 850nm: 3.0 dB/km
• Modal dispersion at 10 Gbps: limits bandwidth to 2.5 km before dispersion is problematic

$$D_{attenuation} = \frac{10 - 1.5}{3.0} = 2.83\text{ km}$$ $$D_{dispersion} \approx 0.5\text{ km (modal dispersion limited)}$$

Dispersion is the limiting factor, hence the ~400m specification for OM4.

Sometimes network requirements exceed standard distance limits. Several technologies allow Ethernet to reach farther than basic specifications.

Media Converter Approach:

Most common for campus networks:

1. Station connects to media converter via 100m UTP
2. Media converter converts to multimode or single-mode fiber
3. Fiber runs to distant building (up to 2 km multimode, 10+ km single-mode)
4. Second media converter converts back to UTP
5. Remote device connects via 100m UTP

Total distance: 100m + fiber distance + 100m

PoE Considerations:

Power over Ethernet (PoE) has stricter distance limits due to voltage drop:

• IEEE 802.3at/bt specifies 100m maximum for PoE
• At 90m, voltage drop may exceed tolerance for high-power devices
• PoE extenders/injectors can relay power at intermediate points
• Fiber links cannot carry PoE; power must be provided at both ends

Maximum segment length is a fundamental network design parameter determined by timing, attenuation, and dispersion constraints. Let's consolidate our knowledge with a comprehensive reference: