Communication between parties often follows a natural turn-taking pattern—one speaks while the other listens, then roles reverse. This conversational model translates directly into half-duplex transmission, a fundamental mode where data can flow in both directions, but not simultaneously.

Half-duplex represents a middle ground between the strict unidirectionality of simplex and the simultaneous bidirectionality of full-duplex. It enables two-way communication using a single channel by alternating the direction of transmission—a design that balances capability against cost and complexity.

Half-duplex transmission is a communication mode where data can flow in both directions, but only one direction at a time. Both stations have the capability to transmit and receive, but they must take turns—when one transmits, the other receives, and vice versa.

The formal definition captures the essential characteristics:

Half-Duplex Communication: A bidirectional data transmission mode using a shared channel where either end can transmit or receive, but mutual exclusion prevents simultaneous transmission in both directions. Direction changes require explicit or implicit turnaround coordination.

This definition highlights several critical aspects:

Channel Sharing Mechanics:

In half-duplex systems, the shared channel alternates between serving as a forward channel and a return channel. This creates a temporal pattern:

Time:      ─────────────────────────────────────────────▶
            ┌─────────┐         ┌─────────┐         ┌───
Station A:  │Transmit │ Receive │Transmit │ Receive │Tx
            └─────────┘         └─────────┘         └───
            ┌─────────┐         ┌─────────┐         ┌───
Station B:  │ Receive │Transmit │ Receive │Transmit │Rx
            └─────────┘         └─────────┘         └───
                      ▲                   ▲
                      │                   │
                  Turnaround          Turnaround

The gaps between transmission phases represent turnaround time—the period required to switch transmission direction. This overhead is a fundamental cost of half-duplex operation.

Turnaround time is the interval required to reverse the direction of transmission on a half-duplex channel. This critical parameter directly impacts system efficiency and is often the determining factor in choosing between half-duplex and full-duplex designs.

Impact on Channel Efficiency:

Turnaround time directly reduces the proportion of time available for actual data transmission. The channel utilization efficiency can be expressed as:

η = T_data / (T_data + T_turnaround)

Where:

• η = channel efficiency (0 to 1)
• T_data = time spent transmitting data
• T_turnaround = total turnaround time

Example Calculation:

Consider a half-duplex wireless system:

• Data transmission: 100 ms
• Turnaround time: 10 ms per switch
• A typical request-response requires 4 turnarounds (request → response → ACK → ready)

Efficiency = 100 ms / (100 ms + 40 ms) = 71.4%

Nearly 30% of potential channel capacity is lost to turnaround overhead. For interactive protocols with frequent direction changes, this loss compounds dramatically.

Strategies to Minimize Turnaround Impact:

1. Message Aggregation — Combine multiple data units into single transmissions to amortize turnaround cost across more data.

2. Sliding Window Protocols — Allow multiple frames to be sent before requiring acknowledgment, reducing turnaround frequency.

3. Piggybacking — Attach acknowledgments to data frames traveling in the opposite direction, eliminating separate ACK turnarounds.

4. Fast Turnaround Hardware — Invest in transceivers with rapid switching times, particularly critical for high-frequency trading or real-time control.

5. Asymmetric Design — If traffic is naturally asymmetric (e.g., mostly downloads), optimize for the dominant direction and accept overhead in the minority direction.

6. Full-Duplex Migration — When turnaround overhead exceeds the cost of full-duplex infrastructure, transition to simultaneous bidirectional operation.

In half-duplex systems, collision occurs when both stations attempt to transmit simultaneously—violating the fundamental mutual exclusion requirement. Managing contention for transmission rights is a central challenge in half-duplex protocol design.

CSMA/CD Deep Dive:

Carrier Sense Multiple Access with Collision Detection was the foundation of classic Ethernet (10BASE5, 10BASE2, 10BASE-T in hub configurations) and exemplifies half-duplex collision management.

Algorithm Steps:

1. Carrier Sense: Before transmitting, listen to the medium. If busy, wait.
2. Transmission: If medium idle, begin transmitting.
3. Collision Detection: While transmitting, monitor the medium. If transmitted signal differs from detected signal, collision has occurred.
4. Jam Signal: Upon collision detection, transmit a 32-bit jam signal to ensure all parties detect the collision.
5. Backoff: Wait a random time (exponential backoff) before attempting retransmission.
6. Retry: Repeat from step 1, with increasing backoff range for subsequent collisions.

Critical Timing Constraint:

For collision detection to work, the minimum frame size must ensure the transmitter is still transmitting when reflections of its signal return from the farthest point on the network. This constraint led to Ethernet's 64-byte minimum frame size:

Minimum Frame Time ≥ 2 × Maximum Propagation Delay

This ensures the transmitter can detect any collision before completing its frame.

Collision Impact on Performance:

Collisions waste channel capacity proportionally to:

• Time spent transmitting collided frames
• Time spent transmitting jam signals
• Time spent in backoff before retry
• Increased backoff times under heavy load

Under light load (<30% utilization), collisions are rare and CSMA/CD performs well. As load increases, collision probability rises exponentially. Beyond ~70% utilization, classic Ethernet can enter collision avalanche—a state where most transmission attempts result in collisions, throughput drops dramatically, and latency becomes unbounded.

This behavior explains the industry's migration to switched Ethernet, where each switch port is a separate collision domain (effectively point-to-point full-duplex), eliminating the half-duplex collision problem entirely.

Half-duplex operations require explicit mechanisms to coordinate direction changes. These protocols range from simple implicit agreements to sophisticated multi-station coordination systems.

Stop-and-Wait: The Simplest Half-Duplex Protocol

Stop-and-Wait represents the most straightforward half-duplex data link protocol:

1. Sender transmits a single frame
2. Sender stops and waits for acknowledgment
3. Receiver processes frame and sends ACK
4. Sender receives ACK and may transmit next frame

Efficiency Analysis:

Let:

• L = Frame length (bits)
• R = Data rate (bits/second)
• RTT = Round-trip propagation time
• T_frame = L/R (transmission time)

Utilization = T_frame / (T_frame + RTT)

This efficiency degrades severely on high-latency links. For a 1000-bit frame at 1 Mbps over a satellite link (RTT = 500 ms):

• T_frame = 1000 / 1,000,000 = 1 ms
• Utilization = 1 ms / (1 ms + 500 ms) = 0.2%

The channel is idle 99.8% of the time! This inefficiency motivated sliding window protocols that allow multiple outstanding frames.

Sliding Window Protocols: Efficient Half-Duplex

Sliding window protocols improve half-duplex efficiency by allowing multiple frames in transit before requiring acknowledgment:

Go-Back-N (GBN):

• Sender may have up to N unacknowledged frames outstanding
• Receiver only accepts frames in order; discards out-of-order frames
• On timeout, sender retransmits from most recent unacknowledged frame
• Simpler receiver; potentially wasteful retransmission

Selective Repeat (SR):

• Sender may have up to N unacknowledged frames outstanding
• Receiver buffers out-of-order frames for later delivery
• On timeout or NAK, sender retransmits only specific lost frames
• More efficient retransmission; more complex receiver

Both protocols can achieve high efficiency on links where BDP is much larger than a single frame, making them essential for practical half-duplex data communication.

Half-duplex transmission remains prevalent in numerous systems where the cost savings from channel sharing outweigh the efficiency loss from turnaround overhead.

Case Study: Push-to-Talk (PTT) Radio

The walkie-talkie is the quintessential half-duplex system, used by emergency services, military, construction, and hobbyists worldwide.

Operational Characteristics:

• Single radio frequency shared for transmit and receive
• Physical push-to-talk button activates transmitter
• Releasing button switches to receive mode
• Verbal conventions ("over") signal transmission complete
• "Roger" confirms receipt (informal ACK)

Technical Implementation:

• Superheterodyne receiver with transmitter sharing some oscillator stages
• Transmit/receive switch isolates antenna from receiver during transmission
• Squelch circuit mutes receiver noise when no signal present
• VOX (voice-operated switch) optional for hands-free operation

Human Protocol: Remarkably, humans naturally develop efficient half-duplex protocols:

• Short, focused transmissions
• Wait for acknowledgment before continuing
• Clear turn-taking signals
• Priority conventions for emergencies

This informal human protocol mirrors engineered data link protocols—a testament to the naturalness of turn-taking communication.

Case Study: Classic Ethernet with Hubs

Early Ethernet networks using hubs operated in half-duplex mode:

Network Topology:

• All stations connected to central hub
• Hub regenerates and broadcasts all received signals to all ports
• Creates a shared collision domain—all stations hear all traffic

Half-Duplex Operation:

• Only one station may transmit at a time
• CSMA/CD protocol governs access
• Collisions detected by voltage level comparison
• Exponential backoff prevents collision storms

Performance Reality:

• 10 Mbps physical rate yields 3-5 Mbps actual throughput under load
• Latency unpredictable during collision conditions
• Performance degrades non-linearly as stations added

Evolution to Full-Duplex: The transition from hubs to switches transformed Ethernet:

• Each switch port becomes isolated collision domain
• Point-to-point links enable full-duplex operation
• No collisions; no CSMA/CD; deterministic performance
• Effective throughput approaches physical rate

This evolution demonstrates the practical limits of half-duplex in high-performance networks.

Half-duplex occupies a strategic middle ground between simplex and full-duplex, offering bidirectional capability at reduced cost—but with inherent performance trade-offs.

Decision Framework: When to Choose Half-Duplex

Half-duplex is optimal when:

1. Bidirectional communication is required — Unlike simplex, either party needs to initiate communication.

2. Traffic is naturally asymmetric or turn-taking — Request-response, master-slave queries, or conversational patterns fit half-duplex well.

3. Cost constraints rule out full-duplex — The expense of separate channels or simultaneous TX/RX capability is prohibitive.

4. Physical constraints require single-channel — Limited frequency allocation or single-wire interfaces mandate sharing.

5. Power budget limits simultaneous operation — Battery-powered devices benefit from half-duplex power savings.

6. Moderate latency is acceptable — Applications can tolerate turnaround delays and occasional contention.

Avoid Half-Duplex When:

• Simultaneous bidirectional traffic is common
• Latency requirements are stringent
• Throughput must approach physical link capacity
• The cost of full-duplex is justifiable
• High reliability and deterministic timing are critical

Understanding half-duplex performance requires quantitative comparison with simplex and full-duplex modes across key metrics.

Throughput Analysis Example:

Consider a 100 Mbps physical link with 1 ms turnaround time and 10 ms average data burst:

Half-Duplex Efficiency:

• Time per exchange = 10 ms (data) + 1 ms (turnaround) = 11 ms
• Efficiency = 10/11 = 90.9%
• Effective throughput = 100 Mbps × 90.9% = 90.9 Mbps (one direction at a time)
• Bidirectional aggregate = 90.9 Mbps (shared between directions)

Full-Duplex Throughput:

• 100 Mbps + 100 Mbps = 200 Mbps aggregate (simultaneous)

Comparison: Full-duplex provides 2.2× the aggregate throughput of half-duplex in this scenario. The gap widens with shorter bursts or longer turnaround times.

Latency Analysis:

For a request-response transaction:

Half-Duplex:

• Request transmission + Turnaround + Response transmission + Turnaround
• = T_req + T_turn + T_resp + T_turn

Full-Duplex: (if simultaneous data exists)

• Response can begin before request completes
• = max(T_req, T_resp) + processing time

For small, frequent transactions, half-duplex latency can be 2× or more than full-duplex.

Half-duplex transmission provides bidirectional communication capability using a shared channel, with direction alternation enforced by protocol. It represents a practical compromise between simplex's simplicity and full-duplex's capability. Let's consolidate the key insights:

Looking Ahead:

With simplex and half-duplex understood, we are now prepared to explore full-duplex transmission—the simultaneous bidirectional mode that maximizes channel utilization. Full-duplex eliminates turnaround overhead entirely but requires either separate channels or sophisticated isolation techniques, introducing its own engineering trade-offs.