Ferkans — Interactive Telecom Tutor

From Physics to Geometry in k-Space

The Taylor expansion of Section 6.2 showed that each measurement maps to a combined wavenumber vector $\kappa_{\mathbf{s},\mathbf{r}} = \boldsymbol{\kappa}_s + \boldsymbol{\kappa}_r$ . We now ask: what region of k-space can we cover with a given set of transmitters, receivers, and frequencies? This is the central question of imaging system design, and its answer is beautifully geometric — the Ewald sphere.

Definition:
The Ewald Sphere

For a fixed frequency $f$ (wavenumber $\kappa = 2\pi(f_0 + f)/\text{c}$ ), the transmitter wavenumber $\boldsymbol{\kappa}_s$ and receiver wavenumber $\boldsymbol{\kappa}_r$ each have magnitude $\kappa$ :

$\|\boldsymbol{\kappa}_s\| = \|\boldsymbol{\kappa}_r\| = \kappa.$

The combined wavenumber $\kappa_{\mathbf{s},\mathbf{r}} = \boldsymbol{\kappa}_s + \boldsymbol{\kappa}_r$ therefore lies within a ball of radius $2\kappa$ centered at the origin:

$\|\kappa_{\mathbf{s},\mathbf{r}}\| \leq 2\kappa.$

For a fixed transmitter direction $\hat{\mathbf{s}}$ and varying receiver direction $\hat{\mathbf{r}}$ , the locus of $\kappa_{\mathbf{s},\mathbf{r}}$ traces a sphere (or circle in 2D) of radius $\kappa$ centered at $\boldsymbol{\kappa}_s$ . This is the Ewald sphere for that transmitter direction.

In the monostatic case ( $\mathbf{s} = \mathbf{r}$ ), we have $\boldsymbol{\kappa}_s = \boldsymbol{\kappa}_r$ and $\kappa_{\mathbf{s},\mathbf{r}} = 2\boldsymbol{\kappa}_s$ , which traces a sphere of radius $2\kappa$ .

The Ewald sphere construction was originally developed by Paul Peter Ewald (1913) for X-ray crystallography. In RF imaging, it tells us exactly which spatial frequencies of the scene are accessible from a given measurement configuration.

, ,

Theorem: k-Space Sampling by a Multi-Static, Multi-Frequency System

For a system with $M$ transmitters, $N$ receivers, and $K$ frequency subcarriers, the set of sampled wavenumber points is

$\mathcal{K} = \{\boldsymbol{\kappa}_{i,j,k} = \boldsymbol{\kappa}_{i,k} + \hat{\boldsymbol{\kappa}}_{j,k} : i = 1, \ldots, M, \; j = 1, \ldots, N, \; k = 1, \ldots, K\},$

where

$\boldsymbol{\kappa}_{i,k} = \frac{2\pi(f_0 + f_k)}{\text{c}} \cdot \frac{\mathbf{p}_{0} - \mathbf{s}_{i}}{d(\mathbf{s}_{i}, \mathbf{p}_{0})}, \qquad \hat{\boldsymbol{\kappa}}_{j,k} = \frac{2\pi(f_0 + f_k)}{\text{c}} \cdot \frac{\mathbf{p}_{0} - \mathbf{r}_{j}}{d(\mathbf{p}_{0}, \mathbf{r}_{j})}.$

The set $\mathcal{K}$ has at most $MNK$ distinct points and determines the achievable imaging resolution and the structure of the point-spread function.

Each transmitter direction fixes a point on the Ewald sphere, each receiver direction traces an arc on it, and each frequency scales the sphere radius. Varying all three sweeps a region of k-space. More coverage = better image.

Proof

Direct substitution

This follows directly from the wavenumber decomposition in TTaylor Expansion and Wavenumber Decomposition. Each measurement $(i, j, k)$ contributes the wavenumber point $\boldsymbol{\kappa}_{i,j,k} = \boldsymbol{\kappa}_{i,k} + \hat{\boldsymbol{\kappa}}_{j,k}$ , which is the vector sum of the Tx and Rx wavenumber vectors at frequency $f_k$ . $\blacksquare$

Ewald Sphere and k-Space Coverage

Visualize how multi-static and multi-frequency configurations tile k-space. Each dot represents a sampled wavenumber point. Adjust the array geometry, number of antennas, and bandwidth to see how the coverage changes.

Parameters

Array configuration

Number of subcarriers (

K

)8

W

(MHz)200

f_0

(GHz)10

Example: k-Space Coverage for a Monostatic System

A monostatic system ( $\mathbf{s} = \mathbf{r}$ ) at angle $\phi$ relative to the target, operating at $K$ uniformly spaced frequencies within bandwidth $W$ centered at $f_0$ . Describe the k-space coverage.

Solution

Monostatic wavenumber

For monostatic operation, $\boldsymbol{\kappa}_s = \boldsymbol{\kappa}_r$ , so $\kappa_{\mathbf{s},\mathbf{r}} = 2\boldsymbol{\kappa}_s = 2\kappa\hat{\mathbf{s}}$ where $\hat{\mathbf{s}}$ is the unit vector from the Tx toward the target. The combined wavenumber always points in the direction $\hat{\mathbf{s}}$ with magnitude $2\kappa$ .

Frequency variation

As $f$ varies from $-W/2$ to $W/2$ , $\kappa$ varies from $2\pi(f_0 - W/2)/\text{c}$ to $2\pi(f_0 + W/2)/\text{c}$ . The sampled points form a radial line segment in k-space, in the direction $\hat{\mathbf{s}}$ , of length

$\Delta\kappa_{\text{radial}} = 2 \cdot \frac{2\pi W}{\text{c}} = \frac{4\pi W}{\text{c}}.$

Interpretation

A single monostatic node with bandwidth provides range resolution (radial extent in k-space) but no cross-range resolution (no angular extent). To image a 2D or 3D scene, we need multiple viewing angles — either by moving the antenna (SAR) or by using multiple static nodes.

How Multi-Static Configurations Fill k-Space

The key system-design insight from the Ewald sphere construction:

Bandwidth controls the radial extent of k-space coverage (range resolution).
Angular diversity (multiple Tx-Rx directions) controls the angular extent (cross-range resolution).
Carrier frequency determines the radius of the Ewald sphere — higher frequency means a larger sphere, hence finer potential resolution.
Bistatic pairs access different regions of k-space than monostatic pairs, improving coverage.

The ideal imaging system tiles k-space as uniformly and densely as possible. Gaps in coverage correspond to missing spatial frequencies, which cause artifacts (sidelobes, ambiguities) in the reconstructed image.

Common Mistake: Narrowband Assumption in the Ewald Sphere

Mistake:

Treating the Ewald sphere radius as constant across the bandwidth, i.e., ignoring the frequency dependence of $\kappa = 2\pi(f_0 + f)/\text{c}$ .

Correction:

For narrowband signals ( $W \ll f_0$ ), the sphere radius varies by less than $W/(2f_0)$ , so the approximation is valid. But for wideband systems (e.g., $W/f_0 > 0.1$ ), the frequency variation of the Ewald sphere radius is significant and helps — it adds radial extent to the k-space coverage, providing range resolution. The point is that this is a feature, not a bug: wideband signals naturally tile a thicker annular region of k-space.

Historical Note: Paul Peter Ewald and the Ewald Sphere

1913

The Ewald sphere construction was introduced by Paul Peter Ewald in 1913 to explain X-ray diffraction from crystals. Ewald showed that a diffraction spot appears when the Ewald sphere (of radius equal to the incident X-ray wavenumber) passes through a reciprocal lattice point of the crystal. This geometric condition encodes Bragg's law in a way that generalizes naturally to non-crystalline objects. When Wolf (1969) and Devaney (1982) brought diffraction tomography to scattering theory, the Ewald sphere became the unifying geometric tool for understanding how measurements sample the spatial Fourier transform of an object.

,

Ewald Sphere

In the wavenumber domain, the locus of combined Tx-Rx wavenumber vectors $\kappa_{\mathbf{s},\mathbf{r}}$ for a fixed transmitter direction and varying receiver direction, at a given frequency. It is a sphere of radius $\kappa$ in 3D (circle in 2D). The set of Ewald spheres for all Tx-Rx pairs and frequencies determines the k-space coverage of the imaging system.

Quick Check

For a monostatic system at carrier frequency $f_0 = 10$ GHz ( $\lambda = 3$ cm), the monostatic k-space points lie on a sphere of radius:

$\kappa = 2\pi/\lambda \approx 209$ rad/m

$2\kappa = 4\pi/\lambda \approx 419$ rad/m

$\kappa/2 \approx 105$ rad/m

$4\kappa \approx 838$ rad/m