Wireless Physical-Layer Techniques

Introduction

A wireless physical layer converts information bits into electromagnetic waveforms and recovers them despite noise, interference, fading, mobility and hardware imperfections. Reliable operation rarely results from one technique alone. Modern links combine modulation, channel coding, interleaving, multicarrier transmission, equalization, diversity, multiple-input multiple-output (MIMO) processing and beamforming.

The principal engineering task is to balance competing goals: data rate, bandwidth, energy efficiency, reliability, latency, coverage and implementation complexity. For example, high-order modulation increases bits per symbol but requires a higher signal-to-noise ratio (SNR), while strong coding improves reliability but consumes transmission resources.

Learning Objectives

After studying this chapter, a student should be able to:

• model time-varying multipath wireless channels;
• select modulation and coding schemes for specified link conditions;
• explain interleaving and spread-spectrum operation;
• derive OFDM subcarrier spacing, cyclic-prefix requirements and throughput;
• apply diversity combining and equalization principles;
• distinguish MIMO spatial multiplexing from diversity and beamforming;
• calculate MIMO spectral efficiency and capacity under stated assumptions; and
• integrate physical-layer techniques into a practical wireless-link design.

1. The Wireless Channel

1.1 Time-varying multipath model

A transmitted signal may arrive through reflection, diffraction and scattering paths. Each path has its own attenuation, delay, phase and Doppler shift. The general linear time-varying baseband model is

Here, x(t) is the transmitted complex-baseband signal, y(t) is the received signal, h(t,τ) is the time-varying channel impulse response at observation time t and excess delay τ, and n(t) is additive noise and interference. A causal channel has h(t,τ)=0 for τ<0.

Only when the channel is time invariant may the model be reduced to ordinary convolution:

Using ordinary convolution for a rapidly time-varying mobile channel would therefore conceal Doppler-induced variation.

A sampled channel can be represented as

where L is the number of significant delay taps and h[ℓ,m] may change with sample index ℓ.

1.2 Delay spread and frequency selectivity

Multipath delay spread determines whether different frequency components experience different gains. The root-mean-square delay spread τ_rms is calculated from the power-delay profile. A commonly used engineering approximation is

where B_c is an approximate coherence bandwidth in hertz. The numerical factor depends on the chosen frequency-correlation threshold, so this expression is a design heuristic rather than a universal equality.

If the signal bandwidth is much smaller than B_c, the channel is approximately frequency-flat. If it is comparable to or greater than B_c, the channel is frequency-selective and can produce intersymbol interference (ISI).

1.3 Doppler spread and coherence time

For a terminal moving at speed v, the maximum Doppler magnitude is

where f_c is carrier frequency, c≈3×10⁸ m/s is propagation speed, and f_D is in hertz. This is the maximum magnitude when motion is aligned with arrival or departure direction; an individual path has Doppler f_D cos θ.

One frequently used coherence-time approximation is

for isotropic scattering under a specified correlation convention. If a frame is much shorter than T_c, the channel may be treated as approximately constant during that frame.

2. Link Performance Measures

The received SNR over bandwidth B is

where P_r is received signal power in watts and N₀ is the one-sided noise power spectral density in W/Hz under the stated complex-baseband convention. Receiver noise figure must be included when determining effective noise power.

For information-bit rate R_b,

because E_b=P_r/R_b. This relationship requires consistent definitions of bandwidth, SNR and information-bit rate.

For an ideal flat additive white Gaussian noise (AWGN) channel, Shannon capacity is

where C is in bit/s, B is in hertz and γ is a linear power ratio. It is an upper bound based on arbitrarily long coding and Gaussian signalling, not the rate of a particular practical modem.

3. Digital Modulation

3.1 Mapping bits to symbols

An M-ary modulation maps groups of

coded bits to each symbol. Common schemes include BPSK, QPSK, quadrature amplitude modulation (QAM) and frequency-shift keying.

Higher-order QAM points are closer together for a fixed average power. Consequently, noise, phase error and nonlinear distortion cause more decision errors.

For coherent BPSK in an AWGN channel, the uncoded bit-error probability is

where Q(x) is the upper-tail probability of a standard normal variable. This formula does not directly apply to fading channels or coded BPSK.

3.2 Pulse shaping

Abrupt rectangular pulses occupy excessive bandwidth. Raised-cosine or root-raised-cosine filtering controls spectral occupancy while satisfying the Nyquist zero-ISI criterion under ideal timing. For symbol rate R_s and roll-off factor α, the approximate null-to-null passband bandwidth for linearly modulated QAM is

where 0≤α≤1. Practical systems must also accommodate oscillator error, filters and regulatory guard bands.

Worked Example: Coded 16-QAM

A link uses 16-QAM at R_s=2 Msymbol/s with code rate R_code=3/4. Each symbol carries four coded bits. The information rate is

Thus, the ideal information rate before pilots, framing and retransmissions is 6 Mbit/s.

4. Channel Coding and Interleaving

A channel encoder introduces controlled redundancy so that a decoder can detect or correct errors. If each block of k information bits produces n coded bits, the code rate is

where 0<R_code≤1. Lower code rates generally provide more protection but reduce the information rate for a fixed coded-symbol rate.

Convolutional codes are commonly decoded using the Viterbi algorithm. Turbo codes and low-density parity-check (LDPC) codes use iterative probabilistic decoding. Polar codes transform channels into reliable and unreliable synthetic subchannels and are used in contemporary cellular control channels.

Coding gain is the reduction in required E_b/N₀ at a specified error rate relative to a reference uncoded system. It depends on code, block length, decoder and target error probability.

4.1 Interleaving

A fading event may corrupt a burst of adjacent coded bits. An interleaver permutes coded-bit positions before transmission; the receiver applies the inverse permutation before decoding. The burst can then appear as separated errors that the code is better able to correct.

Interleaving does not create redundancy and does not itself correct errors. Deep interleaving also increases latency and memory requirements. Its depth should exceed typical fade duration while remaining compatible with delay constraints.

4.2 Hybrid automatic repeat request

Hybrid ARQ combines forward-error correction with retransmission. In chase combining, repeated coded packets are combined to improve SNR. In incremental redundancy, retransmissions supply additional parity bits. Hybrid ARQ improves reliability but can produce variable latency.

5. Spread-Spectrum Techniques

5.1 Direct-sequence spread spectrum

In direct-sequence spread spectrum (DSSS), each data symbol is multiplied by a faster pseudorandom chip sequence. Let R_chip denote chip rate and R_d denote data-symbol rate. For one spreading code of N_c chips per data symbol,

where G_p is processing gain as a linear ratio under ideal synchronization and rectangular-chip assumptions. In decibels, G_p,dB=10log₁₀G_p. The symbol R_chip is deliberately distinct from the coding rate R_code.

DSSS can suppress narrowband interference after despreading and permits code-domain multiple access. It does not create energy or inherently improve performance against thermal noise at fixed E_b/N₀. Performance depends on code correlation, synchronization and the near–far problem.

5.2 Frequency hopping

Frequency-hopping spread spectrum changes carrier frequency according to a hopping sequence. It offers resilience to persistent narrowband interference and can reduce prolonged collisions. Slow hopping uses multiple symbols per hop; fast hopping may use multiple hops per symbol. Synthesizer speed, hop synchronization and regulation constrain implementation.

Worked Example: Processing Gain

A DSSS system transmits 2 Mchip/s and uses 20 ksymbol/s data symbols. Its processing gain is

and

This is an ideal processing-gain figure, not a guarantee of 20 dB end-to-end performance improvement.

6. Orthogonal Frequency-Division Multiplexing

OFDM divides a frequency-selective broadband channel into many narrowband subchannels. Each subcarrier can often be equalized with one complex coefficient.

For N subcarriers, the normalized inverse discrete Fourier transform is

Here, X[k] is the complex symbol on subcarrier k, x[m] is time-domain sample m, and the 1/√N convention preserves energy between domains. Other implementations may place scaling entirely in the inverse or forward transform, but the summation always includes all indices from 0 through N−1.

Subcarriers are orthogonal when their spacing is

where T_u is the useful OFDM symbol duration. Although subcarrier spectra overlap, each other subcarrier has a spectral zero at the desired subcarrier’s sampling point under perfect timing and frequency synchronization.

6.1 Cyclic prefix

A cyclic prefix copies the final samples of an OFDM symbol to its beginning. If a sampled channel has L significant taps, the condition

prevents intersymbol interference from the preceding symbol in the ideal synchronized model and converts linear channel convolution into circular convolution over the useful block. Equivalently, cyclic-prefix duration should be at least the channel’s maximum significant excess delay, with practical design margin.

The total symbol duration and ideal time efficiency are

A longer prefix tolerates more delay spread but reduces throughput.

Worked Example: OFDM Timing

An OFDM system has Δf=15 kHz and cyclic-prefix duration T_CP=4.69 μs. Then

The cyclic prefix therefore consumes about 6.57% of the total symbol time.

6.2 OFDM limitations

OFDM is sensitive to carrier-frequency offset, phase noise and Doppler because they destroy orthogonality and create intercarrier interference. It also has a high peak-to-average power ratio (PAPR), requiring transmitter power-amplifier back-off. Clipping, tone reservation and selected mapping can reduce PAPR, but each introduces distortion, overhead or complexity.

7. Equalization and Channel Estimation

Equalization compensates for channel amplitude and phase distortion. It requires channel knowledge obtained from pilots, training sequences, decision-directed tracking or blind estimation.

For one OFDM subcarrier,

A zero-forcing estimate is

provided H[k]≠0. Zero forcing removes channel distortion but can greatly amplify noise near a spectral null.

If transmitted-symbol power is P_X and noise variance is σ_N², the scalar linear minimum mean-square error equalizer is

with X̂[k]=G_MMSE[k]Y[k]. It balances residual distortion against noise enhancement and approaches zero forcing at high SNR when the channel is not near zero.

In single-carrier systems, a time-domain finite-impulse-response equalizer can suppress ISI. A decision-feedback equalizer cancels postcursor interference using previous decisions, but an incorrect decision can propagate.

8. Diversity

Diversity sends or receives replicas that fade as independently as possible. Separation may be in time, frequency, polarization or space.

8.1 Selection and maximal-ratio combining

Selection combining chooses the branch with the largest instantaneous SNR. Maximal-ratio combining (MRC) phase-aligns and weights all branches according to their channel quality. For independently noisy branches with correctly known channels, its output SNR is

where γ_i is branch SNR and N_r is the number of receive branches.

If three branches have SNRs 2, 4 and 5 as linear ratios, MRC produces γ_MRC=11, or 10log₁₀(11)=10.41 dB. One must add linear SNRs, not decibel values.

8.2 Transmit diversity

The two-transmit-antenna Alamouti space–time block code provides diversity while enabling simple linear detection. Its orthogonality assumes that the channel remains approximately constant over two symbol intervals. Rapid channel variation, inaccurate synchronization and antenna correlation reduce its benefit.

9. MIMO and Beamforming

9.1 MIMO signal model

For a narrowband flat-fading MIMO channel,

where x is an N_t×1 transmit vector, y is an N_r×1 receive vector, H is the N_r×N_t channel matrix, and n is complex Gaussian noise. A wideband channel is commonly transformed into parallel narrowband MIMO models using OFDM.

MIMO can support:

• spatial diversity, improving reliability;
• spatial multiplexing, carrying independent data streams;
• beamforming, concentrating energy in selected spatial directions; and
• multiuser MIMO, serving multiple terminals on shared resources.

The number of independent spatial streams cannot exceed rank(H)≤min(N_t,N_r). Strong antenna correlation or insufficient scattering can reduce rank.

9.2 MIMO spectral efficiency and capacity

Assume a flat, constant channel over bandwidth B, perfect channel knowledge at the receiver, Gaussian signalling, no channel knowledge at the transmitter, equal power across N_t antennas and spatially white receiver noise. Let

where P is total transmitted signal power in the normalized channel model and σ_n² is noise power per receive dimension over bandwidth B. Define ρ=P/σ_n². The channel coefficients in H include the dimensionless link gains consistent with this normalization.

The instantaneous spectral efficiency is

and the corresponding capacity is

The superscript H on the matrix denotes Hermitian transpose. If the transmitter knows the channel, equal power is not always optimal; water-filling across channel eigenmodes can increase capacity subject to a total-power constraint.

Worked Example: MIMO Capacity

Consider

Then

so

For B=10 MHz,

This is an information-theoretic result under the assumptions stated above, not net application throughput.

9.3 Beamforming

Beamforming controls antenna amplitudes and phases to produce constructive interference in desired directions and suppression elsewhere. If a(θ₀) is the steering vector toward direction θ₀, a normalized matched transmit weight is

under a compatible channel and steering-vector convention. In an ideal coherent array, received SNR can scale approximately with the number of transmit elements, but mutual coupling, calibration error, blockage and regulatory power limits reduce practical gain.

Analog beamforming uses one or a few radio-frequency chains and phase shifters. Digital beamforming provides independent complex weights and supports multiple streams but consumes more power and hardware. Hybrid beamforming combines both, particularly at millimetre-wave frequencies.

10. Integrated Illustrative Case: Warehouse Wireless Link

Background

This fictional case considers an automated warehouse using a 3.5 GHz private wireless network. Mobile robots require high-throughput video and control connectivity with bounded packet error and latency.

Problem

Measurements indicate a maximum significant excess delay of 1.2 μs. Robot speed can reach 15 m/s. The designer has a 100 MHz channel and wishes to use 64-QAM, rate-3/4 LDPC coding and two spatial streams.

Analysis

The system selects a subcarrier spacing of 60 kHz. Therefore,

A cyclic prefix of 2.083 μs exceeds the measured 1.2 μs delay and leaves implementation margin. Total symbol duration is

Suppose 1,200 subcarriers carry data. Each 64-QAM symbol carries six coded bits. Information bits per OFDM symbol per spatial layer are

Thus, the ideal physical-layer information rate per layer is

Two spatial layers provide an ideal total of 576 Mbit/s before pilot, control, guard-band, scheduling and retransmission overhead.

Maximum Doppler is

Using the stated approximation,

This spans about 129 OFDM symbols, so pilot spacing must track millisecond-scale variation. Because metal shelves create deep fades and correlation, dual-polarized antenna placement and measured MIMO rank are important.

Approach and Solution

The design uses adaptive modulation and coding: 64-QAM is retained under strong channel quality, while QPSK or 16-QAM with stronger coding is selected during blockage. Frequency-domain MMSE equalization avoids severe noise amplification at weak subcarriers. LDPC coding and limited-depth interleaving address residual burst errors without violating control latency. Two-layer spatial multiplexing is used only when the estimated channel supports adequate rank; otherwise, the antennas provide diversity or beamforming.

Outcome

Link calculations show that the proposed numerology can tolerate the measured delay spread and mobility. The 576 Mbit/s value is treated as a gross ideal rate rather than guaranteed user throughput. Final acceptance requires channel sounding, error-vector-magnitude tests, packet-error measurements and interference assessment.

Learning Points

• Cyclic-prefix duration should follow measured channel delay, not an arbitrary standard value.
• High-order modulation should be adapted to instantaneous channel quality.
• Two antennas do not guarantee two useful streams; channel rank matters.
• Coding, interleaving, equalization and diversity solve different reliability problems.
• Gross physical-layer rate must not be presented as application throughput.

11. Design Challenges and Contemporary Practice

Physical-layer design must account for nonlinear power amplifiers, oscillator phase noise, quantized converters, antenna correlation, imperfect channel estimation and computational delay. Energy-limited devices may prefer low-order modulation and simpler receivers even when greater spectral efficiency is theoretically possible.

Adaptive modulation and coding uses channel-quality reports to select modulation order and code rate. Poor or delayed reports can cause unstable adaptation. Massive MIMO uses large arrays for spatial multiplexing and beamforming, but requires calibration, pilot design and scalable signal processing. Cell-free architectures coordinate distributed access points to improve coverage consistency.

Machine-learning methods are being investigated for channel estimation, beam selection and interference classification. Their use should be justified against conventional estimators, tested under distribution shift and constrained by latency, explainability and safety requirements.

Spectrum regulation specifies permitted bands, emissions, channel access and effective isotropic radiated power. Designers must satisfy spectral masks, exposure limits and coexistence requirements. Beamforming can reduce unintended radiation but also creates privacy and security concerns: adversaries may estimate location from beam-training exchanges or manipulate channel-state feedback. Secure control signalling, authenticated feedback and interference monitoring are therefore part of responsible physical-layer deployment.

Chapter Summary

Wireless reliability depends on matching physical-layer techniques to channel delay, Doppler, noise, interference and spatial structure. Modulation determines bits per symbol, while coding introduces redundancy and interleaving disperses burst errors. Spread spectrum distributes signal energy across bandwidth and can resist narrowband interference. OFDM converts a frequency-selective channel into parallel narrowband subchannels, provided its cyclic prefix and synchronization are adequate.

Equalizers compensate for channel distortion, with MMSE processing offering a practical balance between inversion and noise enhancement. Diversity reduces fading risk by combining independently faded observations. MIMO can provide diversity, spatial multiplexing or beamforming, but achievable gains depend on channel rank, channel knowledge and antenna correlation. Capacity formulas must distinguish spectral efficiency in bit/s/Hz from capacity in bit/s and must state power, noise and bandwidth normalization explicitly. Effective link design integrates these techniques while accounting for latency, hardware limitations, regulation and measured propagation conditions.

Key Terms

Adaptive modulation and coding: Selection of modulation order and code rate according to channel quality.

Beamforming: Coherent antenna weighting used to shape spatial radiation or reception.

Channel coding: Addition of structured redundancy for error detection or correction.

Chip rate: Number of spreading-code chips transmitted per second.

Code rate: Ratio of information bits to coded bits.

Coherence bandwidth: Frequency range over which channel response is strongly correlated.

Coherence time: Approximate duration over which a channel remains strongly correlated.

Cyclic prefix: Repeated OFDM suffix placed before a symbol to control multipath interference.

Delay spread: Distribution or extent of multipath arrival delays.

Diversity: Use of independently fading signal replicas to improve reliability.

Doppler shift: Frequency displacement caused by relative motion.

Equalization: Compensation for channel-induced amplitude, phase or temporal distortion.

Interleaving: Reordering of coded symbols to disperse burst errors.

MIMO: Communication using multiple transmit and receive antennas.

OFDM: Multicarrier modulation using mutually orthogonal subcarriers.

Processing gain: Ratio of spread bandwidth or chip rate to underlying data-symbol rate under stated assumptions.

Spatial multiplexing: Transmission of independent streams through spatial channel dimensions.

Spectral efficiency: Information rate per unit bandwidth, measured in bit/s/Hz.

Review Questions

Short Answer Questions

1. Why must a mobile multipath channel generally be represented by h(t,τ) rather than only h(τ)?
2. What is the difference between coding and interleaving?
3. State the cyclic-prefix condition for an OFDM channel with L significant taps.
4. Why can zero-forcing equalization perform poorly near a channel null?
5. Distinguish spatial diversity, spatial multiplexing and beamforming.

Long Answer Questions

1. Explain how delay spread and Doppler spread influence waveform, pilot and equalizer design.
2. Compare BPSK, QPSK and higher-order QAM in terms of power efficiency, spectral efficiency and hardware sensitivity.
3. Derive the OFDM orthogonality condition and explain how frequency offset destroys it.
4. Discuss channel coding, interleaving and hybrid ARQ as complementary reliability mechanisms.
5. Explain the assumptions behind the equal-power MIMO capacity expression and how transmitter channel knowledge changes power allocation.

Analytical / Application Questions

1. An OFDM link has subcarrier spacing 30 kHz and a cyclic prefix of 2.5 μs. Calculate useful-symbol duration, total-symbol duration and cyclic-prefix efficiency. Assess whether the prefix is sufficient for a maximum significant delay of 3 μs.
2. A 64-QAM system operates at 5 Msymbol/s with coding rate 2/3. Calculate its ideal information rate before overhead. Explain how the result changes if the link falls back to QPSK with rate-1/2 coding.
3. Design a physical-layer strategy for a high-mobility factory link subject to deep fading, impulsive interference and strict latency. Justify modulation, coding, interleaving, equalization, diversity and antenna choices.

Multiple Choice Questions with Answers

Question 1

Which expression correctly models a linear time-varying multipath channel?

A. y(t)=h(t)x(t)+n(t)

B. y(t)=∫h(t,τ)x(t−τ)dτ+n(t)

C. y(t)=x(t−h(t))+n(t)

D. y(t)=∫h(τ)x(t)dτ

Correct Answer: B. y(t)=∫h(t,τ)x(t−τ)dτ+n(t)

Question 2

A rate-3/4 code maps 600 information bits into how many coded bits?

A. 450 bits

B. 600 bits

C. 800 bits

D. 1,200 bits

Correct Answer: C. 800 bits

Question 3

A DSSS link has a chip rate of 4 Mchip/s and a data-symbol rate of 100 ksymbol/s. What is its ideal processing gain?

A. 16

B. 20

C. 40

D. 400

Correct Answer: C. 40

Question 4

For an OFDM subcarrier spacing of 25 kHz, what is the useful symbol duration?

A. 4 μs

B. 25 μs

C. 40 μs

D. 250 μs

Correct Answer: C. 40 μs

Question 5

A sampled OFDM channel has eight significant taps. What is the minimum cyclic-prefix length under the ideal discrete-time model?

A. 4 samples

B. 7 samples

C. 8 samples

D. 16 samples

Correct Answer: B. 7 samples

Question 6

Why is an MMSE equalizer often preferred to a zero-forcing equalizer at low SNR?

A. It eliminates the need for channel estimation.

B. It guarantees zero intersymbol interference in every channel.

C. It balances channel inversion against noise enhancement.

D. It increases the channel coherence bandwidth.

Correct Answer: C. It balances channel inversion against noise enhancement.

Question 7

Three independent receive branches have linear SNRs of 1, 3 and 6. What is the ideal maximal-ratio-combining output SNR?

A. 6

B. 9

C. 10

D. 18

Correct Answer: C. 10

Question 8

Under the equal-power MIMO model in the chapter, log₂ det(I+(ρ/N_t)HH^H) has which units?

A. bit/s

B. bit/s/Hz

C. joule/bit

D. watt/hertz

Correct Answer: B. bit/s/Hz

Question 9

Which condition most directly limits the number of independent MIMO spatial streams?

A. The code rate alone

B. The cyclic-prefix overhead alone

C. The rank of the channel matrix

D. The spreading-code length

Correct Answer: C. The rank of the channel matrix

Question 10

A 16-QAM system transmits 3 Msymbol/s using a rate-1/2 code. Ignoring overhead, what is its information rate?

A. 3 Mbit/s

B. 6 Mbit/s

C. 8 Mbit/s

D. 12 Mbit/s

Correct Answer: B. 6 Mbit/s