Digital Communication System Foundations

Introduction and context

Digital communication is the transfer of information using signals represented, processed, transmitted, and recovered in discrete form. Modern cellular networks, optical-fibre systems, satellite links, wireless sensor networks, industrial control systems, and the Internet all depend on digital communication. Its importance arises not merely from representing information with binary digits, but from the ability to regenerate signals, detect errors, encrypt data, compress sources, share channels efficiently, and integrate communication with computation.

A digital communication system must solve two related problems. First, it must represent a source efficiently and faithfully. Second, it must convey that representation through an imperfect physical channel. The engineering objective is therefore a balance among data rate, bandwidth, transmitted power, reliability, latency, complexity, cost, and security.

The basic intellectual framework is the communication-system model: source, source encoder, channel encoder, modulator, physical channel, demodulator, channel decoder, source decoder, and destination. These blocks are conceptually distinct, although practical systems frequently combine them.

Learning objectives

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

1. Explain the architecture and purpose of a digital communication system.
2. Distinguish analogue and digital representations, symbols, bits, samples, and waveforms.
3. Apply sampling, quantization, entropy, bit-rate, bandwidth, and signal-to-noise concepts.
4. Explain source coding, channel coding, modulation, synchronization, and detection.
5. Interpret the relationship among energy per bit, noise spectral density, bandwidth, and error probability.
6. Evaluate design trade-offs and identify practical impairments and implementation limitations.
7. Analyse a communication link using a structured engineering procedure.

1. The communication-system model

Information begins at a source. The source may produce text, measurements, speech, images, video, or machine-generated data. A transducer converts a physical quantity into an electrical or digital representation when necessary. The transmitter prepares the information for transmission, while the receiver reconstructs an estimate at the destination.

A generalized chain is:

Figure: Functional blocks of a digital communication system.

The source encoder removes redundancy. For example, a lossless text compressor represents frequently occurring characters with fewer bits. A lossy speech or image codec may remove information that is perceptually less important. The channel encoder deliberately adds structured redundancy so that the receiver can detect or correct errors caused by the channel.

The modulator maps bits or coded symbols to a physical waveform suitable for the channel. In a wireless system this includes pulse shaping, frequency conversion, amplification, antenna radiation, and often multiple-antenna processing. The receiver performs the reverse operations, but it generally obtains an estimate rather than the exact transmitted sequence.

2. Information representation and terminology

A bit is a binary digit with two possible values, conventionally 0 and 1. A symbol is one member of a finite alphabet used by a modulation or coding scheme. If an alphabet has M equally likely symbols, each symbol carries log₂M bits before coding overhead. Thus, QPSK has M = 4 and carries 2 bits per symbol under the usual binary mapping.

The bit rate R_b is the number of information or transmitted bits per second, measured in bit/s. The symbol rate R_s, also called baud rate, is the number of transmitted symbols per second. For an uncoded M-ary system with k = log₂M bits per symbol,

Here, R_b and R_s are in bit/s and symbols/s, respectively. With coding rate r_c (the ratio of information bits to coded bits), the information bit rate becomes R_info = r_c R_s log₂M. This distinction is essential: baud is not synonymous with bit/s unless each symbol represents one bit.

A baseband signal occupies frequencies near zero before carrier translation. A passband signal is shifted around a carrier frequency. A waveform is the time-domain signal associated with a symbol or sequence. The constellation is a geometric representation of modulation symbols in the in-phase (I) and quadrature (Q) plane.

3. From analogue sources to binary sequences

Sampling

For a band-limited continuous-time signal x(t) whose highest frequency component is no greater than B hertz, ideal uniform sampling permits perfect reconstruction when

where f_s is the sampling frequency in samples/s. This is the Nyquist sampling condition. It assumes ideal band limitation and an appropriate reconstruction filter. In practical systems, a guard margin is used because real signals are not perfectly band-limited and filters are not ideal.

If f_s is below 2B, spectral replicas overlap and aliasing occurs. Aliasing is irreversible after sampling; increasing later processing cannot recover the lost distinction. An anti-aliasing low-pass filter is therefore placed before the sampler.

Quantization

Sampling discretizes time, whereas quantization discretizes amplitude. An N-bit uniform quantizer with 2ᴺ levels over a full-scale range V_FS has ideal step size

where ∆ is in volts if V_FS is in volts. The quantization error e_q is commonly bounded by −∆/2 ≤ e_q < ∆/2 under a mid-rise or equivalent ideal model. If the error is treated as uniformly distributed and uncorrelated with the signal, its mean-square value is

This model is an approximation and is most credible for sufficiently active signals that exercise many quantizer levels. It is less accurate for low-level, deterministic, or highly correlated inputs.

For an ideal N-bit quantizer driven by a full-scale sinusoid, the approximate signal-to-quantization-noise ratio is

The expression assumes a sinusoid spanning the quantizer range and ideal uniform quantization. It should not be applied indiscriminately to arbitrary signals.

Source coding and entropy

Source coding seeks a compact representation. For a discrete random variable X with outcomes x_i and probabilities p_i, the Shannon entropy is

Entropy is measured in bits per source symbol and is a lower bound on the average length of an ideal lossless code under standard coding assumptions. It is not necessarily the exact bit rate of a practical compressor. Huffman coding, arithmetic coding, predictive coding, transform coding, and vector quantization are examples of source-coding methods.

A source-code design must distinguish compression from error protection. Compression removes redundancy; channel coding adds controlled redundancy. Applying aggressive compression before an unreliable channel can make residual errors more damaging because the compressed stream contains less recoverable redundancy.

4. Channel models and impairments

A channel is a mathematical and physical model of the path between transmitter and receiver. A simple baseband additive-noise model is

where s(t) is the transmitted waveform, r(t) is the received waveform, and n(t) is additive noise. In a wireless fading channel, a more realistic model is

where h(t) is a possibly time-varying complex channel gain. If multipath produces delayed replicas, the channel is dispersive and may create intersymbol interference (ISI).

Important impairments include thermal noise, interference from other systems, attenuation, shadowing, multipath fading, oscillator phase noise, frequency offset, timing error, nonlinear amplifier distortion, quantization effects, and hardware imbalance. These impairments do not have identical consequences. AWGN may cause random symbol errors, whereas a timing offset can corrupt a sequence systematically and a nonlinear amplifier can broaden the spectrum.

For a resistor at absolute temperature T, the thermal-noise power over bandwidth B is commonly modeled as

where k = 1.380649 × 10⁻²³ J/K is Boltzmann’s constant. This expression assumes an idealized white-noise model over the relevant bandwidth. Receiver noise figure and implementation losses may require replacing the ideal noise temperature with an effective system noise temperature.

The signal-to-noise ratio is

where P_s and P_n are average signal and noise powers measured over the same bandwidth. In decibels, SNR_dB = 10 log₁₀(SNR). Power ratios use 10 log₁₀; amplitude ratios use 20 log₁₀ when the impedances are identical.

For digital links, energy per information bit is often more informative than average power alone. The ratio E_b/N₀ compares energy per bit E_b with one-sided noise power spectral density N₀. It permits comparison of modulation schemes at different bit rates and bandwidths. A link budget may also use received power, path loss, antenna gains, implementation loss, and required receiver sensitivity.

5. Pulse modulation and band-limited transmission

A symbol sequence can be represented as

where aₖ is the kth symbol amplitude or complex symbol, p(t) is the transmit pulse, and T = 1/R_s is the symbol interval. The pulse-shaping filter controls occupied bandwidth and ISI.

The ideal Nyquist zero-ISI condition requires p(nT) = 0 for every nonzero integer n and p(0) = 1 at the decision sample. The sinc pulse satisfies this condition mathematically but extends infinitely in time. Practical systems use raised-cosine or root-raised-cosine filtering, which trades excess bandwidth against time-domain decay and implementation complexity.

For a raised-cosine spectrum with symbol rate R_s and roll-off factor α, 0 ≤ α ≤ 1, the approximate one-sided baseband bandwidth is

and the corresponding passband null-to-null bandwidth is approximately (1 + α)R_s. The exact interpretation depends on whether baseband or passband bandwidth is being specified. A larger α eases timing and filter truncation requirements but consumes more bandwidth.

6. Digital modulation and detection

Digital modulation maps symbols to signal parameters. In amplitude-shift keying (ASK), amplitude changes; in frequency-shift keying (FSK), frequency changes; in phase-shift keying (PSK), phase changes; and in quadrature amplitude modulation (QAM), both amplitude and phase vary.

Binary phase-shift keying (BPSK) uses two antipodal symbols. Quadrature phase-shift keying (QPSK) uses four phases and conveys two bits per symbol. M-QAM increases spectral efficiency by using a larger constellation, but adjacent points become closer for a fixed average energy, generally increasing sensitivity to noise and nonlinear distortion.

A coherent receiver requires estimates of carrier phase, frequency, and symbol timing. It correlates or filters the received signal and chooses the most likely symbol. In AWGN with equally likely symbols and equal energy, maximum-likelihood detection selects the constellation point with minimum Euclidean distance from the received point. Noncoherent detection avoids an absolute carrier phase reference but often incurs a performance or complexity trade-off.

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

where Q(x) = (1/√(2π))∫ₓ^∞ exp(−u²/2) du. This formula assumes ideal synchronization, equally likely bits, coherent detection, and additive white Gaussian noise. It is not a universal error formula for fading, coding, or arbitrary modulation.

7. Channel coding and reliability

Channel coding adds redundancy according to a code structure. A block code maps k information bits into n coded bits and has code rate R_c = k/n. A convolutional or modern iterative code processes sequences using memory. Common families include convolutional codes, turbo codes, low-density parity-check (LDPC) codes, and polar codes.

A decoder may use hard decisions, which retain only binary decisions, or soft decisions, which retain reliability information. Soft information usually improves performance at the cost of greater computational and memory requirements. The trade-off is not simply between redundancy and reliability: coding also affects latency, decoder energy, buffer size, and throughput.

The bit-error rate (BER) is the fraction of incorrectly received bits in a specified observation interval. The block-error rate (BLER) is the fraction of blocks containing one or more errors. A low BER does not automatically imply an acceptable application outcome; a single error in a control message may be more consequential than many errors in a robust audio stream.

Automatic repeat request (ARQ) detects errors and requests retransmission. Forward error correction (FEC) corrects errors without retransmission. Hybrid ARQ combines both and is widely useful where feedback exists. Real-time or broadcast systems may favour FEC because retransmission delay is unacceptable.

8. Synchronization and receiver architecture

Synchronization is a foundational receiver task. Timing synchronization identifies the appropriate sampling instants. Carrier synchronization estimates frequency and phase. Frame synchronization identifies boundaries, while network synchronization aligns clocks among nodes.

A receiver commonly includes an RF front end, down-converter, automatic gain control, anti-aliasing filter, analogue-to-digital converter, digital down-converter, matched filter, timing-recovery loop, carrier-recovery loop, equalizer, detector, decoder, and protocol processing. The matched filter maximizes the sampled signal-to-noise ratio at the decision time for a known pulse in AWGN.

Equalization compensates for channel distortion and ISI. Linear equalizers are simple but can enhance noise when the channel has spectral nulls. Decision-feedback equalizers reduce such enhancement but can propagate errors. Modern multicarrier systems such as orthogonal frequency-division multiplexing divide a frequency-selective channel into many narrowband subchannels and use a cyclic prefix to simplify equalization, at the cost of overhead and sensitivity to frequency offset and high peak-to-average power ratio.

9. Worked example: designing a small QPSK link

An uncoded QPSK transmitter must carry an information rate of 2 Mbit/s. The pulse-shaping roll-off factor is α = 0.25. Determine the symbol rate and approximate passband bandwidth, assuming ideal filtering and no coding overhead.

QPSK has M = 4, so log₂4 = 2 bits per symbol. Therefore,

Thus, R_s = 1 Mbaud. With raised-cosine shaping, the approximate passband bandwidth is

The result is approximately 1.25 MHz. If a code with rate R_c = 3/4 were added while preserving the same information rate, the coded bit rate would be 2.667 Mbit/s, the symbol rate would become 1.333 Mbaud, and the approximate bandwidth would become 1.667 MHz. This illustrates that error protection consumes resources unless other parameters are changed.

10. Applied case: a reliable industrial wireless link

Background

Consider an illustrative factory-monitoring system that reports temperature, vibration, and motor-current measurements from battery-powered sensors to a gateway. Metal structures cause multipath, and the system must deliver alarms with high reliability while conserving battery energy.

Problem

A design based only on increasing transmit power is undesirable. It would increase interference, battery consumption, and regulatory risk, yet may not solve deep fading or packet collisions.

Analysis

The engineering team first measures path loss, fading statistics, interference occupancy, and required alarm latency. It separates ordinary telemetry from urgent alarms. Telemetry can tolerate occasional retransmission, whereas alarms require bounded latency and stronger protection. Packet length is also examined because longer packets have a greater probability of containing an error at a fixed BER.

Approach and solution

The system uses source coding to represent sensor values efficiently, a moderate-rate FEC code for alarms, interleaving to spread burst errors, and a robust low-order modulation during poor channel conditions. Pilot symbols support channel estimation; frequency and timing recovery are tested across temperature and oscillator tolerances. Adaptive modulation and coding is used when channel estimates are reliable. Acknowledged retransmission is applied to ordinary telemetry but limited for alarms. The gateway schedules transmissions to reduce collisions, and cryptographic authentication prevents forged control messages.

Outcome

The resulting design does not maximize peak data rate. Instead, it meets the application’s reliability and latency targets with lower average energy than a high-power uncoded design. Field tests remain necessary because laboratory AWGN results do not represent factory multipath, interference, antenna detuning, or installation variability.

Learning points

1. Link design begins with application requirements, not with a modulation name.
2. Coding, diversity, scheduling, antenna design, and retransmission can be more effective than power increase alone.
3. Reliability must be defined using an application metric such as alarm-delivery probability and latency, not BER alone.
4. Security and regulatory compliance are system requirements, not optional add-ons.

11. Design trade-offs, governance, and contemporary developments

The central trade-off is represented by the communication triangle of bandwidth, power, and reliability. Increasing modulation order improves spectral efficiency but reduces distance between constellation points. Increasing coding redundancy improves error performance but reduces net throughput. Increasing bandwidth can reduce noise density per bit under some fixed-rate conditions, but spectrum is scarce and front-end hardware becomes more demanding.

The Shannon capacity of an ideal band-limited AWGN channel is

where C is channel capacity in bit/s, B is bandwidth in hertz, and S/N is received signal-to-noise power ratio over that bandwidth. This is an asymptotic theoretical limit for arbitrarily long codes and acceptably small error probability under the stated channel model. It is not a direct promise about a particular finite-latency implementation.

Practical systems must observe spectrum licences, electromagnetic-compatibility limits, safety requirements, lawful-interception rules where applicable, and privacy and cybersecurity obligations. Encryption protects confidentiality, authentication protects integrity and origin, and robust key management is essential. Designers should avoid collecting unnecessary personal data and should document assumptions behind reliability claims.

Contemporary research addresses massive multiple-input multiple-output systems, millimetre-wave and terahertz links, reconfigurable intelligent surfaces, integrated sensing and communication, semantic and goal-oriented communication, optical wireless systems, machine-learning-assisted receivers, and energy-efficient edge intelligence. Artificial intelligence can assist channel estimation, resource allocation, modulation recognition, and anomaly detection; however, data bias, explainability, adversarial manipulation, computational cost, and out-of-distribution behaviour must be evaluated. A learned receiver should be benchmarked against analytically understood baselines and tested under channel conditions not used during training.

12. Best-practice workflow for link analysis

A disciplined analysis normally proceeds as follows:

1. Define the application’s payload, latency, reliability, mobility, range, and energy requirements.
2. Specify the source and distinguish information rate from gross transmitted bit rate.
3. Select source coding and quantify compression assumptions.
4. Select channel coding, code rate, packet size, and decoding method.
5. Choose modulation and pulse shaping from bandwidth and error-performance requirements.
6. Build a link budget, including antenna gains, path loss, noise figure, fading margin, and implementation loss.
7. Analyse synchronization, equalization, nonlinearities, and hardware tolerances.
8. Evaluate BER, BLER, throughput, latency, energy per delivered bit, and outage probability.
9. Validate analytical predictions with simulation and measurements.
10. Review security, spectrum regulation, safety, maintainability, and lifecycle cost.

The most common analytical mistake is to treat an ideal AWGN result as a complete system prediction. A credible engineering conclusion states the channel model, coding assumptions, synchronization quality, measurement bandwidth, and confidence or uncertainty of the result.

Chapter Summary

Digital communication systems transform source information into structured signals, convey them through a physical channel, and reconstruct an estimate at the destination. Source coding removes redundancy, channel coding adds controlled redundancy, modulation maps symbols to waveforms, and synchronization and detection enable recovery. Sampling requires a rate at least twice the highest signal frequency under ideal band-limited assumptions; quantization introduces amplitude error; and entropy describes the theoretical limit of lossless source compression.

System performance depends on bandwidth, energy per bit, noise, interference, fading, pulse shaping, synchronization, coding, modulation, and implementation constraints. The distinction between bit rate and symbol rate is essential, as is the distinction between BER, BLER, throughput, latency, and application-level reliability. Shannon capacity provides a theoretical benchmark, while practical design requires link budgets, simulation, measurements, security controls, regulatory compliance, and explicit treatment of limitations. Effective communication engineering optimizes the complete system rather than any isolated block.

Key Terms

Bit: A binary digit with value 0 or 1.

Symbol: One element of a modulation or coding alphabet.

Bit rate: Number of bits transmitted or processed per second.

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

Sampling: Conversion from continuous time to discrete time.

Aliasing: Irreversible spectral overlap caused by insufficient sampling.

Quantization: Mapping continuous amplitude values to discrete levels.

Entropy: Average information content of a discrete source.

Source coding: Representation intended to reduce redundancy.

Channel coding: Addition of structured redundancy for error control.

Modulation: Mapping symbols to physical signal waveforms.

Constellation: Geometric representation of complex modulation symbols.

AWGN: Additive white Gaussian noise channel model.

SNR: Ratio of average signal power to average noise power.

E_b/N₀: Energy per information bit divided by noise spectral density.

ISI: Intersymbol interference caused by overlapping channel responses.

Synchronization: Recovery of timing, carrier, frame, or network references.

Equalization: Compensation for channel-induced distortion and ISI.

BER: Fraction of received bits detected incorrectly.

Shannon capacity: Theoretical maximum reliable rate for a specified ideal channel model.

Review Questions

Short Answer Questions

1. Distinguish bit rate from symbol rate and explain why they are not always equal.
2. State the Nyquist sampling condition and explain aliasing.
3. What are the different purposes of source coding and channel coding?
4. Define SNR and E_b/N₀ and explain why both are useful.
5. Why are timing and carrier synchronization necessary in a coherent receiver?

Long Answer Questions

1. Explain the complete digital communication-system model and the function of every major block.
2. Derive the relationship between QPSK symbol rate and information bit rate, including the effect of coding rate.
3. Discuss pulse shaping, the zero-ISI condition, and the bandwidth–roll-off trade-off.
4. Compare ARQ, FEC, and hybrid ARQ in terms of reliability, latency, bandwidth, and feedback requirements.
5. Explain why an ideal AWGN BER result may differ substantially from field performance.

Analytical / Application Questions

1. A 16-QAM link carries 12 Mbit/s using a code rate of 3/4. Calculate the symbol rate and approximate raised-cosine passband bandwidth for α = 0.2. State all assumptions.
2. Design an outline for a battery-powered industrial sensor link requiring low latency and high alarm reliability. Justify source coding, channel coding, modulation, retransmission, synchronization, and security choices.
3. A sampled sensor signal contains frequencies up to 18 kHz. Determine the minimum ideal sampling frequency, then explain why a practical design may choose a higher rate and an anti-aliasing filter.

Multiple Choice Questions with Answers

Question 1

For an uncoded QPSK system, how many bits are represented by each symbol?

A. 1 bit

B. 2 bits

C. 3 bits

D. 4 bits

Correct Answer: B. 2 bits

Question 2

What is the minimum ideal sampling frequency for a signal band-limited to 10 kHz?

A. 5 kHz

B. 10 kHz

C. 20 kHz

D. 40 kHz

Correct Answer: C. 20 kHz

Question 3

Which operation deliberately adds structured redundancy to improve error recovery?

A. Source coding

B. Channel coding

C. Sampling

D. Quantization

Correct Answer: B. Channel coding

Question 4

For a raised-cosine pulse with symbol rate R_s and roll-off α, the approximate passband bandwidth is:

A. R_s/(1 + α)

B. (1 − α)R_s

C. (1 + α)R_s

D. αR_s/2

Correct Answer: C. (1 + α)R_s

Question 5

Which quantity is measured in baud?

A. Information entropy

B. Symbol rate

C. Noise power

D. Channel capacity

Correct Answer: B. Symbol rate

Question 6

Under the ideal uniform quantization-noise model, the quantization-error variance is:

A. ∆/12

B. ∆²/12

C. 12∆²

D. 2∆

Correct Answer: B. ∆²/12

Question 7

For coherent BPSK in AWGN, which expression gives the bit-error probability?

A. Q(√(E_b/N₀))

B. Q(2E_b/N₀)

C. Q(√(2E_b/N₀))

D. Q(E_b²/N₀)

Correct Answer: C. Q(√(2E_b/N₀))

Question 8

What is the principal purpose of an anti-aliasing filter?

A. Increase carrier frequency

B. Prevent spectral overlap before sampling

C. Add channel-coding redundancy

D. Recover frame boundaries

Correct Answer: B. Prevent spectral overlap before sampling

Question 9

A code has 120 information bits and produces 160 coded bits. What is its code rate?

A. 0.25

B. 0.50

C. 0.75

D. 1.33

Correct Answer: C. 0.75

Question 10

Shannon capacity for an ideal AWGN channel is expressed as:

A. C = B log₂(1 + S/N)

B. C = S log₂(1 + B/N)

C. C = B(S/N)²

D. C = N log₂(1 + S/B)

Correct Answer: A. C = B log₂(1 + S/N)