Connected Array
How Continuous Currents Unlock Decade Bandwidth
The starting point is Harold Wheeler's 1965 current-sheet model, which showed that an infinite, uniformly excited sheet of x-directed current radiates a frequency-independent broadside impedance of 377 Ω per square in free space. Real arrays approximate this only over a narrow band because each finite dipole behaves as an isolated resonator: its reactance swings violently away from resonance, the array goes badly mismatched, and bandwidth collapses to perhaps 20 to 40 percent. A connected array attacks the problem directly by bonding the tip of one dipole arm to the tip of its neighbor, so the lattice carries a continuous, slowly varying current distribution instead of a set of independent standing waves.
That galvanic connection adds a strong inductive coupling path between cells. When it is combined with the capacitance of the dipole gaps and the inductance of the ground-plane spacing, the unit cell forms a broadband matching network embedded in the radiator itself. The low-frequency limit is set not by element resonance but by how electrically close the ground plane sits and how much loss the designer accepts in the wideband balun. The high-frequency limit is set by the onset of grating lobes, which forces the lattice pitch below half a wavelength at the top of the band. Between those limits the active impedance stays remarkably flat, which is why connected apertures routinely reach a decade of bandwidth.
The Common-Mode Resonance Problem
The price of connecting everything together is a parasitic even-mode current that the desired differential mode does not control. This common mode resonates when the vertical feed transition or a dielectric superstrate approaches a quarter wavelength, producing a sharp impedance spike and a scan-blindness null inside the band. Practical designs shorten the feed transition, add shorting vias or a thin resistive frequency-selective surface to spoil the common-mode Q, and select baluns that present high common-mode impedance. Managing this resonance is usually the difference between a 4:1 and a 10:1 working array.
Governing Relationships
Zcs = η0 ≈ 377 Ω/□ (free space, per square)
Grating-lobe-free element spacing:
d < λmin / (1 + sin θscan)
Active impedance under scan (E-plane / H-plane):
ZE(θ) = Z0 × cos θ ZH(θ) = Z0 / cos θ
Where η0 = free-space wave impedance, λmin = wavelength at the highest frequency, θscan = scan angle off broadside, Z0 = broadside active impedance. Example: scanning to θscan = 60° requires d < 0.54 λmin, so an X-band array uses a 12 to 15 mm lattice.
Wideband Aperture Comparison
| Architecture | Coupling Mechanism | Bandwidth | Scan Volume | Profile (above gnd) | Best Application |
|---|---|---|---|---|---|
| Connected array | Galvanic dipole-tip bond | 5:1 to 10:1 | ±60° | ≈ 0.1 λlow | Shared multifunction RF panels |
| Tightly coupled dipole array | Capacitive gap coupling | 5:1 to 7:1 | ±60° | ≈ 0.12 λlow | Wideband EW / SIGINT apertures |
| Vivaldi (tapered slot) array | Traveling-wave flare | 10:1+ | ±45° | Tall (several λ) | UWB radar, deep-band sensing |
| Conventional patch array | Weak (isolated elements) | 5 to 15% | ±45° | ≈ 0.03 λ | Single-band comms / radar |
| Stacked-patch array | Aperture / proximity | 20 to 40% | ±50° | ≈ 0.07 λ | Multiband SATCOM, 5G |
Frequently Asked Questions
What is the difference between a connected array and a tightly coupled dipole array?
Both exploit strong interelement coupling for decade bandwidth, but a connected array uses a galvanic (ohmic) bond between adjacent dipole arms so current flows continuously across cell boundaries, while a tightly coupled dipole array uses capacitive coupling across small tip gaps to cancel ground-plane inductance. Connected arrays give the smoothest active impedance and avoid gap-capacitor tuning sensitivity but demand a balanced feed at every node; both routinely reach 5:1 to 10:1 bandwidth.
Why does element spacing have to stay below half a wavelength?
Grating lobes appear when d exceeds λ/(1 + sin θscan). Scanning to 60° forces d < 0.54 λ at the highest frequency, so the lattice is set by the shortest wavelength. At the low end of a 10:1 band the same lattice is only 0.05 to 0.1 λ across, which is exactly why continuous-current-sheet behavior is needed to hold the active reflection coefficient low. X-band connected arrays typically use a 12 to 15 mm pitch.
How is the common-mode resonance that limits scan volume suppressed?
The galvanic connections and feed create an even-mode path that resonates near a quarter-wave feed or superstrate length, causing an impedance spike and scan-blindness null. Designers shorten the vertical feed transition, add shorting vias or a thin resistive frequency-selective surface to spoil the common-mode Q, and pick a balun with high common-mode impedance, holding active VSWR below 2.5:1 while scanning to ±60° in both principal planes.