A phased array is an arrangement of transmitting or receiving elements whose signals are combined with controlled relative phases, allowing the resulting wave field or directional response to be steered and shaped. In a radio-frequency system, the elements are usually antennas; in an acoustic system, they are transducers. Constructive interference reinforces waves in selected directions, while destructive interference reduces them elsewhere. Changing the electrical control settings can redirect a beam without physically turning the aperture.
Phased arrays are important in radar, telecommunications, radio astronomy, sonar, ultrasonic imaging, and optical beam steering. Their principal advantages are rapid directional control, flexible beam patterns, and the ability to share an aperture among several tasks. These capabilities come with substantial requirements for electronic hardware, computation, calibration, and thermal management.
The term is sometimes used narrowly for arrays controlled by phase shifters and sometimes more broadly for systems using true time delays or digital signal processing. Not every antenna array is a phased array: a fixed array may have permanently established feed phases, while some electronically scanned antennas steer primarily by changing frequency. Practical phased-array performance depends not only on the number of elements but also on their spacing, radiation patterns, bandwidth, control accuracy, and interaction with the surrounding structure.
Historical development
The physical basis of phased arrays follows from the interference of waves, but their engineering development required methods for distributing coherent signals and controlling their relative phases. Late nineteenth- and early twentieth-century experiments with multiple radio antennas established that directional patterns could be produced electrically rather than solely through reflectors or mechanical pointing. In 1905, Ferdinand Braun demonstrated directional radio transmission using multiple antennas with controlled phase relationships. Such early systems offered limited switching among directions rather than the continuously programmable scanning associated with later arrays.
Antenna-array theory developed alongside broadcasting, long-distance radio, and direction finding. Engineers learned to treat an array pattern as the combination of an individual element’s response and the interference pattern produced by the positions and excitations of all elements. Work by Dolph in 1946 established a systematic relationship between beamwidth and equal-ripple sidelobes for suitable linear arrays. Taylor’s later aperture distributions offered another widely used approach to controlling sidelobes. These methods remain relevant because electronic steering does not remove the fundamental trade-off between angular concentration and unwanted radiation.
During and after World War II, the growth of radar created strong demand for faster scanning and more flexible tracking. Early radar antennas generally rotated or nodded mechanically. Electronically steered systems became more practical as ferrite phase shifters, microwave feed networks, and electronic control equipment matured. The AN/FPS-85 space-surveillance radar at Eglin Air Force Base, operational from 1969, became an important early example of a large electronically scanned installation. Naval systems such as AN/SPY-1 subsequently demonstrated how fixed array faces could support repeated surveillance and tracking over wide sectors.
The transition from centralized transmitters to distributed solid-state electronics changed both architecture and reliability. In an active electronically scanned array, many elements or small element groups have their own transmit-and-receive electronics. This became increasingly practical through advances in microwave semiconductors, packaging, and digital control during the late twentieth century. Gallium arsenide devices supported compact microwave modules, while gallium nitride later enabled higher power density and demanding operating conditions.
By the early twenty-first century, phased-array techniques were no longer confined to large scientific and defense installations. Cellular base stations, satellite user terminals, automotive radar, and industrial ultrasound adopted related architectures at very different scales. Their development was not a single progression toward one ideal design: low-cost communications panels, high-power surveillance radars, and dense medical transducer arrays each evolved around distinct bandwidth, aperture, power, and manufacturing constraints.
Interference, phase control, and beam formation
The central operation of a phased array is beamforming: imposing relative complex weights on signals associated with different spatial positions. On transmission, these weights determine how individual waves combine in space. On reception, they determine how voltages or sampled signals are combined to favor waves arriving from a selected direction. For a reciprocal antenna operating through reciprocal signal paths, transmitting and receiving patterns are closely related, although practical transmit and receive electronics may require separate calibration.
For a uniform linear array of identical elements separated by distance , an idealized array factor can be written as
where is the amplitude assigned to element , is the wave number, is the wavelength, and is measured from the array’s broadside direction. The progressive phase shift is . With this sign convention, reinforcement toward occurs when
The phase shift compensates for the different propagation distances from the elements toward the selected direction.
This process does not require each individual element to form a narrow beam. Relatively broad element patterns can combine to produce a much narrower array pattern. Under an idealized pattern-multiplication model, the total field pattern is the product of the element pattern and the array factor. The approximation becomes less exact when elements differ substantially or interact strongly. Nevertheless, it explains why an element that radiates poorly at an extreme angle limits the array’s ability to scan there, even if the phase settings are otherwise correct.
Receiving arrays exploit coherence rather than simply adding power. After alignment, a desired plane-wave signal adds consistently across channels, whereas independent receiver noise does not add with the same phase relationship. Under ideal conditions with equal channels and uncorrelated noise, coherent combination of elements improves signal-to-noise ratio by a factor of relative to one element. External noise correlation, unequal gains, and interference modify this result. Transmit comparisons likewise require care: increasing directivity while holding total radiated power fixed differs from increasing both the number of elements and total transmitter power.
A far-field beam describes direction, not a uniquely selected distance. When a target lies in the near field, the wavefront curvature across the aperture matters. Appropriate delays can then focus energy at a finite position rather than align it along a single plane-wave direction. This distinction is essential in ultrasound and increasingly relevant to large communications apertures operating at short range.
Aperture geometry, beamwidth, and sidelobes
Array geometry determines which spatial wave patterns can be distinguished or generated. A linear array primarily controls one angular coordinate; a planar array controls two. Circular, cylindrical, spherical, and conformal arrangements support other coverage requirements. Conformal arrays follow a supporting surface, such as an aircraft body, but require models that account for element orientation, structural blockage, and nonuniform coupling. The relevant aperture is therefore an electromagnetic or acoustic structure, not merely a count of radiating elements.
For a uniformly illuminated continuous rectangular aperture of width , the broadside half-power beamwidth in the corresponding principal plane is approximately
in radians when the beam is narrow. A one-metre aperture at a wavelength of three centimetres therefore has an idealized beamwidth of about in that plane. Finite element spacing, illumination taper, and the element pattern change the precise value. At fixed aperture dimensions, adding elements beyond the density needed to sample the field adequately does not produce an unlimited reduction in beamwidth.
Discrete spatial sampling can create grating lobes, additional directions in which the element signals reinforce almost as strongly as in the intended beam. For a regularly spaced linear array, visible maxima satisfy
where is an integer. For ideal steering throughout the visible angular range, spacing no greater than about half a wavelength is the familiar conservative condition. A restricted scan sector may permit wider spacing. Irregular layouts can break up periodic grating lobes, but the unwanted energy generally reappears as a less orderly sidelobe structure rather than disappearing.
Even an array without grating lobes has ordinary sidelobes caused by its finite aperture. A large uniformly excited linear array has a first sidelobe approximately 13.3 decibels below its main-beam peak. Reducing excitation toward the aperture edges lowers sidelobes, but broadens the main beam and reduces aperture efficiency. Dolph–Chebyshev and Taylor distributions formalize different versions of this trade-off. Low sidelobes are especially valuable when a weak target or astronomical source lies near a much stronger signal.
Scanning away from broadside introduces further penalties. A planar aperture presents a smaller projected area, approximately proportional to , while its element patterns, polarization response, and impedance may also deteriorate. The beam usually broadens in the scan plane. At some angles, coupling to surface or other structural modes can produce severe impedance changes and scan blindness. Consequently, a practical scan limit cannot be inferred from phase-shifter range alone; it is a property of the complete aperture and its electromagnetic environment.
Passive, active, analog, and digital architectures
A passive electronically scanned array, or PESA, generally uses one or a small number of centralized transmitters, with microwave power distributed through a feed network containing controllable phase shifters. “Passive” describes the distribution architecture, not an absence of transmitted energy or powered components. Receive signals may return through a combining network to centralized receivers. Such arrangements can deliver substantial power, but feed loss, high-power switching, and dependence on central components influence efficiency and reliability.
An active electronically scanned array, or AESA, distributes amplification across many transmit-and-receive modules. A module commonly contains a power amplifier, a low-noise receive amplifier, phase and amplitude controls, switching, and protection circuits. The module may serve one element or a small subarray. Amplifying close to the antenna reduces some distribution losses, particularly before the first receive amplifier. Distributed hardware also makes limited element failures less catastrophic, although failed modules still change sensitivity, sidelobes, and calibration.
Analog beamforming combines signals before full digitization, using phase shifters, delay lines, or microwave networks. It can reduce the required number of converters and digital channels, but the number of simultaneously accessible beams is constrained by the feed architecture. Fully digital beamforming preserves separate sampled signals from individual elements, or from very small groups, and forms beams numerically. It offers considerable flexibility but imposes demanding requirements on converters, clock distribution, data movement, processing throughput, and power consumption.
Hybrid beamforming divides the task between analog and digital stages. For example, analog networks may form beams within subarrays, while digital processing combines their outputs. This can make large high-frequency arrays affordable, but the analog stage restricts the spatial information available downstream. A signal rejected by an analog subarray cannot generally be reconstructed later. Subarray spacing and internal beam patterns can also limit scan range and create secondary lobes, particularly across wide bandwidths.
Device technology and packaging strongly affect these choices. Gallium nitride is valuable where high radio-frequency power density and efficiency are important, while silicon-based integration is attractive for densely integrated, high-volume communications products. Neither material alone determines system performance. Waste heat must be removed without excessive temperature gradients, mechanical structures must retain alignment, and control electronics must coexist with sensitive receivers. Cooling plates, power distribution, electromagnetic shielding, and service access can occupy more volume than the radiating surface itself. Thus, an array that appears thin from the front may require substantial equipment behind its aperture.
Bandwidth, time delay, and beam squint
A phase shift and a time delay are equivalent only at a specified frequency, or approximately over a sufficiently narrow band. A true delay introduces a phase proportional to frequency:
An idealized fixed phase shifter instead applies approximately the same phase setting across its operating band. Since wavelength changes with frequency, the phase progression needed to steer toward a fixed angle also changes. This distinction produces beam squint, in which different frequency components point in different directions.
For a linear array set to scan toward at frequency , a fixed phase progression gives, on the corresponding main-beam branch,
At broadside, the required progressive phase is zero and this idealized steering error vanishes. Away from broadside, the error grows with fractional bandwidth and scan angle. A shift of only a few degrees can be severe for a narrow beam, so large electrical apertures may encounter significant broadband losses even when the frequency excursion seems modest.
The underlying scale is the propagation-delay difference across the aperture. For aperture extent in the scan direction, its magnitude is approximately
where is the relevant wave speed. If the product of signal bandwidth and this delay becomes appreciable, phase-only alignment becomes a poor approximation. Effects include reduced peak gain, frequency-dependent coverage, and distortion of short pulses. Wideband radar and high-data-rate communications can therefore require delay-based methods even when their center-frequency beam patterns appear satisfactory.
True-time-delay systems address the problem by compensating actual arrival-time differences. Implementations include switched transmission-line lengths, specialized microwave delay devices, photonic delay networks, and digital delays. Integer-sample digital delays are straightforward, but finer alignment requires fractional-delay filtering or an equivalent frequency-domain operation. Wideband digital arrays often divide signals into frequency channels, apply frequency-dependent steering weights, and then combine or reconstruct the desired outputs.
True time delay does not remove all bandwidth limitations. Antenna matching, polarization, amplifier response, coupling, and element spacing remain frequency-dependent. A spacing acceptable at the lowest operating frequency may permit grating lobes at the highest. Delay circuits themselves can introduce loss, amplitude ripple, dispersion, and quantization error. Practical designs often use a hierarchy: true delays between large subarrays, phase control within smaller subarrays, and digital equalization for residual errors. The relevant choice is therefore not simply phase shifting versus delay, but where each operation is performed and how much uncompensated delay remains.
Adaptive processing, calibration, and error control
Fixed steering weights assume that element positions and channel responses are known. An adaptive beamformer instead adjusts its weights using measured signals or an environmental model. It may preserve gain toward a desired source while placing nulls toward interference. This capability is important in crowded radio environments and electronic warfare, but it is not unlimited: each added constraint consumes spatial degrees of freedom, and desired signals can be damaged if their directions or statistical properties are modeled incorrectly.
A standard example is the minimum-variance distortionless-response beamformer. If is the interference-plus-noise covariance matrix and is the steering vector for the desired direction, its idealized weights are
where denotes conjugate transpose. These weights minimize output variance while maintaining a specified response toward the desired signal. In practice, the covariance must be estimated from finite data, and inversion may be poorly conditioned. Diagonal loading and related regularization methods improve robustness, particularly when snapshots are limited or the steering model is imperfect.
Calibration establishes how commanded settings relate to actual complex channel responses. Errors arise from manufacturing tolerances, unequal cable lengths, amplifier drift, temperature changes, phase-shifter quantization, and the radome. Internal reference injection can measure much of the electronic path, but may not capture the radiating element or its environment. External sources, near-field scanners, astronomical references, and mutual-coupling measurements provide complementary information. A calibration valid at one frequency, power level, or temperature may not remain valid throughout operation.
Mutual coupling makes an element’s behavior depend on its neighbors and their excitations. The embedded element pattern, measured with the other ports appropriately terminated, can differ substantially from the pattern of an isolated radiator. Coupling also affects active impedance as the beam scans. An adequate model may therefore require frequency- and direction-dependent responses rather than one gain and one phase correction per channel. For receiving arrays, correlated external noise further complicates the assumption that every channel contributes independent noise.
Random phase errors reduce coherent gain and raise sidelobes. Under a simplified model with independent, zero-mean Gaussian phase errors of variance , the coherent power contribution is reduced approximately by a factor , with phase measured in radians. Correlated errors can be more damaging in particular ways: a phase gradient shifts the beam, while periodic errors can create distinct unwanted lobes. Calibration quality must therefore be assessed through beam pointing, sidelobes, polarization, and stability, rather than peak gain alone.
Radar operation and resource scheduling
Phased arrays changed radar operation by separating beam direction from mechanical rotation. A radar can switch rapidly among search sectors, established tracks, calibration directions, and other tasks. Electronic steering itself may be extremely fast, but useful radar updates remain constrained by dwell time, propagation delay, waveform duration, processing, and the signal strength required for detection. Rapid switching does not make a weak echo observable without sufficient transmitted energy and coherent integration.
An array radar commonly maintains a schedule of beam dwells. A wide search region may be sampled with a grid of beams, while tracks receive revisits chosen according to uncertainty and expected motion. A newly detected object may receive additional measurements to establish a track. These functions compete for transmitter duty cycle, receiver time, energy, and processing capacity. The scheduler is consequently part of the sensing system, not merely an administrative layer. Increasing the update rate for one task can reduce coverage or sensitivity elsewhere.
A single planar face usually covers a restricted sector effectively. Several fixed faces can provide broad azimuthal coverage without rotation; four approximately orthogonal faces are a familiar arrangement, while other systems use three or combine electronic scanning with mechanical movement. Mechanical rotation is not necessarily an obsolete compromise. It can reduce the number of required faces or permit a larger aperture within a given cost and installation envelope, at the expense of some revisit flexibility and continuous directional access.
Digital receiving architectures can form several beams from the same recorded element data. These beams are separate weighted combinations, not independent collections of physical aperture area. They can support elevation coverage, interference rejection, or multiple simultaneous look directions, but remain subject to shared dynamic range and information limits. Transmission is more restrictive: multiple simultaneous transmit beams divide available power or require waveform and hardware arrangements that distinguish their echoes. Large out-of-beam signals may also overload front-end electronics before digital nulling can act.
Applications extend beyond tracking aircraft and ships. Space-surveillance radars use electronically controlled apertures to observe objects moving rapidly across the sky. Weather-radar research uses rapid beam scheduling to study storms whose internal structures change faster than conventional volume scans can capture. Dual-polarization measurements introduce additional calibration requirements because meteorological variables depend on small differences between polarization channels. In remote sensing, airborne and spaceborne arrays can shape illumination and receive patterns, although electronic steering must be coordinated with platform motion, waveform design, and synthetic-aperture processing rather than treated as an independent improvement.
Communications, satellite links, and radio astronomy
In communications, phased arrays concentrate transmitted energy toward intended receivers and increase receive sensitivity in selected directions. At high radio frequencies, a physically small panel can contain many elements because wavelength is short. For example, half-wavelength spacing at 30 gigahertz is about five millimetres. This supports compact directional terminals, but does not by itself guarantee a strong link: propagation loss, obstruction, atmospheric attenuation, available power, and pointing accuracy still determine performance.
Modern cellular systems combine beamforming with multiple-input multiple-output communication, or MIMO. The concepts overlap but are not identical. Beamforming shapes spatial response, whereas MIMO uses multiple spatial channels to improve reliability or transmit independent data streams. A large array may serve several users by applying different precoding weights, provided their propagation channels are sufficiently distinguishable. The usable number of streams depends on channel rank, scattering, user geometry, and estimation accuracy, not simply on the antenna count.
Obtaining channel information consumes time and signaling resources. A base station must estimate how its element signals reach users, then update its weights before those estimates become stale. Time-division duplex systems can exploit propagation reciprocity between uplink and downlink, but transmit and receive electronics require reciprocity calibration. In high-frequency links, initial access and recovery after blockage may require beam searches. Hybrid arrays add another constraint because a limited number of radio-frequency chains cannot observe every possible spatial response simultaneously.
Satellite terminals use electronic steering to follow spacecraft while remaining mechanically stationary or reducing the required mechanical motion. This is particularly useful with low-Earth-orbit constellations, whose satellites move across the sky. A terminal must coordinate tracking, handover, polarization, frequency, and permitted radiation patterns. At low elevation angles, a flat panel suffers projected-aperture and element-pattern losses, while the atmospheric path becomes longer. Some installations therefore combine electronic steering with mechanical tilt or use several panels to extend usable coverage.
Radio astronomy uses phased arrays both as primary receiving instruments and as components within larger observatories. Low-frequency aperture arrays combine signals from many antennas into station beams. Phased-array feeds place multiple receiving elements near a reflector’s focal region, allowing several beams within its field of view. Instruments such as LOFAR demonstrate the importance of digital beamforming and distributed processing. This must be distinguished from interferometry, which correlates signals between separated antennas to reconstruct spatial information. An observatory may first beamform within each station and then correlate the station outputs, using the two operations at different levels of the system.
Acoustic and optical implementations
The principles of phased arrays apply to waves beyond radio. In acoustics, microphones or underwater hydrophones can be combined to emphasize selected arrival directions. Loudspeaker and sonar-projector arrays can shape transmitted sound fields. Because sound travels much more slowly than electromagnetic radiation, relevant delays are longer for the same physical aperture. Wavelength also depends strongly on medium and frequency: a 40-kilohertz wave in room-temperature air has a wavelength of roughly 8.6 millimetres, while sound at the same frequency in water has a wavelength of about 37 millimetres.
Medical ultrasound uses electronically controlled transducer arrays to steer and focus pulses and to combine returning echoes. For a focal point at distances from the elements, relative delays are chosen to compensate differences in . On reception, the appropriate focus changes as echoes return from increasing depths, enabling dynamic receive focusing. Linear, curved, and small-footprint phased-array probes provide different fields of view and access geometries. Image quality also depends on acoustic attenuation, scattering, pulse bandwidth, and assumptions about tissue sound speed.
Industrial phased-array ultrasonic testing uses related methods to inspect welds, rails, and manufactured components. A probe can sweep a range of refracted angles through a test piece without being mechanically repositioned for every beam direction. Electronic focusing improves sensitivity in selected regions, while encoded scanning links measurements to position. Interpretation requires accounting for the wedge, material sound velocities, refraction, mode conversion, and component geometry. In full-matrix-capture methods, signals are recorded for many transmitter–receiver combinations and subsequently focused computationally; this is related to, but not identical with, conventional real-time phased-array beam steering.
An optical phased array controls the phases of coherent optical emitters to steer or shape light. Integrated implementations may distribute laser light through optical waveguides and adjust phase using thermal or electro-optic effects. Potential uses include free-space communications and lidar. The short optical wavelength permits narrow beams from small apertures, but also makes phase stability demanding. Avoiding unwanted diffraction orders over wide angles can require extremely close emitter spacing, while practical waveguide routing and coupling constrain the layout.
Optical and acoustic systems share mathematical array factors with radio arrays, but their engineering limits differ. Optical devices must address fabrication-scale phase errors, insertion loss, thermal crosstalk, and coherent-light distribution. Acoustic arrays contend with reverberation, medium inhomogeneity, transducer ringing, and sometimes nonlinear propagation. Across these implementations, the useful pattern emerges from the entire wave system: source coherence, element behavior, propagation, geometry, control, and measurement must remain consistent with the intended beam.