Antenna Arrays and Pattern Synthesis

Introduction and Context

A single antenna element rarely provides the directivity, scanning capability, sidelobe control, or adaptive interference rejection required in modern wireless and radar systems. An antenna array overcomes these limitations by combining several radiating elements with controlled spatial separation and excitation. The resulting radiation pattern is determined not only by the pattern of each element but also by the relative positions, amplitudes, and phases of the elements.

Array theory is therefore central to phased-array radar, satellite communication, massive-MIMO base stations, radio astronomy, direction finding, electronic warfare, and millimetre-wave sensing. At postgraduate level, the subject requires both physical intuition and mathematical control of the array factor, beamforming, sidelobe mechanisms, and synthesis constraints.

Learning Objectives

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

1. Explain the distinction between element pattern, array factor, and total array pattern.
2. Derive the array factor for uniform and non-uniform linear arrays.
3. Calculate beam direction, beamwidth, null locations, directivity, and sidelobe levels.
4. Distinguish broadside, end-fire, scanned, binomial, and Dolph–Chebyshev arrays.
5. Apply Fourier-transform concepts to pattern synthesis.
6. Evaluate practical effects such as mutual coupling, phase errors, scan blindness, and finite resolution.
7. Formulate array design problems using numerical optimization and adaptive beamforming.

1. Fundamental Array Model

Consider N identical elements arranged along the z-axis with spacing d. In the far field, the path difference between adjacent elements in direction θ is d cos θ. The corresponding phase difference caused by propagation is −kd cos θ, where

Here k is the free-space wavenumber in radians per metre, λ is wavelength in metres, and θ is measured from the array axis. If the progressive excitation phase of one element relative to the preceding element is β, the total progressive phase difference is

The sign convention depends on the assumed time dependence and element ordering; consistency is essential. With element excitations of equal amplitude, the array factor is

For ψ ≠ 0 modulo 2π, the geometric-series result is

The magnitude, which determines the normalized shape, is therefore

The factor e^{j(N−1)ψ/2} changes phase but not magnitude. At ψ = 0, the limiting value is N, corresponding to coherent addition of all elements.

For an arbitrary array, the total far-field pattern is approximately

This pattern multiplication principle assumes identical, similarly oriented elements and neglects or separately incorporates mutual coupling. In practical arrays, the embedded-element pattern is often more accurate than the isolated-element pattern.

Array Geometry and Excitation

An array is specified by:

• element locations r_n;
• complex excitation weights w_n = a_n e^{jβ_n};
• element orientation and polarization;
• operating frequency and bandwidth;
• scan direction and required sidelobe level.

For a general three-dimensional array, the array factor can be expressed as

where r_n is the position vector of element n and u is a unit vector pointing toward the observation direction. The dot product has units of length, so kr_n · u is dimensionless, as required for a phase.

Figure: Linear-array geometry. The observation angle produces a propagation phase difference between adjacent elements.

2. Uniform Linear Arrays

A uniform linear array (ULA) has equal spacing and equal-amplitude excitation. Its principal-beam direction is obtained from the condition

or, more generally, ψ = 2mπ for integer m. Thus

provided the right-hand side lies between −1 and +1. If β = 0, the main beam is broadside to the array axis, at θ_0 = 90°. If β = −kd, the main beam points toward θ_0 = 0°, producing an end-fire condition for this convention.

Nulls and Sidelobes

Nulls occur when the numerator vanishes but the denominator does not:

Hence

The first-null beamwidth (FNBW) is the angular separation between the first nulls around the main beam. For a broadside ULA with d ≤ λ/2 and sufficiently large N, the approximation in direction cosine u = cos θ is

Near broadside, Δθ_FNBW in radians is approximately 4/N. The half-power beamwidth (HPBW) is approximately

for a uniformly excited, broadside or moderately scanned array, with angles measured in radians and aperture length approximated by Nd. A more precise finite-array calculation should solve the −3 dB condition numerically. The first sidelobe of a uniformly excited large linear array is approximately −13.26 dB relative to the main beam.

Grating Lobes

Grating lobes are unwanted beams of nearly main-beam strength. They occur when a second integer m satisfies

For wide-angle scanning, a common design rule is d ≤ λ/2. This prevents visible grating lobes over the full visible region for ordinary broadside operation and substantially limits their occurrence during scanning. For a specified scan range, larger spacing may sometimes be used, but it requires a visible-region analysis rather than reliance on a simple rule.

3. Broadside, End-Fire, and Scanned Arrays

A broadside array has its main beam perpendicular to the line of elements. For a ULA along z, this corresponds to θ_0 = 90° and β = 0.

An end-fire array has its main beam along the array axis. The ideal Hansen–Woodyard end-fire condition uses a phase progression slightly different from the ordinary end-fire condition, approximately

for one end-fire direction under a particular sign convention. This improves directivity for finite arrays but is narrowband and sensitive to errors. The approximation assumes a linear array, identical elements, and an electrically reasonable aperture.

Electronic scanning is achieved by changing β without mechanically moving the array. If the desired direction is θ_s,

under the convention used above. A phase-shifter network offers fast scanning but introduces quantization error and frequency-dependent beam displacement. True-time-delay networks are preferable for wideband systems because their delay, rather than phase, remains constant with frequency.

4. Amplitude Tapering and Sidelobe Control

Uniform excitation maximizes aperture efficiency for a fixed physical aperture but produces relatively high sidelobes. Amplitude tapering reduces sidelobes by giving greater weight to central elements and lower weight to edge elements. The trade-off is increased beamwidth and reduced directivity.

Important tapers include:

Binomial Arrays

For an N-element binomial array, excitation coefficients are proportional to the binomial coefficients in

For example, a four-element array has relative amplitudes 1:3:3:1. In the ideal theoretical model, sidelobes are eliminated, but the main lobe becomes substantially wider than that of a uniform array. Practical dynamic-range limitations also make extreme tapers difficult to realize.

Dolph–Chebyshev Arrays

Dolph–Chebyshev synthesis uses Chebyshev polynomials to produce equal sidelobes with the narrowest main beam for a specified sidelobe ratio, within the assumptions of a linear array and prescribed symmetry. If R is the positive voltage sidelobe ratio, the sidelobe level in decibels is

For example, a sidelobe level of −30 dB corresponds to R = 10^(30/20) ≈ 31.62. The design then determines the required element weights from a Chebyshev polynomial of suitable order. Tables or numerical routines are normally used for arrays larger than a few elements.

5. Pattern Synthesis as a Fourier Problem

For a broadside linear array, define the direction variable

and let the element positions be z_n. The array factor becomes

for equally spaced elements. This is a discrete Fourier transform (DFT)-like expression. Consequently, the aperture excitation distribution and the far-field pattern form a Fourier-transform pair. Concentrating excitation toward the centre of the aperture smooths the aperture distribution and suppresses sidelobes, while abrupt truncation produces stronger sidelobes.

For a continuous aperture with distribution A(z),

where z is the aperture coordinate and u is the spatial-frequency variable. The finite array is a sampled version of this concept. This connection explains why array synthesis can be treated as approximation, interpolation, convex optimization, or filter design in the spatial-frequency domain.

Common Synthesis Specifications

A synthesis problem may prescribe:

• main-beam direction;
• maximum sidelobe level over one or more angular sectors;
• nulls at interference directions;
• beamwidth or directivity;
• low cross-polarization;
• constraints on amplitude and phase range;
• robustness against manufacturing and calibration errors.

The weights are selected to satisfy these requirements. Exact satisfaction is often impossible because beamwidth, sidelobe level, efficiency, dynamic range, and robustness compete with one another.

6. Adaptive Beamforming and Null Steering

In a receiving array, the complex output is

where x(t) is the vector of element signals, w is the complex weight vector, ᴴ denotes conjugate transpose, and y(t) is the combined output. The array response vector a(θ) contains the phase and element-response information for a signal from direction θ.

A distortionless response toward θ₀ requires

A null toward an interferer at θ_i requires

The linearly constrained minimum-variance (LCMV) formulation minimizes output power while enforcing constraints:

Here R = E[x xᴴ] is the spatial covariance matrix, C contains constraint steering vectors, and f contains desired responses. The solution, when R is nonsingular, is

This result assumes accurate covariance estimation and feasible, linearly independent constraints. In practice, diagonal loading, robust uncertainty sets, and covariance estimation techniques are used because steering-vector errors can cause self-nulling or degraded interference suppression.

7. Worked Example: Four-Element Broadside Array

Consider four isotropic elements arranged along z with d = λ/2, equal amplitudes, and β = 0.

For the main beam, ψ = 0, so θ₀ = 90°.

The null condition is ψ = 2mπ/4 = mπ/2. The first nulls correspond to ψ = ±π/2. Since ψ = π cos θ,

Thus the nulls are at θ = 60° and 120°. The approximate first-null beamwidth is therefore 60°, which agrees with the exact result for this small array. The broadside response is 4 in normalized field units, or 12.04 dB relative to a single isotropic element in field-amplitude terms. Power ratios must instead use 10 log₁₀, not 20 log₁₀.

The first sidelobe is not exactly −13.26 dB for N = 4 because that value is an asymptotic large-array result. For small arrays, sidelobe levels should be obtained directly from the array-factor expression or numerical evaluation.

8. Practical Non-Idealities

Mutual Coupling

The current on one element changes the currents on neighbouring elements. Therefore, the actual element excitation is not necessarily the intended excitation. Coupling changes impedance, embedded patterns, scan performance, and sidelobe levels. Full-wave electromagnetic simulation, measured coupling matrices, or calibration measurements may be required.

Phase and Amplitude Errors

Random phase errors raise the average sidelobe floor and reduce coherent main-beam gain. Systematic phase errors can steer the beam or create asymmetric patterns. Amplitude quantization produces pattern distortion, while phase-shifter quantization introduces pointing error and quantized sidelobes.

Scan Loss and Scan Blindness

As a beam scans away from broadside, projected aperture decreases and element patterns usually weaken. Scan loss depends on element type, spacing, polarization, and coupling. Scan blindness occurs when the array interacts with surface waves or resonant structures so that radiation efficiency collapses at particular scan angles and frequencies.

Bandwidth and Calibration

A phase shift corresponding to β at frequency f does not represent the same time delay at another frequency. This causes beam squint in wideband arrays. Calibration must account for amplitude, phase, cable delay, temperature, receiver gain, oscillator phase, and mutual coupling. Over-the-air calibration is often more representative than conducted calibration alone.

9. Applied Case: Electronically Scanned Weather Radar Array

Background

Consider an illustrative active electronically scanned array (AESA) for weather radar. The system must scan ±45° in azimuth, maintain a sidelobe level below −25 dB in a protected angular sector, and operate over a moderate bandwidth.

Problem

A uniformly excited array would provide a narrow beam but would have sidelobes near −13 dB. These sidelobes could mask weak precipitation returns or permit strong ground clutter to enter the receiver.

Analysis and Approach

The design uses half-wavelength element spacing to reduce grating-lobe risk during scanning. A Taylor taper is selected to target approximately −25 dB sidelobes without the severe beam broadening of a binomial taper. Digital beamforming is used on receive, allowing sector-specific nulls and adaptive suppression of clutter. During calibration, embedded element patterns and mutual coupling are measured. Beam steering is implemented with phase control for narrowband modes and true-time-delay compensation for wideband modes.

The covariance matrix is estimated over multiple pulses, with diagonal loading applied to improve robustness when the number of snapshots is limited. Protection constraints are imposed in known clutter directions, while a distortionless constraint preserves the weather-return direction.

Outcome

The tapered array achieves lower sidelobes and improved clutter rejection, but its beamwidth is wider and its peak gain is lower than that of the uniform array. Adaptive processing provides additional suppression but depends on reliable covariance estimates. The final design is therefore a system-level compromise among angular resolution, detection sensitivity, scan range, calibration effort, and computational cost.

Learning Points

1. No taper simultaneously maximizes gain, minimizes beamwidth, and eliminates sidelobes.
2. Array-factor calculations are necessary but insufficient when coupling and element patterns are significant.
3. Adaptive nulling requires robust estimation and protection against steering errors.
4. Wideband scanning requires attention to beam squint and true time delay.

10. Computational Design and Contemporary Developments

Numerical synthesis commonly uses MATLAB, Python, or electromagnetic co-simulation. A practical workflow is: define geometry; choose an element model; specify angular masks; optimize complex weights; evaluate gain, sidelobes, null depth, and efficiency; then validate with full-wave simulation and measurement.

Convex optimization is effective when constraints can be expressed as linear equalities and magnitude inequalities. Genetic algorithms, particle-swarm optimization, and alternating minimization can address non-convex constraints, although they may require substantial computation and do not guarantee a global optimum. Machine learning can accelerate surrogate modelling, calibration, fault diagnosis, and inverse design, but training data must represent manufacturing tolerances and electromagnetic interactions. AI should support, not replace, physical validation and uncertainty analysis.

Emerging arrays include massive-MIMO panels, reconfigurable intelligent surfaces, holographic apertures, metamaterial antennas, sparse arrays, and distributed cooperative arrays. Sparse arrays reduce hardware count but create complex ambiguity and grating-lobe management problems. Metasurface and holographic approaches offer compact implementations but require careful modelling of losses, coupling, bandwidth, and manufacturability.

11. Design Best Practices and Limitations

A reliable array design should:

• define angle conventions and phase signs explicitly;
• use direction-cosine plots when comparing scanned beams;
• check visible grating lobes over the complete scan and frequency ranges;
• distinguish isolated-element from embedded-element patterns;
• specify whether sidelobe levels refer to field amplitude or power;
• include amplitude, phase, position, and mutual-coupling tolerances;
• evaluate thermal drift, calibration ageing, and failed elements;
• validate analytical results with full-wave simulation and measurement;
• report beamwidth, gain, efficiency, sidelobes, null depth, and scan loss together.

The array-factor model is powerful but idealized. It may fail when elements are strongly coupled, the array is electrically irregular, the environment is near-field, or the element patterns vary substantially across the aperture. In the near field, spherical-wave propagation must replace the plane-wave approximation, and focusing rather than simple beam steering is required.

Chapter Summary

Antenna arrays synthesize radiation through the controlled interference of multiple elements. The total pattern is commonly represented as the product of an element pattern and an array factor, subject to the assumptions of identical elements and manageable coupling. For a uniform linear array, the progressive phase ψ = kd cos θ + β determines beam direction, nulls, and grating lobes. Uniform excitation gives high aperture efficiency but relatively high sidelobes; amplitude tapers such as binomial, Taylor, and Dolph–Chebyshev designs trade beamwidth and gain against sidelobe suppression. Fourier-transform theory provides the conceptual foundation for pattern synthesis. Adaptive beamforming extends fixed synthesis by using covariance information to preserve desired signals and reject interference. Practical arrays require analysis of coupling, calibration, scan loss, beam squint, tolerances, and hardware constraints. Modern synthesis combines electromagnetic modelling, optimization, digital beamforming, and increasingly data-driven surrogate methods.

Key Terms

Array factor: Radiation contribution caused by element positions and excitations, excluding the individual element pattern.

Broadside array: Array whose main beam is perpendicular to its line of elements.

Beam squint: Frequency-dependent change in beam direction caused by phase-based steering.

Beamwidth: Angular width of a beam, commonly measured between half-power or first-null points.

Chebyshev array: Array designed for equal-ripple sidelobes with minimum beamwidth for a specified sidelobe level.

Element pattern: Radiation pattern of an individual element or embedded element.

End-fire array: Array whose main beam is directed along the array axis.

Grating lobe: Unwanted principal beam caused by excessive spacing or scanning.

Half-power beamwidth: Angular separation between points 3 dB below peak power.

LCMV beamformer: Linearly constrained minimum-variance adaptive beamformer.

Main lobe: Principal radiation beam of an array.

Mutual coupling: Electromagnetic interaction among array elements.

Null: Direction in which the ideal array response is zero or strongly suppressed.

Progressive phase: Relative phase change between adjacent array excitations.

Scan loss: Reduction in gain or efficiency as the beam is steered away from broadside.

Sidelobe: Secondary radiation maximum outside the main lobe.

Taylor taper: Amplitude distribution used to control sidelobes with moderate beam broadening.

Uniform linear array: Equally spaced elements with equal excitation amplitude and progressive phase.

Review Questions

Short Answer Questions

1. What is the physical meaning of the array factor?
2. State the condition for the main beam of a uniform linear array.
3. Why does amplitude tapering reduce sidelobe levels?
4. What is a grating lobe, and how is it related to element spacing?
5. Distinguish phase steering from true-time-delay steering.

Long Answer Questions

1. Derive the array factor of an N-element uniform linear array using the geometric-series identity.
2. Explain broadside, ordinary end-fire, and Hansen–Woodyard end-fire arrays.
3. Discuss the Fourier-transform interpretation of aperture distributions and radiation patterns.
4. Compare uniform, binomial, Taylor, and Dolph–Chebyshev excitation distributions.
5. Explain the LCMV beamformer and the practical limitations of adaptive null steering.

Analytical / Application Questions

1. Design the progressive phase required to steer a ULA with spacing λ/2 toward θ = 60° from the array axis, using ψ = kd cos θ + β.
2. A 16-element broadside ULA uses d = λ/2. Estimate its broadside HPBW using HPBW ≈ 0.886λ/(Nd), and discuss why the exact value may differ.
3. For an array required to scan ±60° over a wide frequency band, evaluate whether d = λ/2 is sufficient and recommend additional measures for controlling grating lobes, beam squint, and scan loss.

Multiple Choice Questions with Answers

Question 1

For a uniform linear array, the progressive phase is defined as ψ = kd cos θ + β. Which condition gives the principal maximum?

A. ψ = π

B. ψ = 0 modulo 2π

C. ψ = π/2 only

D. β = 0 for every scan angle

Correct Answer: B. ψ = 0 modulo 2π

Question 2

What is the approximate first sidelobe level of a large uniformly excited linear array?

A. −3 dB

B. −6 dB

C. −13.26 dB

D. −30 dB

Correct Answer: C. −13.26 dB

Question 3

For an N-element array with equal spacing, array-factor nulls occur when which condition holds?

A. ψ = 2mπ/N, with nonzero integer m

B. ψ = mπN

C. ψ = 0 only

D. ψ = π/4 independent of N

Correct Answer: A. ψ = 2mπ/N, with nonzero integer m

Question 4

Which excitation distribution ideally eliminates sidelobes in the classical theoretical model?

A. Uniform

B. Binomial

C. Dolph–Chebyshev

D. Random

Correct Answer: B. Binomial

Question 5

The usual spacing rule d ≤ λ/2 is primarily used to reduce the risk of what phenomenon during scanning?

A. Polarization mismatch

B. Grating lobes

C. Thermal noise

D. Receiver saturation

Correct Answer: B. Grating lobes

Question 6

A sidelobe level of −30 dB corresponds approximately to what voltage-amplitude ratio R, using SLL = −20 log₁₀ R?

A. 1.73

B. 10

C. 31.62

D. 300

Correct Answer: C. 31.62

Question 7

In the LCMV beamformer, the constraint wᴴa(θ₀) = 1 means that the array should:

A. Suppress the desired direction completely

B. Maintain a distortionless response in the desired direction

C. Force all sidelobes to zero

D. Use uniform amplitudes only

Correct Answer: B. Maintain a distortionless response in the desired direction

Question 8

Which effect is most directly associated with using fixed phase shifts over a wide frequency band?

A. Beam squint

B. Elimination of mutual coupling

C. Infinite bandwidth

D. Removal of scan loss

Correct Answer: A. Beam squint

Question 9

For a four-element binomial array, the relative amplitude coefficients are:

A. 1:1:1:1

B. 1:2:2:1

C. 1:3:3:1

D. 1:4:6:4

Correct Answer: C. 1:3:3:1

Question 10

What is the principal trade-off introduced by amplitude tapering?

A. Lower sidelobes at the cost of broader beamwidth and reduced peak efficiency

B. Higher sidelobes and narrower beamwidth in every case

C. Elimination of all coupling effects

D. Automatic removal of beam squint

Correct Answer: A. Lower sidelobes at the cost of broader beamwidth and reduced peak efficiency