10 MAC Protocols: Sharing the Channel

10.1 Start With Who Gets to Talk

This section: Start With Who Gets to Talk — Picture twenty battery sensors trying to report through one receiver after power returns.

Choose the Sharing Rule From the Hard Moment

Picture twenty battery sensors trying to report through one receiver after power returns. IEEE 802.15.4 is one low-power radio and channel-sharing standard. Its name alone does not say how the installed devices behave when they all want the channel.

Write the local traffic story. Name each message type, time need, device count, who can hear whom, sleep limit, reply rule, retry limit, receiver, missed-frame action, and owner.

Test one sender, all senders, two hidden from each other, a busy channel, a sleeping receiver, a lost reply, a low battery, and recovery after a long outage. Measure useful delivery, delay, retries, listen time, and stored energy. An idle channel at one sender is not proof that the receiver is free.

Keep urgent local action safe when the shared channel is busy. The network can retry and report, but the device needs a known rule for a missed deadline.

This opening does not say that one access method always wins. Practitioner compares random, listen-before-send, and scheduled choices. Under the Hood examines backoff, hidden senders, timing, replies, load, mixed strategies, and the proof needed at the final site.

Use a simple field drill. Trigger every sensor at once, then repeat with two sensors hidden from each other. Record who sends, who waits, what the receiver gets, and how much energy each retry uses. Keep missed and late frames in the result.

Run the drill after a node, shelf, antenna, channel, report rate, or retry rule changes. If the hard moment no longer fits the promise, change the rule or narrow the promise before release.

A shared channel is like a narrow doorway: if every device talks whenever it wants, useful messages collide or waste battery waiting. MAC protocols are the rules that decide who may transmit, who must listen, and how the network recovers when two devices choose the same moment.

For IoT design, the question is practical before it is theoretical. Decide whether the workload is periodic, alarm-driven, mobile, sleepy, or dense, then choose the channel-sharing rule that can survive that behavior.

10.2 MAC Rules Decide Who Transmits

This section: MAC Rules Decide Who Transmits

10.2.1 Three Ways to Handle a Collision

Every shared-medium MAC design answers the same question with one of three strategies: let transmissions collide and accept the loss, detect a collision while it happens and stop early, or avoid collisions before they start. Classic wired Ethernet could afford the middle option, carrier-sense multiple access with collision detection (CSMA/CD): a station listens to its own cable while transmitting, and because the transmitted and received signal levels are close enough on a shared copper segment, it can recognize an overlapping signal and abort immediately.

Wireless radios cannot do this cheaply. A transmitting radio's own outgoing signal is far louder at its antenna than any incoming signal it might be trying to hear, so most radios are functionally half-duplex: they cannot listen for a collision while they are the one causing it. Carrier sensing before transmission also only proves the channel looked clear to the sender, not that it will stay clear at the receiver—the hidden-terminal problem covered below. Given both limits, most wireless MAC protocols choose the third option, collision avoidance (CSMA/CA): listen first, back off if busy, and reduce the odds of an overlap instead of trying to detect one after it starts.

Figure 10.1 shows what avoidance costs a sender, one decision at a time.

+

Follow the frame in Figure 10.1 from the left. Before anything is sent there is a wait, and that wait is random rather than fixed, so two radios that became ready at the same moment do not start together. Next comes the local listen. A clear channel lets the frame go, while a busy one sends the sender back to waiting instead of into a collision. After transmission the flow splits on whether an acknowledgement was expected at all, and only then on whether one arrived. A missing acknowledgement is not a verdict. It buys a bounded number of retries and then a declared failure. Read the note at the bottom last: a channel that sounded clear here can still be busy at the receiver.

The busy-channel branch returns to bounded backoff; the success branch closes only after the selected profile’s acknowledgement condition is satisfied.

Contention

Devices compete for the same medium. Carrier sensing, random backoff, acknowledgements, and bounded retries help them share moderate traffic without a central schedule.

Scheduling

A coordinator or network service assigns time, frequency, or link opportunities. Access can become predictable and idle listening can fall when synchronization and slot allocation are maintained.

Random Access

Devices transmit with little coordination. This is attractive for sparse long-range telemetry, but successful throughput falls quickly as competing attempts increase.

MAC Family

Best Fit

Failure Mode

IoT Examples

CSMA/CA

Bursty traffic with moderate contention and devices that can sense much of the local channel.

Hidden terminals, retry storms, and battery drain when many nodes compete.

Wi-Fi and IEEE 802.15.4 style networks.

TDMA or scheduled access

Predictable reporting, bounded latency, or duty-cycled devices that can follow a schedule.

Wasted slots, synchronization overhead, and coordinator dependency.

Cellular scheduling and industrial time-slotted networks.

ALOHA-style access

Very sparse, low-duty-cycle telemetry where coordination would cost more than occasional retransmission.

Collisions rise rapidly if many devices transmit at the same time.

LoRaWAN-style uplink behavior and simple event reporting.

10.3 Choose MAC Rules from Evidence

This section: Choose MAC Rules from Evidence

Start with the workload instead of the protocol. Record whether messages are periodic, event-driven, alarm-first, bulk, or interactive. Then test whether devices can hear one another, whether a gateway can coordinate them, how long a battery device may listen, and what happens when a frame is missed.

That record should be tied to a physical layout, because shelves, machinery, walls, and antenna height can change who hears whom even when every node reports the same configured channel.

Look at the two sensing circles in Figure 10.2 before you look at the nodes.

+

Each sensor sits at the centre of its own range, and the gateway sits where the two ranges overlap. Node A and node C are outside each other’s circle, so neither can hear the other. Both can reach gateway B. That geometry is the whole failure. Each sensor checks the channel, finds it idle, and is correct about what it can hear. Both transmit, the signals overlap at the gateway, and the frame is lost. Figure 10.2 is a warning about where the test is run. Carrier sensing happens at the sender, but collisions happen at the receiver.

Evidence

What to Measure

Design Implication

Common Mistake

Traffic pattern

Inter-arrival time, burst size, alarm priority, retry tolerance, and reporting windows.

Periodic traffic can use schedules; irregular traffic may need contention or event slots.

Using a fixed schedule with no event or shared capacity, which may waste slots or delay unpredictable alarms.

Reachability

Which nodes hear each other, which only hear the gateway, and where obstacles create radio shadows.

Hidden terminals may need RTS/CTS, placement changes, more gateways, or scheduled access.

Assuming low average utilization means collisions will be rare.

Latency

Worst-case wait, backoff distribution, retransmission time, and queueing under bursts.

Safety or control messages need bounded access, priority, or a separate path.

Reporting average latency while hiding peak delay during contention.

Energy

Listen time, retry count, wake windows, synchronization cost, and acknowledgement behavior.

Sleeping outside a scheduled slot can save power, but missed synchronization can be costly.

Counting transmit energy while ignoring idle listening and retries.

Minimum retest record: record the traffic pattern and physical layout used for the choice; measure delivery after the retry limit, tail latency, retry or CCA-busy behaviour, and radio-on energy; retest after placement, density, traffic, schedule, or firmware changes.

10.3.1 Practitioner: Compare Mitigations

A warehouse has periodic inventory sensors behind metal shelving. The average channel load is low, but sensors on opposite aisles cannot hear each other and both can reach the gateway. Basic CSMA/CA without receiver-coordinated reservation may still lose frames because each hidden sensor believes the channel is idle. There are three defensible responses.

Improve Visibility

Move gateways, add gateways, or adjust antennas so more nodes can hear each other and the coordinator has cleaner reception.

Add Handshakes

Use RTS/CTS when control-frame overhead costs less than hidden-terminal retransmissions.

Schedule Traffic

Use slots when reports are predictable and deterministic delivery matters more than immediate unscheduled access.

The hidden-terminal problem is one of geometry, so the fix has to travel through the one node that hears both senders. Figure 10.3 walks that exchange through in five steps.

+

The setup line at the top of Figure 10.3 is the whole reason the handshake exists: two senders cannot hear each other, but both reach the same gateway. Step one has the sender ask for a stated length of airtime. Step two does the real work, because the gateway answers out loud, so the hidden node hears a grant it could never have worked out by listening for the sender. In step three that node starts a timer and stays quiet, which is virtual carrier sensing. Data and an acknowledgement follow. The closing lines keep the claim honest: the request itself can still collide, and nothing here removes interference or a weak receiver.

In IEEE 802.11, an optional RTS/CTS exchange can reserve the medium before a unicast data frame. Stations that hear the reservation information defer for the advertised duration through virtual carrier sensing, represented by the Network Allocation Vector (NAV). This handshake descends from MACA (Multiple Access with Collision Avoidance) and its refinement MACAW, media-access designs built specifically for wireless links that cannot rely on wired-style collision detection.

RTS/CTS can reduce some hidden-terminal collisions, but control overhead, control-frame collisions, interference geometry, and the configured frame-size threshold mean that it is not a universal cure.

10.4 MAC Collision and Timing Bounds

This section: MAC Collision and Timing Bounds

10.5 Hidden and Exposed Terminals

This section: Hidden and Exposed Terminals — Carrier sensing answers a sender-side question: “Can I detect activity here?” Delivery is decided at the receiver.

Carrier sensing answers a sender-side question: “Can I detect activity here?” Delivery is decided at the receiver. In a hidden-terminal case, two senders cannot hear one another, both judge the channel idle, and their frames can overlap at the same receiver.

An idle CCA result is not proof that the receiver is clear. Treat rising retries or missing acknowledgements as symptoms, not proof of a collision. Confirm the geometry: the senders cannot hear one another, both can reach or interfere at the same receiver, and controlled overlap makes delivery worse.

Compare the two ranges drawn in the diagram at Figure 10.4 to see why a correct carrier check can still give the wrong answer.

+

Node B is sending to node A. Node C wants to send to node D, and C sits inside B’s sensing range, so C hears the channel as busy and waits. But node D sits outside B’s interference area, so the two transmissions would never have met at either receiver. Nothing is lost in Figure 10.4, and that is the point. This is wasted capacity, not a failed frame. It is the mirror image of the hidden-terminal case, and it needs the opposite fix. One problem asks for more coordination, the other for less waiting.

The opposite pattern is an exposed terminal. A sender waits because it hears a nearby transmission even though its own receiver could have accepted a concurrent frame. The cost is unnecessary waiting and lost spatial reuse, not necessarily packet loss. The Practitioner workbench tests this case.

10.6 Summary

This section: Summary

• MAC protocols decide when a local device may use a shared medium.
• CSMA/CA fits moderate contention when carrier sensing sees enough of the local channel.
• TDMA or scheduled access fits predictable traffic, bounded timing, and duty-cycled devices that can follow a schedule.
• ALOHA-style access can be acceptable for sparse telemetry but degrades quickly as offered load increases.
• Hidden terminals cause receiver-side collisions; exposed terminals waste spatial reuse.
• Production MAC decisions need evidence about traffic, reachability, timing, energy, retries, and retest triggers.

10.7 Key Takeaway

This section: Key Takeaway

10.8 See Also

This section: See Also