DIGITAL COMMUNICATIONS

Chapter 6
Digital Passband Modulation

Post Graduate Program in Electronics and Communication Engineering
DIGITAL COMMUNICATIONS   •   Post Graduate

Digital Passband Modulation

Introduction and context

Digital passband modulation converts a sequence of symbols into a waveform whose energy is concentrated around a carrier frequency. It is required whenever the communication medium is bandpass, as in wireless links, satellite systems, microwave radio, optical intensity systems, and many cable networks. In contrast, baseband transmission sends pulses whose spectrum is concentrated near zero frequency.

A digital passband transmitter performs three principal operations: symbol mapping, pulse shaping, and carrier modulation. The receiver performs matched filtering or correlation, timing recovery, carrier or phase recovery, symbol detection, and decoding. The central engineering objective is to transmit the largest reliable information rate within available bandwidth and power while controlling complexity and spectral interference.

At postgraduate level, passband modulation should be understood through signal-space geometry, probability of error, bandwidth efficiency, synchronization requirements, and implementation constraints—not merely through waveform sketches.

Learning objectives

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

  1. Explain the passband modulation model and distinguish symbols, waveforms, and carriers.
  2. Derive the relationship between complex envelopes and passband signals.
  3. Analyze ASK, PSK, FSK, QAM, DPSK, and continuous-phase modulation.
  4. Relate constellation geometry to Euclidean distance and error probability.
  5. Calculate bandwidth, energy per bit, and approximate bit-error probability under AWGN.
  6. Explain pulse shaping, Nyquist criteria, synchronization, and receiver architecture.
  7. Select a modulation scheme for a practical communication requirement.
  8. Evaluate implementation limitations and current developments.

1. Passband signal model

Let the transmitted symbol sequence be ak, where ak belongs to an alphabet of size M. A general linear passband signal may be written as

s(t) = Re{u(t)ej2πfct}

where u(t) is the complex baseband equivalent or complex envelope, fc is the carrier frequency in hertz, j = √(-1), and Re{·} denotes the real part. If u(t) = I(t) + jQ(t), then

s(t) = I(t)cos(2πfct) − Q(t)sin(2πfct)

I(t) and Q(t) are the in-phase and quadrature components, measured in volts when the signal is a voltage waveform. The two carriers are orthogonal over an appropriate observation interval, allowing two independent dimensions to be transmitted at the same carrier frequency.

A pulse-shaped complex envelope is commonly expressed as

u(t) = ∑k ak p(t − kTs)

where Ts is the symbol duration in seconds and p(t) is the unit-energy or appropriately normalized shaping pulse. The symbol rate is [[MATH{R_s = 1/T_s}]] symbols per second (baud). For an uncoded system carrying [[MATH{log_2 M}]] bits per symbol, the bit rate is

Rb = Rs log2 M

This formula assumes equally likely symbols and no coding overhead. With a code rate [[MATH{R_c}]], the net information rate is approximately [[MATH{R_b,net = R_c R_s log_2 M}]].

Energy definitions

The energy per symbol is [[MATH{E_s}]], and the energy per bit is [[MATH{E_b}]]. For an uncoded [[MATH{M}]]-ary modulation,

Es = Eb log2 M

The equality applies when the average symbol energy is used and each symbol represents [[MATH{log_2 M}]] information bits. In a channel with one-sided noise power spectral density [[MATH{N_0}]] in watts per hertz, performance is often expressed using [[MATH{E_b/N_0}]].

2. Signal-space representation

A passband modulation scheme can be represented by a constellation: a set of points corresponding to possible signal vectors. For two-dimensional modulation, a symbol is represented by

si(t) = Ii φ1(t) + Qi φ2(t)

where [[MATH{φ_1(t)}]] and [[MATH{φ_2(t)}]] are orthonormal basis functions and [[MATH{I_i,Q_i}]] are constellation coordinates. The squared Euclidean distance between symbols [[MATH{i}]] and [[MATH{l}]] is

dil2 = (Ii − Il)2 + (Qi − Ql)2

Under additive white Gaussian noise (AWGN), larger minimum distance generally produces lower error probability for a fixed average energy. Define

dmin = mini ≠ l dil

The nearest-neighbor approximation is especially useful for large constellations, although exact error probability depends on all decision-region boundaries and symbol probabilities.

Gray labeling assigns adjacent constellation points bit labels differing in one bit. It does not change symbol-error probability, but it often minimizes bit errors when symbol errors predominantly occur between neighboring points.

3. Major digital passband modulation schemes

3.1 Amplitude-shift keying

In binary amplitude-shift keying (BASK), the carrier amplitude represents the binary symbol. On-off keying (OOK) is the special case in which one symbol has zero carrier amplitude:

s1(t) = A p(t)cos(2πfct), s0(t) = 0

ASK has simple transmitters but is sensitive to amplitude fading, interference, and nonlinear amplification. OOK is widely used in low-cost optical and infrared links, remote controls, and low-power devices. Coherent detection requires carrier phase information; noncoherent energy detection is simpler but generally has inferior performance.

3.2 Phase-shift keying

In M-PSK, information is carried by carrier phase:

sm(t) = A p(t)cos(2πfct + θm), θm = 2πm/M

where [[MATH{m = 0,1,…,M−1}]]. BPSK uses phases separated by 180° and has two antipodal constellation points. QPSK uses four phases and carries two bits per symbol. With suitable pulse shaping, QPSK has the same theoretical bit-error probability as BPSK at the same [[MATH{E_b/N_0}]], while doubling the bit rate at the same symbol rate.

For coherent BPSK in AWGN,

Pb = Q(√(2Eb/N0))

where [[MATH{Q(x) = (1/√(2π))∫_x^∞ e^(−z^2/2) dz}]]. This result assumes ideal synchronization, equally likely bits, coherent detection, and AWGN.

For coherent Gray-coded QPSK, the bit-error probability is also

Pb = Q(√(2Eb/N0))

although its symbol-error probability is different because two bits are decided per symbol.

Offset QPSK delays the Q-channel transitions by one-half symbol relative to the I channel. Consequently, the phase cannot jump directly between opposite constellation points, reducing envelope fluctuations and easing operation with nonlinear power amplifiers.

3.3 Differential PSK

Differential PSK encodes information in phase changes rather than absolute phase. In DBPSK, a binary one may cause a phase reversal and a binary zero may cause no reversal, according to the chosen convention. Differential detection avoids an explicit carrier-phase reference but introduces a performance penalty. For ideal DBPSK in AWGN, a common result is

Pb = 1/2 exp(−Eb/N0)

under the conventional noncoherent differential detector model. The exact result depends on the differential encoding and detector assumptions.

3.4 Frequency-shift keying

In FSK, symbols select different frequencies. Binary FSK can be expressed as

si(t) = A p(t)cos(2πfi t + θi), i ∈ {0,1}

Coherent orthogonal BFSK has

Pb = Q(√(Eb/N0))

when the two signals are orthogonal and detected coherently in AWGN. Compared with coherent BPSK, orthogonal BFSK requires approximately 3 dB more [[MATH{E_b/N_0}]] for the same high-SNR error probability. Noncoherent FSK removes phase recovery but generally requires still more energy.

In continuous-phase FSK, phase is constrained to remain continuous at symbol boundaries. Minimum-shift keying (MSK) is binary CPFSK with modulation index [[MATH{h = 0.5}]]. Gaussian MSK (GMSK) filters the data before frequency modulation and is spectrally compact, although filtering introduces controlled intersymbol interference that the receiver must handle.

3.5 Quadrature amplitude modulation

QAM simultaneously changes amplitude and phase. A square 16-QAM constellation carries four bits per symbol, while 64-QAM carries six. A signal can be represented as

s(t) = [Ik p(t)cos(2πfct) − Qk p(t)sin(2πfct)]

QAM provides high spectral efficiency but requires accurate amplitude and phase recovery. It is sensitive to additive noise, phase noise, nonlinear distortion, and amplitude imbalance. Higher-order QAM therefore requires a larger [[MATH{E_b/N_0}]] or stronger coding for a target error rate.

For square M-QAM with Gray coding in AWGN, a widely used approximate bit-error expression is

Pb ≈ [4/log2 M] (1 − 1/√M) Q(√([3log2 M/(M−1)] Eb/N0))

This approximation assumes square constellations, equally likely symbols, Gray labeling, coherent detection, and sufficiently small error probability. It should not be treated as exact at low SNR.

4. Pulse shaping and bandwidth

If rectangular pulses are used, the spectrum has slowly decaying sidelobes and substantial adjacent-channel interference. Practical systems use raised-cosine or root-raised-cosine filtering. The raised-cosine roll-off factor is [[MATH{α}]], where [[MATH{0 ≤ α ≤ 1}]]. For ideal raised-cosine signaling, the null-to-null baseband bandwidth is

B = (1 + α)Rs/2

for a real low-pass equivalent spectrum. The corresponding passband occupied bandwidth is approximately

BRF = (1 + α)Rs

The precise occupied bandwidth also depends on filtering definitions and spectral masks. A root-raised-cosine filter is commonly split between transmitter and receiver; the cascade produces a raised-cosine response satisfying the zero-ISI condition at ideal sampling instants.

The Nyquist zero-ISI criterion requires the overall pulse [[MATH{g(t)}]] to satisfy

g(0) = 1, g(nTs) = 0 for every nonzero integer n

This condition eliminates intersymbol interference at sampling times in a noiseless, perfectly synchronized channel. It does not eliminate noise, timing error, multipath distortion, or implementation imperfections.

5. Detection and receiver architecture

A coherent receiver estimates carrier phase and frequency, filters the received waveform, samples at symbol timing, and selects the nearest constellation point. The received signal is modeled as

r(t) = s(t) + n(t)

where [[MATH{n(t)}]] is AWGN with zero mean and two-sided power spectral density [[MATH{N_0/2}]]. A matched filter maximizes the output signal-to-noise ratio at the sampling instant when the noise is white and the expected signal pulse is known.

A practical receiver includes:

  • a radio-frequency front end and down-converter;
  • automatic gain control;
  • carrier-frequency and carrier-phase recovery;
  • matched or root-raised-cosine filtering;
  • timing recovery, often using a timing-error detector and a digital loop;
  • equalization when the channel is dispersive;
  • symbol decision and channel decoding.

Carrier recovery may use a phase-locked loop, Costas loop, pilot symbols, or blind algorithms. Decision-directed loops operate using detected symbols and may suffer cycle slips or error propagation. Timing recovery is equally important: a phase-accurate receiver sampling at the wrong instant can perform poorly even in a high-SNR channel.

6. Performance measures and design trade-offs

The principal measures are bit-error rate (BER), symbol-error rate (SER), spectral efficiency, power efficiency, peak-to-average power ratio (PAPR), implementation complexity, and robustness to channel impairments.

Spectral efficiency is

η = Rb/BRF

in bits per second per hertz. Ignoring coding overhead and using the approximate passband bandwidth above,

η ≈ log2 M/(1 + α)

for single-carrier M-ary modulation. Increasing [[MATH{M}]] improves spectral efficiency but reduces minimum constellation distance for a fixed average energy. This creates the fundamental practical trade-off between bandwidth efficiency and power efficiency.

PAPR is important in power-amplifier design:

PAPR = maxt |s(t)|2 / E{|s(t)|2}

where the numerator is the maximum instantaneous power and the denominator is average power. QAM and multicarrier waveforms can have substantial envelope variation. A nonlinear amplifier causes constellation compression, spectral regrowth, and adjacent-channel interference. Constant-envelope schemes such as MSK tolerate nonlinear amplification better.

Worked example: selecting a modulation order

A link must transmit 24 Mbit/s using a raised-cosine passband channel with roll-off factor [[MATH{α = 0.25}]] and occupied bandwidth no greater than 6 MHz. Determine the minimum square-QAM order, ignoring coding overhead.

The required spectral efficiency is

ηreq = 24/6 = 4 bit/s/Hz

For single-carrier QAM,

η ≈ log2 M/(1 + α)

Therefore,

log2 M ≥ 4(1.25) = 5

The smallest square-QAM constellation with at least five bits per symbol is 64-QAM, for which [[MATH{log_2 64 = 6}]]. Its approximate efficiency is [[MATH{6/1.25 = 4.8}]] bit/s/Hz. The corresponding symbol rate is

Rs = 24/6 = 4 Mbaud

and the estimated occupied bandwidth is

BRF = 1.25 × 4 = 5 MHz

Thus 64-QAM meets the bandwidth requirement, but the design must verify that the available [[MATH{E_b/N_0}]], amplifier linearity, phase-noise performance, and coding scheme can support the target BER.

7. Channel impairments and mitigation

AWGN is a useful reference model but is not sufficient for most wireless links. Multipath propagation produces frequency-selective fading and intersymbol interference. Doppler shift causes time variation and frequency offset. Shadowing changes average received power, while oscillator phase noise rotates the constellation. In-phase/quadrature imbalance, carrier leakage, amplifier nonlinearity, quantization, and sampling-clock offset create additional impairments.

Mitigation methods include adaptive equalization, decision-feedback equalizers, orthogonal frequency-division multiplexing, channel coding, interleaving, pilot-assisted estimation, automatic frequency control, phase tracking, and diversity. OFDM converts a frequency-selective channel into many approximately flat subchannels, but its high PAPR and sensitivity to frequency offset require careful engineering.

Link adaptation selects modulation and coding according to channel quality. A robust mode such as QPSK with strong coding is used at low SNR; higher-order QAM with weaker coding is selected when the channel supports it. The decision should be based on target block-error rate, latency, retransmission policy, mobility, and power constraints—not solely on instantaneous SNR.

8. Illustrative application case: adaptive wireless backhaul

Background

Consider a microwave backhaul link connecting a cellular base station to the core network. The licensed channel has fixed bandwidth, but received SNR varies because of rainfall, multipath fading, and antenna alignment.

Problem

A fixed 256-QAM mode provides high throughput during clear conditions but experiences unacceptable errors during fading. A permanently robust QPSK mode preserves availability but wastes capacity.

Analysis

The link controller estimates SNR and monitors decoder block errors. Each modulation and coding mode has a required SNR margin. Higher-order QAM has closer constellation points and is more vulnerable to phase noise and nonlinear distortion. The power amplifier must also operate with sufficient back-off to prevent spectral regrowth.

Approach and solution

The system uses adaptive modulation and coding. It selects QPSK, 16-QAM, 64-QAM, or 256-QAM according to filtered channel estimates, with hysteresis to prevent rapid mode switching. Pilot symbols support channel and phase estimation. Forward-error correction and interleaving address random and burst errors, while a link-alarm threshold forces a robust mode before service becomes unavailable.

Outcome

Clear-channel periods use 256-QAM and deliver high throughput. During fading, the link automatically reduces modulation order and increases coding protection. Availability improves compared with fixed high-order QAM, while average throughput remains greater than that of a permanently conservative mode.

Learning points

The case illustrates that modulation selection is a systems problem. Constellation size, coding, amplifier efficiency, channel estimation, latency, and availability must be optimized jointly. A theoretically efficient modulation can be operationally inferior if synchronization and linearity requirements cannot be met.

9. Contemporary developments and research directions

Modern systems combine passband modulation with powerful forward-error correction, iterative detection, massive multiple-input multiple-output processing, beamforming, and machine-assisted link adaptation. Probabilistic constellation shaping changes the probabilities of constellation points so that the transmitted distribution approaches a power-efficient Gaussian-like distribution while retaining a finite QAM structure.

Machine learning is being investigated for modulation classification, equalization, channel estimation, and end-to-end waveform optimization. However, data-driven methods must be evaluated for generalization, computational cost, explainability, robustness to distribution shift, and compliance with spectral regulations. They supplement rather than eliminate the need for signal-processing principles.

Research also examines low-power constant-envelope or low-PAPR waveforms, joint communication and sensing, optical wireless modulation, and modulation formats designed for high-mobility and high-frequency channels. Regulatory constraints remain essential: occupied bandwidth, out-of-band emissions, maximum power, coexistence, and electromagnetic compatibility are engineering requirements as well as legal obligations.

10. Best-practice design procedure

A defensible passband-modulation design should proceed as follows:

  1. Specify information rate, latency, reliability, coverage, and available bandwidth.
  2. Characterize the channel, including fading, delay spread, Doppler, interference, and oscillator impairments.
  3. Select candidate constellations using spectral and power efficiency requirements.
  4. Choose pulse-shaping filters and verify the spectral mask and timing response.
  5. Analyze BER or BLER using appropriate channel and receiver models.
  6. Include synchronization, equalization, coding, interleaving, and amplifier nonlinearity in simulations.
  7. Validate with measured hardware impairments and worst-case operating conditions.
  8. Define adaptation rules, alarms, fallback modes, and regulatory limits.

Challenges and limitations

Analytical BER formulas frequently assume ideal synchronization, AWGN, independent symbols, linear amplification, and perfect channel knowledge. Real systems violate several of these assumptions simultaneously. Simulation results can also be misleading if the channel model, random seeds, oversampling, or stopping criteria are unsuitable. Laboratory performance does not guarantee field performance under interference and mobility.

A second limitation is that bandwidth efficiency cannot be increased without cost. Higher-order modulation requires improved SNR, linearity, phase accuracy, and channel estimation. Similarly, aggressive pulse shaping reduces bandwidth but may increase sensitivity to timing error and filter implementation. Engineering decisions therefore require multi-objective optimization rather than maximization of a single metric.

Chapter Summary

Digital passband modulation maps information symbols onto amplitude, phase, frequency, or combinations of these properties of a carrier. The complex-envelope model separates modulation from the carrier and provides a convenient basis for analysis. Signal-space geometry explains why minimum Euclidean distance, constellation labeling, and average energy determine error performance.

ASK is simple but amplitude-sensitive; PSK is power-efficient; FSK and continuous-phase schemes offer robustness and constant-envelope operation; QAM provides high spectral efficiency at the cost of greater SNR, synchronization, and linearity requirements. Pulse shaping controls bandwidth and intersymbol interference, while matched filtering and accurate timing maximize detection quality.

Practical design must consider BER, spectral efficiency, PAPR, fading, phase noise, nonlinear amplification, coding, synchronization, regulation, and adaptation. Contemporary systems use adaptive modulation, advanced coding, MIMO, probabilistic shaping, and increasingly machine-assisted signal processing.

Key Terms

Passband modulation: Transmission of digital information using a carrier-centered waveform.

Complex envelope: Complex baseband representation of a passband signal.

Symbol: A waveform or constellation point representing one or more bits.

Symbol rate: Number of transmitted symbols per second, measured in baud.

Constellation: Geometric representation of possible signal vectors.

ASK: Amplitude-shift keying, in which amplitude carries information.

PSK: Phase-shift keying, in which phase carries information.

FSK: Frequency-shift keying, in which frequency carries information.

QAM: Quadrature amplitude modulation using independent in-phase and quadrature amplitudes.

BPSK: Binary phase-shift keying with two antipodal phases.

QPSK: Quadrature phase-shift keying carrying two bits per symbol.

GMSK: Gaussian minimum-shift keying.

Raised-cosine filter: Pulse-shaping filter satisfying the Nyquist zero-ISI criterion.

Roll-off factor: Parameter controlling excess bandwidth of a raised-cosine response.

Intersymbol interference: Distortion caused by overlap of neighboring symbol responses.

Matched filter: Filter maximizing sampled SNR for a known signal in white noise.

Gray coding: Labeling in which adjacent constellation points differ by one bit.

Spectral efficiency: Information rate per unit bandwidth.

PAPR: Ratio of peak instantaneous power to average power.

Link adaptation: Dynamic selection of modulation and coding based on channel conditions.

Review Questions

Short Answer Questions

  1. What is the difference between symbol rate and bit rate?
  2. Why are in-phase and quadrature carriers useful in passband modulation?
  3. State the purpose of Gray labeling in a constellation.
  4. What does the roll-off factor of a raised-cosine filter control?
  5. Why is high-order QAM more sensitive to noise than QPSK?

Long Answer Questions

  1. Derive the complex-envelope representation of a passband signal and explain the roles of its in-phase and quadrature components.
  2. Compare BPSK, QPSK, BFSK, and 16-QAM in terms of constellation geometry, bandwidth, power efficiency, and implementation requirements.
  3. Explain the Nyquist zero-ISI criterion and the practical use of root-raised-cosine filtering.
  4. Describe the architecture and synchronization requirements of a coherent digital passband receiver.
  5. Discuss the effects of fading, phase noise, amplifier nonlinearity, and frequency offset on QAM performance.

Analytical / Application Questions

  1. A system must transmit 12 Mbit/s in a 4 MHz passband channel with roll-off factor 0.25. Determine the minimum ideal value of [[MATH{log_2 M}]] and identify a practical square-QAM order.
  2. Explain how you would design an adaptive modulation and coding policy for a mobile link whose SNR varies rapidly.
  3. A nonlinear power amplifier is available with limited back-off. Evaluate whether QAM, OQPSK, or MSK is most appropriate, considering spectral efficiency, envelope variation, and receiver complexity.

Multiple Choice Questions with Answers

Question 1

Which quantity primarily determines the number of information bits carried by one uncoded symbol?

A. Carrier frequency

B. log₂ M

C. Roll-off factor

D. Noise bandwidth

Correct Answer: B. log₂ M

Question 2

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

A. Q(√(2E_b/N_0))

B. Q(√(E_b/N_0))

C. 1/2 exp(−E_b/N_0)

D. Q(E_b/N_0)

Correct Answer: A. Q(√(2E_b/N_0))

Question 3

What is the principal advantage of Gray labeling?

A. It increases carrier frequency

B. It eliminates all symbol errors

C. It often minimizes bit errors for nearest-neighbor symbol errors

D. It removes the need for timing recovery

Correct Answer: C. It often minimizes bit errors for nearest-neighbor symbol errors

Question 4

For a raised-cosine passband signal, the approximate occupied bandwidth is

A. R_s/(1+α)

B. (1+α)R_s

C. α/R_s

D. 2R_s/α

Correct Answer: B. (1+α)R_s

Question 5

Which modulation generally has the highest spectral efficiency among the following at comparable symbol rate?

A. BPSK

B. QPSK

C. BFSK

D. 256-QAM

Correct Answer: D. 256-QAM

Question 6

Why is MSK relatively tolerant of nonlinear amplification?

A. It has no carrier

B. It is a constant-envelope modulation

C. It uses infinitely many amplitudes

D. It eliminates noise

Correct Answer: B. It is a constant-envelope modulation

Question 7

A system using 64-QAM without coding carries how many bits per symbol?

A. 4

B. 5

C. 6

D. 8

Correct Answer: C. 6

Question 8

The matched filter is designed primarily to

A. maximize sampled signal-to-noise ratio in white noise

B. increase carrier frequency

C. remove all multipath without channel knowledge

D. double the symbol rate

Correct Answer: A. maximize sampled signal-to-noise ratio in white noise

Question 9

If the symbol rate is 4 Mbaud and QPSK is used without coding overhead, the bit rate is

A. 2 Mbit/s

B. 4 Mbit/s

C. 8 Mbit/s

D. 16 Mbit/s

Correct Answer: C. 8 Mbit/s

Question 10

Increasing QAM order while keeping average symbol energy fixed generally

A. increases minimum constellation distance

B. reduces spectral efficiency

C. reduces minimum constellation distance and increases SNR requirements

D. eliminates phase-noise sensitivity

Correct Answer: C. reduces minimum constellation distance and increases SNR requirements
DIGITAL COMMUNICATIONS
Chapter 6 of 6