If collision domains are the territories of Ethernet, then slot time is its heartbeat—the fundamental timing interval that synchronizes the entire CSMA/CD protocol. Every critical parameter in classic Ethernet—minimum frame size, backoff intervals, collision detection windows—derives from slot time.

Slot time is precisely 512 bit times for 10 Mbps and 100 Mbps Ethernet. This seemingly arbitrary number encapsulates the physics of signal propagation, the engineering of network hardware, and the mathematics of collision detection into a single, elegant parameter.

Slot time is the worst-case time required for a transmitted signal to reach all corners of the collision domain AND for a collision indication to return to the transmitter. It includes all propagation delays, repeater delays, and hardware processing times, plus safety margins.

More formally:

$$\text{Slot Time} = 2 \times t_{propagation} + t_{hardware} + t_{safety}$$

Where:

• t_propagation = one-way propagation delay across maximum network diameter
• t_hardware = total repeater/transceiver processing delays
• t_safety = engineering margin for component variation

Let's derive the slot time for 10BASE5 Ethernet step by step, examining each component that contributes to the total.

Component 1: Cable Propagation Delay

For 10BASE5 with maximum configuration (5 segments × 500m):

• Total cable length: 2500 meters
• Signal velocity: ~0.77c (for thick coax) = 2.31 × 10⁸ m/s
• One-way propagation: 2500 / (2.31 × 10⁸) = 10.82 μs
• Round-trip cable delay: 21.64 μs

Component 2: Repeater Delays

Maximum of 4 repeaters in the path:

• Per-repeater delay: ~7.5 μs (varies by implementation)
• Total repeater delay: 4 × 7.5 = 30 μs (round-trip, since signal passes through each twice)

Note: In practice, signal only passes through each repeater once per direction, but we account for round-trip.

Component 3: Transceiver and DTE Delays

End devices (DTEs) and transceivers add processing time:

• Transceiver delay: ~1-2 μs per end
• DTE interface delay: ~1-2 μs per end
• Total end delays: ~6-8 μs round-trip

Component 4: Safety Margin

Engineers add margin for:

• Component tolerance (−5% to +5%)
• Cable length measurement errors
• Manufacturing variation
• Temperature effects on propagation speed
• Margin: ~5-10 μs

When a collision occurs, stations cannot immediately retry—they would just collide again. The Binary Exponential Backoff algorithm uses slot time as its fundamental unit to randomize retransmission timing and resolve contention.

The Backoff Algorithm:

1. After the n-th collision (n = 1, 2, 3, ..., 16):
2. Choose k = min(n, 10) — cap k at 10
3. Select random integer r from uniform distribution [0, 2^k - 1]
4. Wait for r × slot_time before retrying
5. If n > 16, abort transmission and report failure

Example Backoff Ranges:

Why Binary Exponential?

The doubling of the backoff window serves critical purposes:

1. Low Delay at Low Contention: After the first collision, stations choose from only 0 or 1 slot times. If only two stations collided, there's a 50% chance they won't collide again. Average wait is 0.5 slots.

2. Adaptation to High Contention: As collisions persist, the window grows exponentially. After 10 collisions, stations spread retransmissions across 1024 slot times, dramatically reducing collision probability.

3. Eventual Resolution: Even with many competing stations, the expanding window eventually provides enough space for all to find non-overlapping transmission opportunities.

The Cap at k=10:

Why cap at 1024 slots instead of continuing to double?

• Maximum wait of 52.4 ms is long enough
• Larger windows don't significantly reduce collision probability
• Very high contention indicates a saturated network—need fewer stations or more bandwidth, not longer waits

The collision window (or contention window) is the period during frame transmission when collisions can still occur. Understanding this window is crucial for analyzing network behavior and identifying configuration problems.

The Collision Window Timeline:

1. T = 0: Station A begins transmitting
2. T = 0+ to T < slot_time: Collision window is OPEN
   • Other stations may not yet have heard A's transmission
   • If they transmit, collision occurs
   • A must still be transmitting to detect any collision
3. T = slot_time: Collision window CLOSES
   • A's signal has reached all stations
   • Any collision would have propagated back to A by now
   • If no collision detected, the frame will complete successfully
4. T = slot_time to T = frame_end: Safe transmission period
   • No collisions possible (all stations are deferring)
   • A continues transmitting until frame complete

Key Insight:

Once the collision window closes at T = slot_time (512 bit times), the transmitter is guaranteed to complete the frame without collision. This is why minimum frame size must equal slot time—the transmitter must still be transmitting when the window closes.

As Ethernet evolved to higher speeds, maintaining the same slot time in actual time would have required prohibitively short cables. Different strategies were employed to handle this tension between speed and network diameter.

10 Mbps Ethernet:

The original design. 512 bit times = 51.2 μs allows for up to 2500 meters of cable with repeaters.

100 Mbps Fast Ethernet:

Kept 512 bit times but physical time is only 5.12 μs. This severely limits network diameter:

• Maximum hub-to-hub distance: ~200 meters
• Maximum station-to-hub distance: ~100 meters (Cat5 UTP)

This worked because networks were consolidating into building-centric topologies with short cable runs.

1 Gbps Gigabit Ethernet (Half-Duplex):

The IEEE faced a dilemma:

• 512 bit times at 1 Gbps = 512 ns
• This would limit cables to ~20 meters—impractical!

Solution: Carrier Extension

The slot time was increased to 4096 bit times (4.096 μs), maintaining 200-meter cable support. Frames shorter than 4096 bits are "extended" with a special carrier symbol to fill the slot time.

• Minimum frame: still 64 bytes (for compatibility)
• Minimum transmission: 4096 bit times (including extension)
• Extension doesn't contain data—just holds the medium

Frame Bursting was introduced to improve efficiency: multiple short frames could be sent in succession, sharing the extension overhead.

1 Gbps+ in Full Duplex:

In practice, Gigabit Ethernet almost universally uses full duplex with switches. No shared medium = no collisions = no CSMA/CD = no slot time requirement. The slot time is simply not applicable.

For networks that still use half-duplex segments (increasingly rare), slot time calculations are essential for valid design. Let's examine how network designers work with slot time constraints.

Path Delay Value (PDV) Model:

The IEEE 802.3 standard provides a detailed delay budget approach:

1. Calculate round-trip delay for each segment type
2. Sum all delays along the worst-case path
3. Verify total is ≤ 575 bit times (for 10/100 Mbps)

Wait—why 575 bit times if slot time is 512?

The 575 bit times includes:

• 512 bit times for slot time
• 64 bit times for detecting the collision
• Safety margin

Delay Budget Components:

Each network element contributes to the path delay value:

• Cables: Delay = length × velocity factor adjustment

  • Cat5 UTP: 1.112 bit times per meter (at 100 Mbps)
  • Fiber: varies by type

• Repeaters/Hubs: Each class has specified maximum delay

  • Class I repeater (100 Mbps): 140 bit times
  • Class II repeater (100 Mbps): 92 bit times

• NICs/DTEs: Each manufacturer specifies NIC delay

  • Typical: 50-100 bit times per end

Let's formalize the relationship between slot time and network efficiency through mathematical analysis.

The Efficiency Parameter 'a':

Network efficiency in CSMA/CD is heavily influenced by the parameter 'a':

$$a = \frac{t_{propagation}}{t_{frame}} = \frac{t_{propagation} \times Bandwidth}{Frame Size}$$

Since slot time ≈ 2 × t_propagation:

$$a \approx \frac{\text{Slot Time}}{2 \times t_{frame}}$$

For minimum-sized frames where t_frame = slot_time:

$$a_{min} = \frac{1}{2} = 0.5$$

Maximum Channel Efficiency:

Under ideal conditions with Poisson arrivals:

$$\eta_{max} = \frac{1}{1 + 2a}$$

For minimum-sized frames (a = 0.5):

$$\eta_{max} = \frac{1}{1 + 1} = 0.5 = 50%$$

For maximum-sized frames (1518 bytes = 12144 bits):

$$a = \frac{512}{2 \times 12144} = 0.021$$

$$\eta_{max} = \frac{1}{1 + 0.042} \approx 96%$$

Key Insight:

Slot time represents the overhead of the CSMA/CD protocol. The larger the slot time relative to frame size, the more overhead and the lower the efficiency. This is why:

1. Maximum-sized frames are more efficient than minimum-sized frames
2. Higher-speed networks need proportionally shorter slot times (or accept shorter cables)
3. Batch processing (sending large frames) is more efficient than small messages

Numerical Example:

Consider a 10 Mbps Ethernet with 40% utilization sending minimum-sized frames:

• Useful throughput: 10 Mbps × 40% = 4 Mbps
• But minimum-sized frame efficiency caps at 50%
• Actual maximum with min frames: 10 Mbps × 50% × utilization_factor
• Heavy contention further reduces this

With maximum-sized frames at 96% efficiency:

• Achievable throughput: 10 Mbps × 96% × 0.4 = 3.84 Mbps for same utilization

The choice of frame size significantly impacts network capacity.

Slot time is the fundamental timing parameter that makes CSMA/CD work. Understanding it reveals the deep engineering behind Ethernet's success. Let's consolidate our knowledge:

What's Next:

Slot time directly determines how far Ethernet signals can travel. In the next page, we'll explore Maximum Segment Length—the cable distance limits for different Ethernet standards and how they derive from slot time constraints, signal attenuation, and physical layer characteristics.