Representing Antenna Fields with Plane-Wave Expansions
A common representation of antenna fields is in terms of plane wave expansions. In AntennaFieldRepresentations.jl, plane-wave expansions are represented by a struct PlaneWaveExpansion{P, S, C} which is a subtype of AntennaFieldRepresentation{P, C}. The type parameters have the following meaning
| Parameter | Short Description |
|---|---|
P <: PropagationType | Can be Radiated, Absorbed, or Incident |
C <: Complex | Element type of the coefficient vector |
S <: SphereSamplingStrategy, C} | Struct defining the sampling of the $k$-space unit sphere. Can be RegularθRegularϕSampling or GaussLegendreθRegularϕSampling. |
A PlaneWaveExpansion of Radiated type corresponds to a far field pattern $\bm{F}(\vartheta, \varphi)$ and a PlaneWaveExpansion of Incident type corresponds to a plane wave spectrum $\bm{P}(\vartheta, \varphi)$ as described in more detail in the theory section.
Constructors for a PlaneWaveExpansion
To generate a PlaneWaveExpansion, use the following constructor:
PlaneWaveExpansion(P::PropagationType, samplingstrategy::S, Eθ::Matrix{C}, Eϕ::Matrix{C}, wavenumber::Number) where{S <: SphereSamplingStrategy, C}The input arguments for the constructor are
P <: PropagationType: Can beRadiated,Absorbed, orIncidentsamplingstrategy::SphereSamplingStrategy: Struct defining the sampling of the $k$-space unit sphere. Can beRegularθRegularϕSamplingorGaussLegendreθRegularϕSampling.Eθ::Matrix{C}: θ-component amplitudes of the plane waves. Dimensions must match thesamplingstrategy.Eϕ::Matrix{C}: ϕ-component amplitudes of the plane waves. Dimensions must match thesamplingstrategy.wavenumber: Wavenumber $\omega = 2\pi \, f$
Stored Samples of a PlaneWaveRepresentation
In a PlaneWaveExpansion, an antenna far field pattern $\bm{F}(\vartheta, \varphi)$ or a plane wave spectrum $\bm{P}(\vartheta, \varphi)$ is representated by a discrete set of samples in the $\vartheta, \varphi$-domain [1] .
The samples should be chosen such that
- integration over the sphere is fast and accurate
- local interpolation is sufficently accurate
- the conversion $\bm{F}(\hat{\bm{k}}) \rightarrow \bm{F}(-\hat{\bm{k}})$ has small computational cost (ideally should not rely on interpolation)
- redundancy in the representation is kept minimal
By default, any far-field pattern or plane-wave spectrum with a corresponding spherical mode order of $L$ is sampled according to a GaussLegendreθRegularϕSampling with
\[N_\vartheta=L+1\]
and
\[N_\varphi=2L+2 \,.\]
Although the $\varphi$-sampling is slightly redundant ($2L+1$ samples would suffice) the even number of $\varphi$-samples ensures that for every sampling point there is another sampling point exactly in the oppsoite direction. This has obvious benefits for the conversion $\bm{F}(\hat{\bm{k}}) \rightarrow \bm{F}(-\hat{\bm{k}})$ as well as for pattern interpolation, when the interpolation scheme requires samples at either side of the coordinate poles at $\vartheta\in \{0,\pi\}$.
PlaneWaveExpansions can be sampled according to any arbitrary SphereSamplingStrategy.
Methods Special to PlaneWaveExpansions
In addition to the methods defined in the interface of AntennaFieldRepresentation, a PlaneWaveExpansion supports the following methods:
References
This should not lead to any ambiguities because the relation between $\hat{\bm{k}}$ and the tuple $\vartheta, \varphi$ is one-to one, as each encodings uniquely define the same point on the unit sphere.
- 1Since the propagation vector $\bm{k}$ and the corresponding unit vector into the same direction $\hat{\bm{k}}$ encode the propagation direction of the plane wave, we may use the shorthand notation $\bm{P}(\hat{\bm{k}})$ to represent the slightly longer expression $\bm{P}(\vartheta, \varphi)$ whenever convenient (sometimes we may even mix $\vartheta, \varphi$ with $\hat{\bm{k}}$ in the same expression).