9.1 Uniform linear arrays (ULA)

Uniform linear arrays fundamentals

A uniform linear array (ULA) arranges multiple identical antenna elements along a straight line with equal spacing between them. This simple geometry is the foundation for most array processing and beamforming techniques because it enables directional signal reception, interference rejection, and straightforward mathematical analysis.

ULAs show up across radar, sonar, wireless communications, and radio astronomy. Mastering ULA fundamentals is essential before moving to more complex array geometries or adaptive algorithms.

Definition of ULA

A ULA consists of identical antenna elements placed along a line with a fixed inter-element spacing $d$. The elements are typically omnidirectional (dipoles, microstrip patches, etc.), meaning they radiate uniformly in the plane perpendicular to the array axis. Because every element is identical and equally spaced, the math simplifies considerably, and standard beamforming techniques apply directly.

Key properties of ULAs

• Directionality: A ULA focuses its response toward a chosen direction while suppressing signals from other angles.
• Resolution vs. aliasing trade-off: The inter-element spacing $d$ controls how well the array can distinguish closely spaced signals, but spacing that's too large introduces spatial aliasing (more on this below).
• Scalable gain: Adding more elements narrows the beam and increases gain, enabling longer-range detection or transmission.

Applications in signal processing

• Radar: Electronic beam steering for target detection, tracking, and imaging without physically rotating the antenna.
• Wireless communications: Beamforming to boost signal quality, reject interference, and support multiple users simultaneously.
• Sonar: Underwater acoustic imaging and target localization using arrays of hydrophones.
• Radio astronomy: Large arrays that achieve the angular resolution needed to observe distant celestial objects.

Array geometry and spacing

The geometry and spacing of a ULA directly determine its angular resolution, steering range, and susceptibility to grating lobes. Getting the spacing right is one of the most important design decisions.

Element positioning in ULAs

Each element sits along a line at a distance $nd$ from the reference point (usually the first element or the array center), where $n = 0, 1, \ldots, N-1$. Performance depends not on $d$ alone but on the ratio $d/\lambda$, where $\lambda$ is the signal wavelength. This ratio ties the physical spacing to the electrical behavior of the array.

Inter-element spacing considerations

The spacing $d$ must satisfy the Nyquist spacing criterion:

$d \leq \frac{\lambda}{2}$

This prevents grating lobes (spurious main lobes in undesired directions). Spacing smaller than $\lambda/2$ allows wider steering angles without grating lobes, but elements that are too close together suffer from increased mutual coupling, where the electromagnetic fields of neighboring elements interfere with each other.

Relationship between spacing and performance

There's a fundamental trade-off here:

• Larger $d$ improves angular resolution (the ability to distinguish two closely spaced sources) but shrinks the grating-lobe-free steering range.
• Smaller $d$ gives a wider usable steering range but coarser resolution.

Most practical designs start at $d = \lambda/2$ and adjust from there based on the application's resolution and steering requirements.

Array steering and beamforming

Steering and beamforming let you electronically point the array's main beam in any desired direction, no mechanical rotation needed. You do this by applying phase shifts (or time delays) across the elements so their signals add constructively in the target direction.

Principles of array steering

A signal arriving from angle $\theta$ reaches each successive element with a small time delay. To steer the beam toward $\theta$, you apply a progressive phase shift across the elements that compensates for these arrival-time differences. The result is constructive interference in the desired direction and destructive interference elsewhere. Because the phase shifts are electronic, beam steering can happen in microseconds.

Beamforming techniques for ULAs

• Delay-and-sum (conventional) beamforming: Apply time delays (or equivalent phase shifts) to align the signals from all elements, then sum them. Simple and robust.
• Adaptive beamforming: Dynamically adjust complex weights on each element to maximize signal-to-interference-plus-noise ratio (SINR). Examples include the Capon (MVDR) beamformer.
• Null steering: Place nulls in the radiation pattern at known interference directions to suppress jammers or co-channel interference.

Steering vector calculation

The steering vector encodes the phase shifts needed to point the beam at angle $\theta$. For a ULA with $N$ elements and spacing $d$:

$\mathbf{a}(\theta) = \begin{bmatrix} 1,\ e^{j\frac{2\pi}{\lambda}d\sin\theta},\ \ldots,\ e^{j\frac{2\pi}{\lambda}(N-1)d\sin\theta} \end{bmatrix}^T$

Each entry represents the phase shift at the $n$-th element relative to the reference element. The steering vector is the building block for computing the array factor, forming beams, and running DOA estimation algorithms.

Array factor and radiation pattern

The array factor (AF) and radiation pattern describe where the array sends or receives energy. They're the primary tools for analyzing ULA performance.

Definition of array factor

The array factor captures the combined effect of all element weights and positions. For a ULA with $N$ elements, spacing $d$, and complex weights $w_n$:

$AF(\theta) = \sum_{n=0}^{N-1} w_n\, e^{j\frac{2\pi}{\lambda}nd\sin\theta}$

This is essentially the discrete-space Fourier transform of the weight vector. The AF determines the locations and levels of the main lobe, side lobes, and nulls.

Calculating radiation patterns

The total radiation pattern equals the product of the array factor and the individual element pattern (the radiation characteristic of a single antenna). For isotropic elements, the element pattern is unity in all directions, so the radiation pattern equals the AF directly.

You can plot the pattern in polar coordinates (intuitive for visualizing beam shape) or in Cartesian coordinates with angle on the x-axis and magnitude in dB on the y-axis (better for reading sidelobe levels precisely).

Main lobe and side lobe characteristics

• Main lobe: The region of peak response, pointing in the steered direction. Its angular width is the beamwidth, which determines angular resolution.
• Side lobes: Smaller lobes flanking the main lobe. They represent sensitivity to signals from undesired directions.
• Side lobe level (SLL): The ratio of the highest side lobe peak to the main lobe peak, usually expressed in dB. Lower SLL means better interference rejection. Tapering the weights (e.g., using Hamming or Chebyshev windows) reduces SLL at the cost of a slightly wider main lobe.

Grating lobes and aliasing

Grating lobes are the array-processing equivalent of aliasing in time-domain sampling. They create ambiguous "copies" of the main beam that can cause false detections or corrupt DOA estimates.

Conditions for grating lobe formation

Grating lobes appear when the phase difference between adjacent elements reaches $2\pi$, producing constructive interference in an unintended direction. The general condition for grating lobe formation is:

$d > \frac{\lambda}{1 + |\sin\theta_{\max}|}$

where $\theta_{\max}$ is the maximum steering angle. For full hemispherical scanning ($\theta_{\max} = 90°$), this simplifies to $d > \lambda/2$.

Aliasing in ULAs

Spatial aliasing occurs when $d > \lambda/2$. At that point, multiple angles produce the same inter-element phase difference, making it impossible to tell them apart. This limits the unambiguous field of view of the array. The analogy to temporal sampling is direct: just as sampling a signal below the Nyquist rate creates frequency ambiguity, spacing elements too far apart creates angular ambiguity.

Techniques to avoid grating lobes

1. Nyquist spacing: Keep $d \leq \lambda/2$. This guarantees no grating lobes for any steering angle.
2. Array thinning: Selectively remove elements from a dense ULA to break the periodicity that causes grating lobes, while preserving a large aperture.
3. Non-uniform spacing: Use logarithmic, prime-number-based, or random element positions to disrupt the regular pattern that produces grating lobes.

Each alternative to Nyquist spacing introduces trade-offs (higher sidelobes, more complex calibration), so $d = \lambda/2$ remains the default starting point.

Directivity and gain

Directivity and gain quantify how well a ULA concentrates energy in the desired direction. They're closely related but not identical.

Directivity of ULAs

Directivity is the ratio of the peak radiation intensity to the average intensity over all directions. A higher directivity means a tighter beam. For a ULA with $N$ elements and spacing $d$, the approximate directivity is:

$D \approx \frac{2Nd}{\lambda}$

This tells you that directivity grows linearly with both the number of elements and the spacing (i.e., with the total aperture length $Nd$).

Gain enhancement using ULAs

Gain equals directivity multiplied by radiation efficiency (which accounts for ohmic losses, feed network losses, etc.). In the ideal lossless case, the maximum gain of an $N$-element ULA is approximately:

$G_{\max} \approx N$

This means doubling the number of elements roughly doubles the gain (a 3 dB improvement). The gain enhancement comes from coherently combining signals across elements, which boosts the desired signal while uncorrelated noise partially cancels.

Trade-offs between directivity and beamwidth

Directivity and beamwidth are inversely related:

• More elements or wider spacing increases directivity but narrows the beam, reducing angular coverage.
• A wider beam covers more angles but sacrifices directivity and the ability to reject nearby interferers.

The right balance depends on the application. A radar tracking a single target benefits from a narrow beam and high directivity. A communications base station serving a wide sector needs broader coverage and accepts lower directivity per beam.

ULA design considerations

Designing a real ULA involves balancing performance goals against practical constraints like cost, size, power, and manufacturing tolerances.

Number of elements vs. performance

Adding elements improves directivity, gain, angular resolution, and sidelobe control. But each additional element adds:

• Hardware cost (antenna, receiver chain, ADC)
• Computational load (more channels to process)
• Physical size and weight

The design process typically starts from a required beamwidth or gain specification, calculates the minimum number of elements needed, and then checks whether that number is feasible given the budget and platform constraints.

Mutual coupling effects

When elements are spaced closer than about $\lambda/2$, the current on one element induces currents on its neighbors. This mutual coupling distorts element impedances and radiation patterns, degrading beamforming accuracy.

Mitigation strategies include:

• Element isolation: Physical barriers or electromagnetic band-gap structures between elements.
• Impedance matching networks: Compensate for the impedance changes caused by coupling.
• Coupling compensation algorithms: Measure or model the coupling matrix and correct for it in software.

Practical implementation challenges

• Manufacturing tolerances: Small errors in element positions or orientations shift nulls and raise sidelobes. Tight fabrication tolerances or post-manufacturing calibration are essential.
• Feed network design: The network distributing signals to all elements must maintain equal amplitude and precise phase across a wide bandwidth. Losses and imbalances here directly degrade array performance.
• Environmental effects: Nearby structures (radomes, mounting platforms, ground planes) scatter and reflect signals, altering the effective radiation pattern.
• Calibration: Real arrays require periodic calibration to correct for component drift, temperature effects, and aging. Calibration typically involves injecting known signals and measuring the per-element response.

ULAs in DOA estimation

Direction of arrival (DOA) estimation is one of the most important applications of ULAs. The goal is to determine the angles from which signals are arriving, using the phase differences measured across the array.

Principles of DOA estimation

A signal from angle $\theta$ arrives at each element with a predictable phase shift determined by the steering vector $\mathbf{a}(\theta)$. By analyzing the received data across all elements, you can infer which angles have active sources. The core challenge is extracting these angles from noisy, potentially overlapping signals.

ULA-based DOA estimation algorithms

The main families of algorithms, in order of increasing sophistication:

1. Conventional beamforming (Bartlett): Sweep the steering vector across all angles and plot the output power. Peaks indicate source directions. Simple but has limited resolution; two sources closer than about one beamwidth blur together.

2. Capon (MVDR) beamformer: Minimizes output power subject to unity gain in the look direction. Provides sharper peaks than Bartlett but still struggles with very closely spaced sources.

3. Subspace methods (MUSIC, ESPRIT): Decompose the covariance matrix of the received data into signal and noise subspaces.

   • MUSIC searches for angles where the steering vector is orthogonal to the noise subspace, producing sharp spectral peaks.
   • ESPRIT exploits the shift-invariance structure of ULAs to estimate DOAs directly from eigenvalues, avoiding the spectral search entirely.

4. Maximum likelihood (ML) methods: Jointly estimate all source angles by maximizing the likelihood of the observed data. Statistically optimal but computationally expensive, especially with many sources.

Performance comparison of DOA techniques

Subspace methods (MUSIC, ESPRIT) are the most widely used in practice because they offer high resolution at reasonable computational cost. Performance for all methods improves with more elements, higher SNR, and more data snapshots.

Advanced ULA configurations

Standard ULAs use uniform spacing, but relaxing this constraint opens up design possibilities that can suppress grating lobes, reduce hardware, or conform to physical platforms.

Non-uniform spacing in ULAs

Non-uniform spacing places elements at unequal intervals along the array axis. This breaks the periodicity responsible for grating lobes and can also reduce mutual coupling between closely spaced pairs.

Common approaches include:

• Logarithmic spacing: Inter-element distances increase logarithmically, providing good wideband performance.
• Prime-number spacing: Element positions based on prime numbers create an aperiodic layout that suppresses grating lobes.
• Minimum redundancy arrays: Positions are chosen so that the set of pairwise element separations covers as many unique spacings as possible, maximizing the effective aperture for a given number of elements.

The trade-off is increased design complexity and the loss of the shift-invariance property that algorithms like ESPRIT rely on.

Sparse and thinned arrays

Sparse arrays use fewer active elements than a fully populated array while maintaining a large total aperture. Thinned arrays are a specific type where you start with a full ULA and selectively remove elements.

The benefits are clear: fewer elements means lower cost, less hardware, and reduced power consumption. A well-designed sparse array can match the angular resolution of a full array (since resolution depends on aperture, not element count). However, the reduced number of elements typically leads to higher sidelobe levels and lower gain compared to the fully populated version.

Conformal and curved ULAs

Conformal arrays mount elements on a curved surface (an aircraft fuselage, a ship hull, a vehicle body) rather than a flat plane. This provides several advantages:

• Reduced aerodynamic drag compared to protruding planar arrays
• Better platform integration
• The ability to steer beams over a wider angular range, including both azimuth and elevation

The non-planar geometry complicates the array manifold (the set of steering vectors is no longer a simple function of one angle), requiring more sophisticated beamforming and calibration. Element patterns also vary across the array because each element faces a different direction, so the assumption of identical elements no longer holds. Despite these challenges, conformal arrays are increasingly common in aerospace and automotive applications where platform integration is a priority.