Channels, Propagation and Link Engineering

Introduction

A communication system is successful only when the transmitted information can be recovered with acceptable accuracy at the receiver. Between transmitter and receiver lies a physical channel: free space, atmosphere, optical fibre, copper cable, water, or a combination of media. The channel introduces attenuation, delay, distortion, noise, interference, fading and, in some cases, nonlinear effects. Channel and propagation engineering therefore connects electromagnetic theory with system-level design.

At postgraduate level, the channel should not be regarded merely as a passive connection. It is a dynamic, frequency-selective and statistically uncertain system whose characteristics determine modulation, coding, antenna configuration, power allocation, link budget and reliability. Link engineering is the disciplined process of converting channel knowledge into design decisions that satisfy coverage, capacity, availability, quality-of-service and regulatory requirements.

Learning Objectives

After completing this chapter, the reader should be able to:

1. distinguish deterministic, stochastic, time-invariant and frequency-selective channel models;
2. explain the principal propagation mechanisms in radio, guided and optical channels;
3. calculate free-space loss, received power, noise power, link margin and basic fade probability;
4. interpret delay spread, coherence bandwidth, Doppler spread and coherence time;
5. design and evaluate a practical communication link budget;
6. compare diversity, equalization, adaptive modulation and coding as mitigation techniques;
7. assess engineering trade-offs involving capacity, reliability, energy, spectrum and regulation.

1. The Communication Channel as a System

A baseband channel may be represented by an impulse response h(t,τ), where t describes time variation and τ describes propagation delay. For a transmitted signal x(t), the received signal is

where y(t) is the received signal, n(t) is additive noise, and the integration is over all physically significant delays. This linear time-varying model assumes that the channel is linear and that noise can be represented separately. In a time-invariant channel, h(t,τ)=h(τ), and convolution notation gives y(t)=h(t)*x(t)+n(t).

For a narrowband flat-fading link, a frequently used complex-envelope model is

Here h(t) includes path loss, phase rotation and fading. If the signal bandwidth is sufficiently small relative to the channel coherence bandwidth, all spectral components experience approximately the same complex gain. Otherwise, different frequencies experience different gains and the channel is frequency selective.

Important classifications

2. Propagation Mechanisms

2.1 Free-space propagation

In unobstructed line-of-sight propagation, the received power is described by the Friis transmission equation:

P_t and P_r are transmitted and received powers in watts, G_t and G_r are dimensionless antenna gains in the direction of propagation, λ is wavelength in metres, and d is separation in metres. The assumptions are aligned polarization, far-field operation, free space, and a line-of-sight path. Antenna gains must be expressed as ratios, not decibels.

The corresponding free-space path loss is

or, in decibels,

Using λ=c/f, where c is the speed of light and f is frequency, the engineering form is

The constant is approximately 32.44 when frequency is in megahertz and distance in kilometres. This equation describes spreading loss, not atmospheric absorption, rain attenuation or obstruction loss.

2.2 Reflection, diffraction and scattering

Reflection occurs when a wave encounters a surface whose dimensions are large relative to wavelength. Ground reflection can produce constructive or destructive interference. Diffraction allows energy to reach shadowed regions around edges and rooftops, although with reduced strength. Scattering results from objects comparable to or smaller than the wavelength, such as foliage, street furniture and rough surfaces.

A two-ray ground-reflection model illustrates why received power may depart substantially from free-space predictions. At sufficiently large distance and under suitable antenna-height assumptions, the received power approximately follows

where h_t and h_r are effective transmitter and receiver antenna heights. Thus power can decrease approximately as d^-4, rather than d^-2. This is an asymptotic approximation and should not be used near the crossover distance or where terrain and reflection coefficients are unknown.

2.3 Atmospheric and weather effects

Gaseous absorption is significant at particular microwave and millimetre-wave frequencies, especially near oxygen and water-vapour resonance bands. Rain attenuation becomes important as frequency and path length increase. A simplified rain-loss calculation is

where A_R is attenuation in decibels, γ_R is specific attenuation in decibels per kilometre, and d_eff is effective rain-affected path length in kilometres. In engineering practice, γ_R is obtained from a propagation standard and depends on frequency, polarization, rain rate and statistical availability target. Using a local measured or standardized rain-rate distribution is essential.

2.4 Guided propagation

In metallic cables, propagation is governed by distributed resistance, inductance, conductance and capacitance. For a transmission line with primary parameters per unit length R, L, G and C, the propagation constant is

where α is attenuation constant in nepers per metre, β is phase constant in radians per metre, and ω=2πf. For a low-loss line, R≪ωL and G≪ωC, so attenuation and phase velocity can be approximated more simply. Impedance matching is needed to reduce reflections.

Optical fibre guides light by refractive-index contrast. The normalized frequency or V-number is

where a is core radius and n₁ and n₂ are core and cladding refractive indices. For an ideal step-index fibre, single-mode operation occurs approximately when V<2.405. Fibre links are affected by attenuation, chromatic dispersion, polarization-mode dispersion, connector loss and nonlinear effects.

3. Path-Loss and Large-Scale Channel Models

A practical terrestrial model often begins with a reference distance and a path-loss exponent:

PL(d) is path loss in decibels at distance d, d₀ is a reference distance in the far field, n is the path-loss exponent, and X_σ is a zero-mean Gaussian shadowing variable in decibels with standard deviation σ. The model assumes that the environment is statistically similar over the area being studied. It is not a substitute for site-specific measurements when high availability is required.

Typical values of n may be near 2 in free space, approximately 2 to 4 in many urban outdoor environments, and higher in heavily obstructed indoor environments. These are indicative ranges, not universal constants.

The log-distance model separates large-scale variation from small-scale fading. Shadowing represents slow changes caused by buildings, terrain and blockage. Small-scale fading results from interference among multipath components and may change over distances of a fraction of a wavelength.

4. Multipath, Delay Spread and Doppler Effects

A multipath channel can be described by discrete components:

where a_i(t), φ_i(t) and τ_i(t) are the amplitude, phase and delay of path i. Different path delays cause inter-symbol interference (ISI) when delayed replicas overlap adjacent symbols.

The power-delay profile describes average received power versus excess delay. Its first moment is the mean excess delay:

The root-mean-square delay spread is

where P_i is average power in path i. A rough engineering relationship is B_c≈1/(5σ_τ) for about 0.9 correlation, though the numerical factor depends on the chosen correlation criterion and channel definition. Consequently, delay spread does not determine a unique coherence bandwidth without specifying the criterion.

Motion produces Doppler shift. For a single path,

where v is relative speed and θ is the angle between velocity and wave-arrival direction. Maximum Doppler frequency is f_D,max=v/λ. Coherence time is often approximated as inversely proportional to Doppler spread; one commonly used estimate is

This approximation is model- and definition-dependent. A channel is approximately slow fading when the symbol duration is much smaller than coherence time; it is fast fading when channel changes significantly within a symbol interval or coding block.

5. Statistical Fading Models

In a line-of-sight channel with a dominant component, the envelope is often modeled by a Rician distribution. When many scattered components exist without a dominant path, a Rayleigh model is commonly used. The Rician K-factor is

where powers are linear, not decibel quantities. K=0 corresponds to the limiting Rayleigh case, while large K indicates a strong dominant component.

For a flat-fading link with instantaneous signal-to-noise ratio γ, the average received SNR is

where N₀ is one-sided noise spectral density in watts per hertz and B is receiver noise bandwidth in hertz. Thermal noise power is

where k=1.380649×10⁻²³ J/K is Boltzmann’s constant, T is absolute temperature in kelvin, and F is the dimensionless noise factor. In decibel form, at reference temperature 290 K,

where NF=10log₁₀F.

6. Link-Budget Engineering

A link budget accounts for all gains and losses between transmitter output and receiver input:

L_p includes propagation loss, while L_misc includes feeder, connector, polarization, atmospheric, implementation and pointing losses. Effective isotropic radiated power is

The link margin is

where P_r,min is the receiver sensitivity required for the target data rate, modulation, coding and error performance. A positive nominal margin does not guarantee availability because fading and interference are random.

Worked example: terrestrial microwave link

An 18 GHz point-to-point link spans 12 km. The transmitter output is 20 dBm, the transmit and receive antenna gains are 34 dBi and 36 dBi, and feeder losses are 2 dB at each end. Assume an additional 3 dB allowance for miscellaneous losses and a receiver sensitivity of −72 dBm.

Using the free-space formula:

L_fs=32.44+20log₁₀(18000)+20log₁₀(12)

Since log₁₀(18000)≈4.2553 and log₁₀(12)≈1.0792,

L_fs≈32.44+85.11+21.58=139.13 dB.

The received power is

P_r=20+34+36−139.13−2−2−3≈−56.13 dBm.

Therefore,

M=−56.13−(−72)=15.87 dB.

This is the clear-sky nominal margin. A complete design must add rain attenuation, diffraction loss, antenna misalignment, interference and the required fade margin for the specified availability.

7. Capacity, Reliability and the Engineering Trade-off

For an additive white Gaussian noise channel, the theoretical capacity is

where C is capacity in bits per second, B is bandwidth in hertz, and S/N is linear signal-to-noise ratio. This is an upper bound under the AWGN assumptions and arbitrarily long capacity-achieving codes; practical throughput is lower because of protocol overhead, finite block length, fading, interference and implementation limitations.

A designer must distinguish:

• coverage, the geographical region meeting a target service condition;
• availability, the percentage of time a service meets its performance target;
• reliability, the probability of correct operation over a specified interval;
• capacity, the sustainable information rate;
• latency, including propagation, processing, queueing and retransmission delay.

Increasing transmit power can improve coverage but may increase interference and energy consumption. Increasing bandwidth can increase capacity but may raise noise power and spectrum cost. Higher carrier frequency may support wider bandwidth and smaller antennas, but often increases blockage and weather sensitivity.

8. Channel Equalization and Diversity

When delay spread causes ISI, a receiver may use a linear equalizer, decision-feedback equalizer, or maximum-likelihood sequence estimator. Modern multicarrier systems commonly use orthogonal frequency-division multiplexing (OFDM). A cyclic prefix longer than the significant channel delay spread converts linear convolution into approximately circular convolution, allowing frequency-domain one-tap equalization. The prefix reduces spectral efficiency, so its length should not be unnecessarily large.

Diversity reduces the probability that all independently varying replicas fade simultaneously. Forms include:

• space diversity, using separated antennas;
• frequency diversity, transmitting across separated frequencies;
• time diversity, using interleaving and coding;
• polarization diversity, using different polarizations;
• path diversity, exploiting distinct propagation routes.

The branches must be sufficiently uncorrelated for substantial diversity gain. Simply adding antennas close together in a highly correlated field may provide little benefit.

Adaptive modulation and coding selects a mode according to channel quality. A high-order modulation gives high spectral efficiency but requires larger SNR; robust coding and lower-order modulation improve coverage and reliability. Link adaptation should include measurement delay, feedback errors and hysteresis to avoid unstable mode switching.

9. Practical Link-Engineering Workflow

A defensible link design follows a traceable sequence:

1. define service objectives, data rate, latency, outage probability and geographical area;
2. select frequency, bandwidth, duplexing method, antenna configuration and regulatory limits;
3. identify the propagation environment and choose an appropriate channel model;
4. calculate nominal path loss and complete the link budget;
5. calculate receiver noise, interference and required sensitivity;
6. include rain, gaseous absorption, diffraction, polarization, feeder and implementation losses;
7. evaluate fading statistics, availability and fade margin;
8. select mitigation such as diversity, adaptive coding, power control or relay placement;
9. validate the model with measurements, drive tests, channel sounding or pilot statistics;
10. document assumptions, uncertainty ranges and acceptance criteria.

A link budget should be auditable. Every gain and loss needs a physical explanation, a unit, a source and an uncertainty estimate. Double counting is common—for example, including antenna feeder loss once in EIRP and again in total losses.

10. Applied Case: Illustrative Urban 5G Small-Cell Design

Background

Consider an illustrative, fictional deployment of small cells in a dense urban district. The operator seeks 100 Mb/s downlink service to pedestrians and vehicles, with 99% area reliability under normal loading. The candidate band is 3.5 GHz, where blockage and street-canyon multipath are significant.

Problem

A free-space estimate alone predicts adequate coverage, but initial field measurements show deep fades near building corners and large variation between line-of-sight and non-line-of-sight streets. The planning team must determine whether more transmit power, additional sites, or advanced antenna processing is the best response.

Analysis

The team constructs a site-specific model using building data, terrain, antenna heights and measured shadowing. The channel is represented as a combination of deterministic dominant paths and stochastic residual fading. Drive tests estimate path-loss exponent, shadowing deviation, delay spread and spatial correlation. The 100 Mb/s target is translated into a minimum SINR after accounting for coding, reference signals and scheduling overhead.

The analysis shows that increasing power improves some cell-edge locations but raises inter-cell interference. The dominant problem is not average path loss alone; it is correlated blockage and interference at street intersections. A larger fade margin cannot efficiently solve a systematic coverage hole.

Approach and solution

The design uses moderate transmit power, carefully tilted massive-MIMO panels, coordinated scheduling at overlapping sites, and additional low-power nodes near the most obstructed corners. The system applies adaptive modulation and coding, hybrid automatic repeat request, and beam management. Measurements are repeated after installation, and the model is recalibrated using actual reference-signal received power and SINR distributions.

Outcome

The final arrangement improves the lower-tail SINR distribution and reduces outage without excessive interference. The solution costs more sites but requires less radiated power per site and provides greater capacity under load. This illustrates a central link-engineering principle: reliability is often improved more effectively by controlling geometry, interference and diversity than by increasing nominal power.

Learning points

1. Coverage predictions should include uncertainty and spatial variability.
2. A propagation model must be validated against measurements.
3. Link margin, capacity and interference are coupled design variables.
4. The best engineering solution is usually a system-level combination of antennas, sites, coding and scheduling.

11. Contemporary Developments and Research Directions

Massive MIMO and beamforming exploit spatial degrees of freedom, but performance depends on channel-state information, calibration, pilot contamination and user mobility. Reconfigurable intelligent surfaces seek to control propagation through programmable scattering; their practical benefit depends on insertion loss, channel estimation, control overhead and deployment geometry. Terahertz systems offer very wide bandwidth but face severe molecular absorption, blockage and hardware limitations.

Machine learning is increasingly used for propagation-map construction, beam prediction, handover optimization and anomaly detection. It should complement, not replace, physical models. A model trained in one city, frequency band or weather regime may fail elsewhere. Engineering governance therefore requires representative data, uncertainty quantification, explainable validation and protection against biased or incomplete measurements.

For satellite and high-altitude links, long propagation delay, Doppler variation, atmospheric losses and link availability dominate design. For optical wireless systems, eye safety, ambient light, pointing and obstruction are central. For underwater acoustic channels, very low propagation speed, severe multipath and frequency-dependent absorption require different models from radio systems.

12. Regulatory, Safety and Ethical Considerations

Link engineering must comply with spectrum allocation, emission masks, maximum EIRP, antenna installation rules and electromagnetic-exposure limits. Interference coordination is both a technical and regulatory responsibility. Accurate records of frequencies, powers, antenna patterns and site geometry support accountability.

Environmental considerations include energy consumption, equipment lifecycle, tower construction and electronic waste. Measurements involving people or locations may reveal sensitive movement patterns; data collection should therefore follow privacy, security and data-minimization principles. AI-assisted planning must not conceal uncertainty or systematically disadvantage communities through unequal service quality.

Challenges and Limitations

No single propagation model is universally accurate. Free-space equations are elegant but idealized; empirical models depend on calibration; ray tracing can be computationally expensive and sensitive to incomplete maps; stochastic models may omit rare but consequential blockage events. Noise-limited assumptions fail in dense networks where interference dominates. Similarly, a nominal link margin does not directly equal availability unless the fading, rain and interference distributions are known.

The most reliable practice is model triangulation: use analytical bounds, empirical or ray-based prediction, measurements and statistical uncertainty analysis. Designs should report percentile performance rather than only average received power.

Best-Practice Checklist

• Use linear ratios inside power equations and decibels for additive budgets.
• State frequency, distance, antenna gains, polarization and reference conditions.
• Separate path loss, fading margin, interference margin and implementation loss.
• Check near-field and far-field validity before applying Friis’ equation.
• Use channel bandwidth, not automatically symbol rate, in noise calculations.
• Validate propagation predictions using measurements.
• Design to outage and availability targets, not only median performance.
• Document model uncertainty and avoid double counting losses.
• Reassess the design under mobility, weather, blockage and traffic loading.

Chapter Summary

Communication channels impose attenuation, delay, distortion, noise, interference and time variation. Their behavior is modeled through impulse responses, path-loss laws, multipath statistics and propagation mechanisms. The Friis equation is appropriate for ideal far-field free-space links, whereas practical systems require empirical, deterministic or stochastic models that account for terrain, buildings, weather and motion.

Delay spread determines the severity of frequency selectivity and ISI; Doppler spread determines temporal variation. Noise power depends on bandwidth, temperature and receiver noise figure. A link budget combines transmitter power, antenna gains and all relevant losses to obtain received power and link margin. Capacity, reliability, coverage, latency and interference must be optimized together. Equalization, OFDM, diversity, adaptive modulation, coding, beamforming and coordinated deployment are principal mitigation strategies.

Professional link engineering is an iterative process involving requirements, modeling, calculation, measurement, uncertainty analysis and regulatory compliance. Contemporary systems increasingly use massive MIMO, intelligent surfaces, machine learning and higher frequencies, but physical validation and transparent engineering judgment remain essential.

Key Terms

Channel: Physical medium and associated input-output behavior through which a signal propagates.

Propagation: Physical transmission of electromagnetic, optical or acoustic energy.

Path loss: Reduction in average received power caused by propagation.

Friis equation: Free-space relation between transmitted and received power.

EIRP: Equivalent isotropically radiated power.

Link budget: Accounting of gains and losses in a communication path.

Link margin: Difference between received power and required receiver sensitivity.

Shadowing: Slow spatial variation caused by blockage and environmental structure.

Multipath: Arrival of multiple replicas of a transmitted signal.

Delay spread: Statistical dispersion of multipath arrival times.

Coherence bandwidth: Frequency range over which channel gains are strongly correlated.

Doppler spread: Frequency broadening caused by relative motion.

Coherence time: Approximate interval over which the channel remains correlated.

Rayleigh fading: Fading model without a dominant received component.

Rician fading: Fading model with a dominant component plus diffuse scattering.

Noise figure: Degradation of signal-to-noise ratio introduced by a receiver.

Inter-symbol interference: Symbol overlap caused by channel dispersion.

Diversity: Use of multiple independently faded signal replicas.

Availability: Fraction of time a specified service-performance condition is met.

Beamforming: Spatial shaping of transmitted or received signals using antenna arrays.

Review Questions

Short Answer Questions

1. What assumptions are required for the Friis transmission equation?
2. Distinguish flat fading from frequency-selective fading.
3. Define delay spread and explain its relationship to inter-symbol interference.
4. What is the difference between antenna gain and EIRP?
5. Why is a positive nominal link margin not sufficient to guarantee availability?

Long Answer Questions

1. Explain the major propagation mechanisms in terrestrial wireless communication and discuss when each becomes important.
2. Derive and interpret a complete link-budget equation, including receiver sensitivity and link margin.
3. Discuss the origins of multipath fading and compare Rayleigh and Rician channel models.
4. Explain how OFDM, equalization and diversity mitigate channel impairments.
5. Critically evaluate analytical, empirical and ray-tracing propagation models.

Analytical / Application Questions

1. A 6 GHz line-of-sight link is 8 km long. The transmitter power is 17 dBm, antenna gains are 28 dBi and 30 dBi, and total miscellaneous loss is 5 dB. Calculate free-space loss, received power and margin if receiver sensitivity is −78 dBm.
2. A channel has three paths with delays 0, 1 and 3 μs and average powers 1, 0.5 and 0.25 in arbitrary but consistent units. Calculate the mean excess delay and RMS delay spread, then discuss whether a signal with bandwidth much greater than the reciprocal of that spread is likely to experience frequency selectivity.
3. An urban network has adequate median received power but poor fifth-percentile SINR near building corners. Recommend a measurement and redesign strategy, justifying the roles of site placement, beamforming, interference coordination, diversity and adaptive coding.

Multiple Choice Questions with Answers

Question 1

Which equation gives the received power for an ideal far-field free-space link?

A. P_r = P_t G_t G_r (λ/(4πd))²

B. P_r = P_t + G_t + G_r + d

C. P_r = kTBF

D. C = B log₂(1+S/N)

Correct Answer: A. P_r = P_t G_t G_r (λ/(4πd))²

Question 2

In the log-distance path-loss model, what does the term X_σ represent?

A. Antenna gain

B. A zero-mean shadowing variable in decibels

C. Receiver bandwidth

D. Maximum Doppler frequency

Correct Answer: B. A zero-mean shadowing variable in decibels

Question 3

A channel is most likely frequency selective when:

A. Its signal bandwidth is much smaller than its coherence bandwidth

B. Its signal bandwidth is approximately zero

C. Its signal bandwidth exceeds its coherence bandwidth

D. It has no multipath components

Correct Answer: C. Its signal bandwidth exceeds its coherence bandwidth

Question 4

At 290 K, a receiver with 1 MHz bandwidth and 5 dB noise figure has approximately what thermal noise power?

A. −169 dBm

B. −159 dBm

C. −139 dBm

D. −94 dBm

Correct Answer: C. −139 dBm

Question 5

What is the principal purpose of a cyclic prefix in OFDM?

A. To increase antenna gain

B. To reduce carrier frequency

C. To mitigate inter-symbol interference from multipath

D. To eliminate thermal noise

Correct Answer: C. To mitigate inter-symbol interference from multipath

Question 6

The Rician K-factor is defined as:

A. The ratio of line-of-sight power to scattered power

B. The ratio of bandwidth to delay spread

C. The ratio of noise power to signal power

D. The ratio of antenna height to wavelength

Correct Answer: A. The ratio of line-of-sight power to scattered power

Question 7

If a link has received power −60 dBm and receiver sensitivity −75 dBm, its nominal link margin is:

A. −135 dB

B. −15 dB

C. 15 dB

D. 135 dB

Correct Answer: C. 15 dB

Question 8

Which change generally increases maximum Doppler frequency for a fixed carrier frequency?

A. Reducing receiver noise figure

B. Increasing relative velocity

C. Reducing bandwidth

D. Increasing coding gain

Correct Answer: B. Increasing relative velocity

Question 9

Why may simply increasing transmit power fail to solve an urban coverage problem?

A. Power cannot affect received signal

B. It may increase interference while leaving blockage geometry unchanged

C. It always reduces antenna gain

D. It eliminates multipath and therefore causes outage

Correct Answer: B. It may increase interference while leaving blockage geometry unchanged

Question 10

The Shannon AWGN capacity formula is best understood as:

A. An exact throughput for every practical modem

B. An upper theoretical bound under stated channel assumptions

C. A formula for free-space path loss

D. A formula valid only for optical fibre

Correct Answer: B. An upper theoretical bound under stated channel assumptions