Antenna Arrays: Uniform Linear Arrays

From Single Antennas to Arrays

A single antenna has a fixed radiation pattern — to steer the beam or increase directivity, one must physically rotate it or use a larger aperture. An antenna array solves both problems electronically: by combining signals from NN elements with appropriate phase shifts, one can steer the beam to any direction and sharpen it far beyond what a single element achieves. This is the foundational concept behind MIMO, massive MIMO, and 5G beamforming.

Definition:

Uniform Linear Array (ULA)

A uniform linear array (ULA) consists of NN identical antenna elements placed along a line (taken as the zz-axis) with equal inter-element spacing dd:

  • Element positions: zn=ndz_n = n d, for n=0,1,,N1n = 0, 1, \ldots, N-1
  • All elements have the same radiation pattern (the element pattern)
  • The array is characterised by NN and d/λd/\lambda

The ULA is the simplest and most widely analysed array geometry. Its properties extend naturally to planar and circular arrays (Section 7.3).

Definition:

Steering Vector

For a ULA with spacing dd, the steering vector for a plane wave arriving from direction θ\theta (measured from broadside) is

a(θ)=[1,  ejkdsinθ,  ej2kdsinθ,  ,  ej(N1)kdsinθ]TCN\mathbf{a}(\theta) = \bigl[1,\; e^{jkd\sin\theta},\; e^{j2kd\sin\theta},\; \ldots,\; e^{j(N-1)kd\sin\theta}\bigr]^T \in \mathbb{C}^N

where k=2π/λk = 2\pi/\lambda is the wavenumber. The nn-th component encodes the phase shift ejnkdsinθe^{jnkd\sin\theta} due to the extra path length ndsinθnd\sin\theta to element nn.

The steering vector is a spatial frequency concept: it maps angle θ\theta to a complex exponential sequence with frequency ψ=kdsinθ\psi = kd\sin\theta.

Definition:

Array Factor

The array factor is the beam pattern produced by the array of isotropic elements, obtained by summing the element outputs with complex weights w=[w0,w1,,wN1]T\mathbf{w} = [w_0, w_1, \ldots, w_{N-1}]^T:

AF(θ)=wHa(θ)=n=0N1wnejnkdsinθ\mathrm{AF}(\theta) = \mathbf{w}^H \mathbf{a}(\theta) = \sum_{n=0}^{N-1} w_n^* \, e^{jnkd\sin\theta}

The total radiation pattern is the product of the element pattern and the array factor:

Ftotal(θ)=Felement(θ)AF(θ)2F_{\text{total}}(\theta) = F_{\text{element}}(\theta) \cdot |\mathrm{AF}(\theta)|^2

This factorisation is the pattern multiplication principle — it holds exactly when mutual coupling is negligible.

,

Theorem: ULA Array Factor with Uniform Weights

For a ULA with NN elements, spacing dd, and uniform weights wn=1Nejnkdsinθ0w_n = \frac{1}{N} e^{-jnkd\sin\theta_0} (steering to θ0\theta_0), the array factor is

AF(θ)=1Nn=0N1ejnkd(sinθsinθ0)=1Nsin ⁣(Nψ2)sin ⁣(ψ2)\mathrm{AF}(\theta) = \frac{1}{N} \sum_{n=0}^{N-1} e^{jnkd(\sin\theta - \sin\theta_0)} = \frac{1}{N} \cdot \frac{\sin\!\bigl(\frac{N\psi}{2}\bigr)}{\sin\!\bigl(\frac{\psi}{2}\bigr)}

where ψ=kd(sinθsinθ0)\psi = kd(\sin\theta - \sin\theta_0) is the progressive phase difference. The power pattern is

AF(θ)2=1N2sin2 ⁣(Nψ2)sin2 ⁣(ψ2)|\mathrm{AF}(\theta)|^2 = \frac{1}{N^2}\,\frac{\sin^2\!\bigl(\frac{N\psi}{2}\bigr)}{\sin^2\!\bigl(\frac{\psi}{2}\bigr)}

This is the spatial analogue of the Dirichlet kernel from Fourier analysis. The array samples the incoming wavefront at NN equally spaced points, just as a DFT samples a time signal. The sin(Nψ/2)/sin(ψ/2)\sin(N\psi/2)/\sin(\psi/2) function has a main lobe of width 2/(Ndcosθ0)\sim 2/(Nd\cos\theta_0) and sidelobes at 13.3-13.3 dB — identical to the spectral leakage of a rectangular window in spectral analysis.

,

Example: 8-Element ULA Beam Pattern

An 8-element ULA with half-wavelength spacing (d=λ/2d = \lambda/2) is steered to broadside (θ0=0\theta_0 = 0).

(a) Find the 3 dB beamwidth.

(b) Find the direction of the first null.

(c) What is the first sidelobe level?

ULA Beam Steering Animation

Watch the beam pattern of a 16-element ULA sweep from 60-60^\circ to +60+60^\circ and back. Observe how the main lobe narrows at broadside and broadens at large scan angles due to the 1/cosθ01/\cos\theta_0 foreshortening effect.
Cinematic beam sweep of a 16-element ULA (d=λ/2d = \lambda/2). The beam broadens as it scans away from broadside, consistent with the HPBW formula 0.886λ/(Ndcosθ0)0.886\lambda/(Nd\cos\theta_0).

ULA Beam Pattern

Explore how the beam pattern of a ULA changes with the number of elements NN, element spacing d/λd/\lambda, and scan angle θ0\theta_0. Observe the main lobe narrowing as NN increases, grating lobes appearing when d>λ/2d > \lambda/2, and beam broadening at large scan angles.

Parameters
8
0.5
0

Beamwidth and Directivity vs Number of Elements

See how the 3 dB beamwidth decreases and the array directivity increases as the number of elements NN grows. The beamwidth scales approximately as 1/N1/N for half-wavelength spacing. The theoretical approximation HPBW0.886λ/(Nd)\text{HPBW} \approx 0.886\lambda/(Nd) is overlaid for comparison.

Parameters
0.5
64

Theorem: ULA 3 dB Beamwidth

For a ULA with NN elements, spacing dd, steered to angle θ0\theta_0, the approximate half-power beamwidth is

HPBW0.886λNdcosθ0\text{HPBW} \approx \frac{0.886\,\lambda}{Nd\cos\theta_0}

The directivity of the ULA (for d=λ/2d = \lambda/2) is approximately

DND \approx N

or in dBi: DdBi10log10ND_{\text{dBi}} \approx 10\log_{10} N.

The aperture of the array is L=(N1)dNdL = (N-1)d \approx Nd. The beamwidth is inversely proportional to the aperture measured in wavelengths, just as the time-domain resolution of a filter is inversely proportional to its bandwidth (the space-frequency duality). The cosθ0\cos\theta_0 factor accounts for the foreshortening of the projected aperture when the beam is steered away from broadside.

Common Mistake: Confusing Element Spacing with Array Aperture

Mistake:

Believing that increasing element spacing dd always narrows the beam without consequences.

Correction:

While increasing dd does increase the total aperture L=(N1)dL = (N-1)d and thus narrows the main lobe, when d>λ/2d > \lambda/2 grating lobes appear — additional main-lobe-level peaks at angles where ψ=±2π\psi = \pm 2\pi. Grating lobes satisfy

sinθg=sinθ0±mλ/d,m=1,2,\sin\theta_g = \sin\theta_0 \pm m\lambda/d, \quad m = 1, 2, \ldots

For d=λ/2d = \lambda/2, grating lobes are pushed to θg=±90\theta_g = \pm 90^\circ (invisible endfire). For d>λ/2d > \lambda/2, they enter the visible region and cause ambiguity and power loss.

Quick Check

A ULA with d=λd = \lambda is steered to broadside (θ0=0\theta_0 = 0). At what angles do grating lobes appear?

±30\pm 30^\circ

±90\pm 90^\circ

No grating lobes appear

±45\pm 45^\circ

Key Takeaway

The ULA is a spatial DFT: the steering vector a(θ)\mathbf{a}(\theta) maps physical direction θ\theta to spatial frequency ψ=kdsinθ\psi = kd\sin\theta. Everything from the DFT world carries over — the Dirichlet kernel shapes the beam, d=λ/2d = \lambda/2 is the spatial Nyquist rate, and exceeding it causes grating lobes (spatial aliasing).

Uniform Linear Array (ULA)

An array of NN identical antenna elements arranged along a line with equal spacing dd. The simplest and most fundamental array geometry.

Related: Steering Vector, Array Factor, Beamforming Architecture Comparison

Steering Vector

a(θ)CN\mathbf{a}(\theta) \in \mathbb{C}^N: the vector of phase shifts across array elements for a plane wave from direction θ\theta. The nn-th entry is ejnkdsinθe^{jnkd\sin\theta}.

Related: Uniform Linear Array (ULA), Array Factor, Spatial Frequency

Array Factor

AF(θ)=wHa(θ)\mathrm{AF}(\theta) = \mathbf{w}^H \mathbf{a}(\theta): the spatial response of an array of isotropic elements to a plane wave from direction θ\theta, determined by the weight vector w\mathbf{w}.

Related: Steering Vector, Beamforming Architecture Comparison, Pattern Multiplication

Grating Lobes

Spurious main-lobe-level peaks in the array pattern that appear when the element spacing exceeds λ/2\lambda/2. Caused by spatial aliasing, analogous to spectral aliasing in temporal sampling.

Related: Uniform Linear Array (ULA), Element Spacing, Spatial Aliasing