1

CS 168, Spring 2026 @ UC Berkeley

Slides credit: Sylvia Ratnasamy, Rob Shakir, Peyrin Kao, Prabal Dutta (and others)

Wireless Links

Lecture 24 (Wireless 1)

Why is Wireless Different?

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Brief History of Wireless Communication

Wireless communication predates the Internet!

• 1880s: Photophone (Bell, Tainer) sent data wirelessly using a light beam.
• 1890s: Wireless telegraph (Marconi) sent data using radio waves.
• 1890s: Experiments with millimeter waves (Bose).
• This is becoming an active area of research again!

​

We now live in a world where wireless communication is everywhere.

Wireless Signals

Wireless signals are not packets of data floating in space.

Wireless signals are waves that propagate in all directions.

• Analogy: Ripples in a pond.
• Waves interact with each other, and with the environment.

A

B

Wired vs. Wireless: Key Differences

1. Wireless is fundamentally a shared medium. (Wired is not.)
2. Wireless signals attenuate significantly with distance. (Wired signals do not.)
3. Wireless environments can change rapidly. (Wired environments do not.)
4. Wireless packet collisions are hard to detect. (Wired packet collisions are not.)

​

Differences mostly affect Layer 1 (Physical) and Layer 2 (Link).

Difference: Wireless is a Shared Medium

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Difference: Wired vs. Wireless Links

Wired links:

• Point-to-point (private) by default.��
• Creating multi-point buses requires work.�
• Fairly easy to shield from external interference.��
• Uses electrical signals to transmit data.

Wireless links:

• Broadcast (shared) by default.��
• Creating point-to-point private links requires work.�
• Fairly hard to shield from external interference.��
• Modulate electromagnetic fields to transmit data.

Encoding Data Over Wireless Links

Wired link: Encode bits as electrical signals.

• High voltage = 1.
• Low voltage = 0.

​

Wireless link:

• Draw the bits as a wave?
• Problem: Resulting wave is low-frequency, and hard to transmit.

0 0 1 1 0 0 0 0 0 1 0 1 0 1 1 1 1 0 0 0 0

Encoding Data Over Wireless Links – Modulation

Modulation: Impose our data signal on top of a carrier signal.

• Carrier signal: A high-frequency, constant wave that contains no information.
• The combined wave is easy to transmit, and contains our data!

Amplitude Modulation (AM):

1 = Taller wave.

0 = Shorter wave.

Frequency Modulation (FM):

1 = Oscillate fast.

0 = Oscillate slow.

Original signal

+ Carrier signal

= Modulated signal

Other modulation strategies exist.

Measuring Noise and Interference – SINR

Shared medium → other signals can corrupt our data!

• Noise: Background, ambient signals.
• Interference: Another transmitter sending signals.

SINR (Signal to Interference and Noise Ratio) lets us measure connection quality:

​

​

​

• Ratio of power to noise+interference at the the receiver.
• Higher SINR is better.
• If there's more noise+interference, the signal must be stronger.
• If signal is weak, can employ coding gain (error-correcting codes).

Measuring Noise and Interference – SINR

SINR is dimensionless (it's a ratio).

Decibels let us measure ratios on a logarithmic scale.

• Ratio is 10 times greater = increase of 10 dB.

SINR formula.

SINR (measured in dB).

Measuring Noise and Interference – Noisy Channel Shannon Capacity

Shannon capacity: Theoretical limit of how much data can be sent on a noisy channel.

• Higher bandwidth = can send more data.
• SINR increases = can send more data.
• Stronger signal, or less noise+interference.

How much data can be sent? (bits/sec)

Bandwidth of channel (range of frequencies we can use).

Ratio of signal power to noise+interference.

Measuring Noise and Interference – Noisy Channel Shannon Capacity

Shannon capacity: Theoretical limit of how much data can be sent on a noisy channel.

Example: The plain old telephone system:

• B = 3000 Hz. (Telephones understand frequencies between 300 Hz and 3300 Hz.)
• SINR = 100. (20 dB signal-to-noise ratio.)
• C = 3000 · log2(1 + 100) ≈ 20000 = 20 kbps.

How much data can be sent? (bits/sec)

Bandwidth of channel (range of frequencies we can use).

Ratio of signal power to noise+interference.

Difference: Wireless Signals Attenuate

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Difference: Attenuation

Wireless signals attenuate – they get much weaker over longer distances.

• Our design must account for attenuation.
• Wired signals also attenuate, but effect is far smaller.

​

Trade-off:

• Maximize performance: Accuracy, speed, range.
• Minimize resource use: Power, use less of the frequency spectrum (costs money).
• Trade-off: Better signal requires more resources.

Measuring Attenuation – Free-Space Model

Free-space model (aka line-of-sight model):

• Transmitter and receiver exist in empty space.
• No obstacles, not even Earth's surface.

In this model, the inverse square law applies.

• 10 times as far = signal is 100 times weaker.
• k times as far = signal is k² times weaker.

​

Receiver power

d = distance between transmitter and receiver

Transmitter power

Intuition: Signal propagates out like a sphere.

Signal power is spread over surface of sphere.�Surface area of sphere = 4πr².

"is proportional to"

Distance

Signal Strength

Measuring Attenuation – Friis Equation

The Friis equation accounts for:

• Gain of the transmitter and receiver antennas.
• Aperture (area) of the receiver antenna. Proof omitted.
• Intuition: Larger antenna can capture more signal.
• Distance between antennas (inverse-square law).

Receiver power

Transmitter power

Gains of antennas

Aperture of receiver antenna

Distance (inverse square law)

Measuring Attenuation – Friis Equation, Rewritten

The equation is sometimes written like this:

Or in terms of decibels (logarithmic scale):

Measuring Attenuation – Link Budget

How do we know if the link will actually work?

• Compute a link budget: Add all gains, subtract all losses.

​

​

• Compare:
• Signal power at receiver.
• Receiver sensitivity (minimum signal the receiver can hear).
• If link budget is positive: Pr > Sensitivity. Link works!
• If link budget is negative: Pr < Sensitivity. Link doesn't work!

​

Link margin is difference between receiver signal and sensitivity.

• Bigger link margin = more robust signal.

Measuring Attenuation – Link Budget

Example of computing link budget:

Transmitter

Receiver

Sensitivity: –80 dBm

Transmitted Signal:�+10 dBm

Cable: –0.44 dB

Cable: –2.21 dB

Cable: –0.44 dB

Cable: –2.21 dB

Antenna: +25 dB

Antenna: +25 dB

Signal travels 10km: –120 dB

Received Signal: –65.5 dBm

Link margin = 14.5 dBm > 0.�Our connection works!

Note: dBm = Power relative to 1 milliwatt.

Difference: Changing Environments

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Difference: Environments Change

Wireless environments change rapidly.

• Devices move around.
• Signals reflect and refract off physical obstacles (e.g. buildings, Earth's surface).

Distance

Signal Strength

Free-space model:�Signal weakens over distance.

Distance

Signal Strength

After accounting for obstacles: Signal strength fluctuates!

Path Loss is Messy

Wireless propagation is messy.

• In theory: Signal propagates in all directions.
• In real life: Signal strength depends on environment.

Color = strength of signal.

Characteristics of Path Loss

3 characteristics affect signal strength:

Free-space loss:

– Due to inverse� square law.�– Fluctuates very slowly.

Shadowing:

– Due to obstructions.�– Fluctuates quickly.

Multipath fading:

– Due to signal colliding� with itself.�– Fluctuates very quickly.

Modeling Path Loss – Two-Ray Model

Two-ray model assumes the signal waves travel along two paths:

• Line-of-sight path: Wave arrives with no obstacles.
• Ground-bounce path: Wave reflects off the Earth's surface.

Earth's surface

Sender tower

Receiver tower

Line-of-sight path

Ground bounce path

Modeling Path Loss – Two-Ray Model

Assuming sender and receiver are far enough:

• Waves arrive phase-shifted at the receiver, causing destructive interference.
• Signal strength ∝ 1/d⁴. Drops off much faster than inverse-square law!

Earth's surface

Sender tower

Receiver tower

Line-of-sight path

Ground bounce path

Time

Signal

Time

Signal

=

Signal arrives with phase offset...

...canceling each other out!

Modeling Path Loss – Two-Ray Model

If sender and receiver are far enough:

• Waves arrive phase-shifted at the receiver, causing destructive interference.
• Signal strength ∝ 1/d⁴. Drops off much faster than inverse-square law!

Distance

Signal Strength

Free-space model:

– Signal strength ∝ 1/d².�– Idealized, no obstacles.

Distance

Signal Strength

Two-ray model:

– Signal strength ∝ 1/d⁴.�– Signal bounces off ground.� Causes destructive interference.

Modeling Path Loss – General Ray Tracing Models

General ray tracing models account for other obstacles.

• Signals reflect, scatter, and diffract.
• Most signals arriving at receiver are reflections.
• Run simulations in software. Requires information about environment.

Earth's surface

Sender

Receiver

Obstacle

Obstacle

Obstacle

Modeling Path Loss – General Ray Tracing Models

Free-space model: Signal strength ∝ 1/d².

Two-ray model: Signal strength ∝ 1/d⁴.

General ray tracing model:

​

• Signal dominated by reflections.
• Exponent γ is determined empirically. Usually between –2 and –8.

Receiver power

Transmitter power

Distance

K and γ are empirically determined by the model.

Difference: Collision Detection

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Difference: Collision Detection

Wired collisions are easy to detect.

• On a point-to-point link, collisions might not happen at all.
• There's just one signal on the wire to sense.

​

Wireless collisions are much harder to detect.

• There's a spatial aspect to collisions.
• Signals can collide in one place, but not another place.

A

B

Signals collide here.

No collision here!

No collision here!

Recall: Multiple Access Protocols

Recall: Many ways for devices to share a link.

• Let's start with CSMA: Listen, and transmit when it's quiet.
• Then we'll design some protocols for wireless networks.

CSMA/CD

CSMA

ALOHA

Tokens

Polling

Multiplexing

Taking Turns

Random Access

Multiple Access Protocols

Frequency

Time

CSMA in Wireless Networks

If pairs are well-separated, no problem!

• Goal: A→B and C→D.
• A transmits to B.
• C transmits to D at the same time.
• No collisions!

​

Notice: Signals propagate in all directions (not just toward the destination).

• Arrows are just drawn for convenience.

A

C

B

D

CSMA in Wireless Networks

If pairs are in range of each other, no problem!

• Goal: A→B and C→D.
• A transmits to B.
• C detects transmission. Must wait to transmit to D.
• No collisions! A–B and C–D take turns.

A is quiet now.�My turn!

A

C

B

D

I hear A transmitting!

I'll wait for A to finish.

CSMA in Wireless Networks – Hidden Terminal Problem

Hidden terminal problem:

• Goal: A→B and C→B.
• A senses quiet, and starts transmitting.
• C senses quiet, and starts transmitting.
• Collision at B!

Problem: A and C are out-of-range. They can't detect each other sending.

A

C

B

All is quiet.�Time to send!

All is quiet.�Time to send!

???

CSMA in Wireless Networks – Exposed Terminal Problem

Exposed terminal problem:

• Goal: B→A and C→D.
• B senses quiet, and starts transmitting.
• C senses a collision and doesn't send.

Notice: We could have actually sent simultaneously.

• Some areas have collision, but we don't care. No collisions at the receivers.

A

C

B

D

All is quiet.�Time to send!

I hear B.

I'll be quiet.

This would have been okay!�But C didn't send.

MACA (Multiple Access with Collision Avoidance)

Key problem: CSMA detects collisions at the sender.

• But we only care about collisions at the receiver.

​

Solution: Let's have the receiver announce if it detects collisions.

• New protocol for shared medium:�MACA (Multiple Access with Collision Avoidance).
• Note: In this new protocol, we're not doing carrier sense anymore.

MACA (Multiple Access with Collision Avoidance)

To communicate over MACA:

1. Sender transmits Request to Send (RTS) with length of data.
2. Receiver transmits a Clear to Send (CTS) with length of data.
• This tells sender that it's safe to send. No collisions at receiver.
• This tells everyone in receiver's range to be quiet.

A

B

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up.�I need to receive 1 MB."

A sends 1 MB.

MACA (Multiple Access with Collision Avoidance) – Solving Hidden Terminal Problem

MACA solves the hidden terminal problem.

• Goal: A→B, C→B.
• B now tells everyone in its range to be quiet, using the CTS.
• C won't send anymore. Collision avoided!

A

C

B

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up.�I need to receive 1 MB."

B told me to be quiet.�I won't send.

A sends 1 MB.

C would have sent, but MACA saved the day!

MACA (Multiple Access with Collision Avoidance) – Rules

If you hear a CTS, be quiet until the data is sent.

• CTS contains length of data.
• You can use data length to estimate how long you need to wait.

​

If you hear an RTS, be quiet for one time slot.

• Give the receiver time to send the CTS.
• If you don't be quiet, you might clobber out the CTS.
• After waiting:
• If you hear the CTS: You're in receiver range. Be quiet.
• If you don't hear the CTS: You're not in receiver range. You can send again.

MACA (Multiple Access with Collision Avoidance) – Solving Exposed Terminal Problem

MACA solves the exposed terminal problem, under certain assumptions.

• Goal: B→A and C→D.
• C hears the RTS and defers for one time slot.
• C doesn't hear the CTS, so it's out-of-range of the other receiver, and can send.

A

C

B

D

I hear an RTS.�I'll be quiet for 1 time slot�to avoid clobbering the CTS.

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up. I'm receiving 1 MB."

B sends 1 MB.

C quiet at this time.

I didn't hear the CTS.�I'm not in receiver range.�I can send data!

C can send here too!�(After its own RTS/CTS.)

MACA (Multiple Access with Collision Avoidance) – Solving Exposed Terminal Problem

MACA solves the exposed terminal problem, under certain assumptions.

• Assumes that C can hear the CTS from D over B's data.
• Key problem: MACA requires the sender to listen (for the CTS).
• Contrast with CSMA: Sender just sends.

A

C

B

D

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up. I'm receiving 1 MB."

B sends 1 MB.

C quiet at this time.

RTS: "I want to send 1 MB."

CTS from D.

Did I hear CTS from D,�or data from B?

MACA (Multiple Access with Collision Avoidance) – Collisions

If we send RTS, but don't hear CTS, that means there was a collision!

Apply exponential backoff and wait up to twice as long before sending another RTS.

• Each device maintains a CW (Contention Window) value.
• Pick a random number in [0, CW]. Wait that long before re-sending RTS.

Rules for adjusting CW:

• Minimum value: CW = 2.
• Maximum value: CW = 64.
• On successful RTS/CTS: Set CW ← 2.
• On failed RTS/CTS: Set CW ← 2 × CW, clamped at 64.

Optimization: Acks for Reliability

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

Introducing MACAW (Multiple Access Collision Avoidance for Wireless)

MACAW (Multiple Access Collision Avoidance for Wireless) offers improvements over MACA.

• Acks for reliability.
• Better backoff for fairness.
• DS packets for synchronization.
• RRTS packets for synchronization.

MACAW Feature: Acks for Reliability

MACAW implements acks for reliability:

• If data is lost: No ack! Sender tries again, starting over with a new RTS.
• If ack is lost: No ack! Sender tries again, starting over with a new RTS.
• Since receiver already got data, it can immediately ack, instead of CTS.

Recall end-to-end principle: Reliability implemented at end hosts for correctness.

• We're adding acks as a performance optimization, not for correctness.

A

B

RTS

CTS

Data

Ack

Optimization: Better Backoff for Fairness

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

MACAW Feature: Better Backoff for Fairness

MACA is unfair: Winners keep winning. Losers keep losing.

A

B

CW = 2

CW = 2

A and B try to reserve simultaneously. Suppose A wins.

CW = 2

CW = 4

B loses and doubles CW.

A wins and sets CW = 2.

A has lower CW, will try again sooner.

A wins again!

CW = 2

CW = 8

B loses and doubles CW.

A wins and sets CW = 2.

A has lower CW, will try again sooner.

A wins again!

CW = 2

CW = 16

B loses and doubles CW.

A wins and sets CW = 2.

MACAW Feature: Better Backoff for Fairness

MACAW solution: Have everybody share the same CW.

• Packet header has a field with CW value.
• If you receive a packet, set your CW to value in the packet.

​

MACAW solution: Change CW update rules to be more gentle.

• Multiplicative Increase, Linear Decrease (MILD).
• On successful RTS/CTS/data/ack: CW ← CW – 1, clamped at 2.
• Contrast with MACA, setting CW = 2.
• On failed RTS/CTS: CW ← 1.5 × CW, clamped at 64.
• Contrast with MACA, setting CW ← 2 × CW.

We're simplifying a bit. Technically, this slide is only true if all devices are in range of each other.

Optimization:�DS for Synchronization

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

MACAW Feature: Admit Defeat on Exposed Terminals

Recall the exposed terminal problem.

• Goal: B→A, C→D.
• C must hear an CTS from D before it can start sending.
• But C can't hear the CTS. Gets clobbered by the data from B!

A

C

B

D

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up. I'm receiving 1 MB."

B sends 1 MB.

C quiet at this time.

RTS: "I want to send 1 MB."

CTS from D.

Did I hear CTS from D,�or data from B?

MACAW Feature: Admit Defeat on Exposed Terminals

MACAW (and MACA) admits defeat on the exposed terminal problem.

• Ideally, B→A and C→D send simultaneously. In reality, they must take turns.
• Punchline: If you're in range of sender, you need to be quiet.
• Because you can't hear the CTS you'd need to start sending yourself.

A

C

B

D

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up. I'm receiving 1 MB."

B sends 1 MB.

C quiet at this time.

RTS: "I want to send 1 MB."

CTS from D.

Did I hear CTS from D,�or data from B?

MACAW Feature: Admit Defeat on Exposed Terminals

Punchline: If you're in range of sender, you need to be quiet.

• Solution: Sender transmits a Data Sending (DS) packet before the data.
• Warns everyone in sender's range to be quiet.

A

C

B

D

RTS: "I want to send 1 MB."

CTS: "If you hear this, shut up. I'm receiving 1 MB."

B sends 1 MB.

C quiet at this time.

C stays quiet, as intended.

(Because C can't hear the CTS it'd need to start sending.)

DS: "If you hear this, shut up. I'm sending 1 MB."

MACAW Feature: DS Packet for Fairness

The DS packet also helps with synchronization for fairness.

• To see why, let's see what happens without the DS packet.
• Goal: B→A, C→D.

​

A

C

B

D

B busy sending.

I know exactly when my transmission ends.

When should I try an RTS again?

C attempts RTS.�Fails, because C won't hear CTS.

B attempts RTS immediately after its data. B will win!

C attempts RTS.�Fails, because B sent RTS first.

MACAW Feature: DS Packet for Fairness

B has a huge advantage, because B knows when its data transmission ends.

• B can immediately recapture the channel with another RTS.
• C must guess when to send the RTS.

A

C

B

D

B busy sending.

I know exactly when my transmission ends.

When should I try an RTS again?

C attempts RTS.�Fails, because C won't hear CTS.

B attempts RTS immediately after its data. B will win!

C attempts RTS.�Fails, because B sent RTS first.

MACAW Feature: DS Packet for Fairness

The DS packet synchronizes by telling everybody when the transmission ends.

• Now C can also send the RTS after B is done.
• C and B both have a fair shot at winning!

A

C

B

D

B busy sending.

I know exactly when my transmission ends.

The DS packet tells me when B's transmission ends.

C attempts RTS after B is done.�Fair shot at winning!

B attempts RTS after B is done. Fair shot at winning!

DS: I'm sending 1 MB.

Optimization:�RRTS for Synchronization

Lecture 24, CS 168, Spring 2026

Why is Wireless Different?

• Shared Medium
• Attenuation
• Changing Environments
• Collision Detection

MACAW Optimizations

• Acks for Reliability
• Backoff for Fairness
• DS for Synchronization
• RRTS for Synchronization

MACAW Feature: RRTS Packet for Fairness

Another case where we need synchronization for fairness:

• Goal: A→B, D→C.
• D is doomed. D can't hear the CTS or the DS, and has no idea when to try again.

A

C

B

D

I know when to RTS again.

I heard CTS, so I need to be quiet.

D attempts RTS.

Fails, because C cannot respond.

RTS

CTS

DS

Data

Next RTS (wins).

Next RTS (loses).

MACAW Feature: RRTS Packet for Fairness

Solution: Let C contend on behalf of D.

• Goal: A→B, D→C.
• C sends an RRTS to tell D: Now's a good time to send a RTS.

A

C

B

D

I know when to RTS again.

I heard CTS, so I need to be quiet.

RRTS

RTS

CTS

DS

Data

Next RTS (fair).

Next RTS (fair).

A's done!�I'll tell D.

MACAW Feature: RRTS Packet for Fairness

If C hears an RTS, but can't respond: Send an RRTS when channel frees up.

• This tells D to immediately send an RTS.
• If you hear an RRTS: Be quiet for 2 time slots so the RTS/CTS can happen.

A

C

B

D

I know when to RTS again.

I heard CTS, so I need to be quiet.

RRTS

RTS

CTS

DS

Data

Next RTS (fair).

Next RTS (fair).

A's done!�I'll tell D.

Summary

Wireless is different!

• Shared medium by default
• Attenuation due to distance and obstacles
• Rapidly shifting environments
• Collision detection is different

Use MACA instead of CSMA for collision detection at receivers.

Optimize with MACAW for:

• Reliability
• Backoffs
• Synchronization

​