Loading field map
Communication systems is the branch of electrical engineering concerned with the reliable transmission of information from one point to another. Its central problem is deceptively simple: given a message that must be delivered, how can it be represented, sent over a physical medium, and recovered at the destination with acceptable fidelity, despite noise, distortion, and limits on power and bandwidth? The field studies both the theoretical limits of what is possible and the practical techniques for approaching those limits.
Every communication system consists of the same logical elements. A source produces a message—speech, text, images, video, or data. A transmitter transforms this message into a signal suitable for the channel, a process called modulation. The channel is the physical medium: a wire, optical fiber, radio waves through free space, or even a storage medium. The channel inevitably degrades the signal, adding noise (random disturbances) and distortion (systematic alterations). A receiver attempts to invert the transmitter's operations and recover the original message from the corrupted signal.
The field's central questions follow from this structure. How much information can a channel carry reliably? This is the domain of information theory, which provides fundamental limits. How should signals be designed to be robust against noise? This is the domain of modulation and coding. How can a receiver best estimate the transmitted message from a noisy observation? This is the domain of detection and estimation theory. How can multiple users share the same channel? This is the domain of multiple access and networking.
The modern field begins with Claude Shannon's 1948 work, which created information theory. Shannon asked a question that had not been precisely posed before: what is information, and how much of it can be transmitted reliably?
Shannon defined information in terms of entropy, a measure of the uncertainty or surprise associated with a message source. A source that produces highly predictable messages carries less information per symbol than one that produces equally likely alternatives. The channel capacity is the maximum rate at which information can be transmitted over a given channel with arbitrarily low error probability. Shannon's noisy-channel coding theorem proved that if the transmission rate is below capacity, there exists a coding scheme that achieves arbitrarily reliable communication; if the rate exceeds capacity, reliable communication is impossible.
This result was revolutionary because it separated the problem of representing information from the problem of protecting it. Before Shannon, engineers thought of noise as an unavoidable degradation that could only be mitigated by increasing power. Shannon showed that the fundamental limit was not power but the ratio of signal to noise, and that clever encoding could approach the limit. The theorem is a possibility result: it proves that good codes exist but does not say how to find them. Much of subsequent communication theory has been the search for practical codes that approach Shannon's limit.
Information theory also introduced the concept of source coding (compression), which removes redundancy from a message, and channel coding (error correction), which deliberately adds redundancy to protect against noise. These two operations are logically distinct and can be performed separately, a separation that is both a theoretical convenience and a practical design principle.
Before Shannon, and continuing alongside information theory, is the classical framework of modulation theory. This tradition treats the design of signals themselves: how to map information onto waveforms that can be transmitted through physical channels.
The basic distinction is between analog modulation and digital modulation. Analog modulation, such as amplitude modulation (AM) and frequency modulation (FM), varies a continuous carrier parameter in proportion to the message. These techniques were developed in the early twentieth century for radio and telephony. Digital modulation, by contrast, maps discrete symbols onto a finite set of waveforms. Common schemes include phase-shift keying (PSK), frequency-shift keying (FSK), and quadrature amplitude modulation (QAM), which varies both amplitude and phase.
The receiver's task in this framework is detection: given the received waveform, decide which symbol was most likely sent. The optimal rule, under the assumption of additive white Gaussian noise, is the maximum-likelihood detector, which chooses the symbol that maximizes the probability of the observed signal. This framework treats the channel as a probabilistic mapping and the receiver as a statistical decision maker. The key performance measure is the bit error rate as a function of signal-to-noise ratio.
The classical framework is fundamentally about the trade-off between bandwidth and power. A modulation scheme that uses more bandwidth can transmit more symbols per second, but each symbol is more vulnerable to noise. A scheme that uses more power can support more symbols per second at a given error rate. The design space is characterized by the bandwidth-efficiency plane, where different schemes occupy different regions.
The gap between what modulation alone could achieve and what Shannon's theorem promised was the motivation for error control coding. The idea is to add structured redundancy to the transmitted sequence so that the receiver can detect and correct errors.
The first major class of codes, developed in the 1950s, were block codes, which group data into fixed-length blocks and append parity symbols. Hamming codes, Reed–Solomon codes, and BCH codes are examples. These codes have algebraic structure that allows efficient decoding, but they were limited in how close they could approach Shannon's capacity.
A second major class, developed in the 1960s, were convolutional codes, which process a continuous stream of data through a shift register, producing a sequence of output symbols that depend on the current and previous inputs. These codes are decoded with the Viterbi algorithm, a dynamic programming method that finds the most likely transmitted sequence. Convolutional codes were widely used in deep-space missions and early digital cellular systems.
The field reached a turning point in the 1990s with the discovery of turbo codes and the rediscovery of low-density parity-check (LDPC) codes, originally invented by Robert Gallager in the 1960s but computationally impractical at the time. These codes use iterative decoding algorithms that exchange probabilistic information between the code's structure and the channel observations. They achieve performance within a fraction of a decibel of Shannon's limit, effectively closing the gap that had motivated decades of research. Modern systems, from satellite links to Wi-Fi to 5G cellular, use these near-capacity codes.
A communication receiver must do more than detect symbols. It must also estimate unknown parameters of the channel and the transmitter. Synchronization is the problem of aligning the receiver's clock and phase with the transmitter's. Channel estimation is the problem of determining how the channel has attenuated and delayed the signal. These are problems of estimation theory, which treats unknown parameters as quantities to be inferred from noisy observations.
The standard framework distinguishes between coherent and non-coherent reception. Coherent receivers know the carrier phase and can use it to make better decisions; non-coherent receivers do not. The cost of coherence is the need for training sequences or pilot symbols that consume bandwidth. This trade-off between estimation overhead and detection performance is a recurring theme in system design.
Estimation theory also underlies the treatment of fading channels, where the signal strength varies randomly due to multipath propagation. In mobile radio, signals arrive at the receiver via multiple paths, adding constructively or destructively. This creates fading, which can cause deep nulls that temporarily destroy the signal. The statistical characterization of fading channels, and the design of systems that mitigate their effects through diversity (transmitting over multiple independent paths) or interleaving (spreading errors so they can be corrected), is a major subfield.
When many users share a single channel, the system must allocate resources among them. This is the problem of multiple access. The classical approaches are frequency-division multiple access (FDMA), which gives each user a distinct frequency band; time-division multiple access (TDMA), which gives each user a distinct time slot; and code-division multiple access (CDMA), which gives each user a distinct spreading code so that all transmit simultaneously over the same band.
The choice among these schemes involves a fundamental trade-off. FDMA and TDMA are simple and avoid interference but waste resources when users are idle. CDMA is more flexible and robust to interference but requires careful power control and more complex receivers. The modern trend is toward orthogonal frequency-division multiplexing (OFDM), which divides the channel into many narrow subcarriers and assigns them dynamically to users. OFDM is the basis of Wi-Fi, 4G, and 5G because it is robust to frequency-selective fading and allows flexible resource allocation.
A deeper question is whether multiple users can share a channel more efficiently than by simply partitioning it. Multiuser information theory studies the fundamental limits of such systems. The key result is that the sum capacity of a multiple-access channel can exceed what is achievable by orthogonal partitioning, because users can be separated by their distinct signatures rather than by exclusive resources. This insight underlies modern massive MIMO (multiple-input multiple-output) systems, where base stations with many antennas serve many users simultaneously, separating them in space.
The contemporary field is characterized by the convergence of these traditions. Modern systems are adaptive: they measure the channel and adjust their modulation, coding, and power accordingly. Adaptive modulation and coding is standard in Wi-Fi and cellular systems, where the transmitter selects a rate based on the current channel quality.
The field has also become increasingly iterative. The turbo principle—exchanging soft information between decoding stages—has been generalized to iterative receivers that jointly perform detection, decoding, and channel estimation. This blurs the classical separation between these functions and approaches the joint optimality that information theory suggests is possible.
Another major development is the extension of communication theory to wireless networks with many nodes, where the problem is not just point-to-point transmission but the coordination of a distributed system. Network information theory studies the fundamental limits of such systems, including relaying, interference management, and broadcast. This is an active research area with many open problems.
The field also increasingly intersects with signal processing and machine learning. Deep learning has been applied to channel decoding, channel estimation, and even end-to-end system design, where a neural network learns a modulation and coding scheme directly from data. These approaches are promising but not yet fully understood theoretically, and their relationship to the classical framework remains an active research question.
Several tensions structure the field and are unlikely to disappear. The first is between theory and practice: information theory provides limits and existence proofs, but practical systems must be implementable with finite complexity and latency. The gap between what is theoretically possible and what is practically achievable is a permanent feature of the field.
The second is between bandwidth and power. Every system operates under constraints on both, and the optimal design depends on which is scarcer. Satellite links are power-limited; terrestrial wireless is bandwidth-limited. The same theory applies, but the design choices differ.
The third is between simplicity and performance. The best-performing codes and receivers are often the most complex. The history of the field is partly the history of finding clever simplifications—algebraic structure, iterative algorithms, suboptimal but near-optimal detectors—that capture most of the theoretical gain at a fraction of the cost.
The fourth is between universality and specialization. The core theory—Shannon's information theory, the statistical framework of detection and estimation—applies to all channels. But the details of a fiber-optic system, a satellite link, a cellular network, and an underwater acoustic channel are so different that practitioners in each area develop specialized knowledge. The field is unified by its mathematics but diversified by its applications.
Communication systems is thus a mature field with a solid theoretical core and a rich set of practical techniques. Its central questions—how much information can be sent, how to send it reliably, how to share the channel fairly—remain the same as they were at the field's founding, but the answers have become increasingly sophisticated, and the gap between what is possible and what is achieved continues to narrow.