The architecture of a radar array — how elements are grouped, fed, and combined — determines the system’s beamforming flexibility, sidelobe performance, bandwidth, cost, and reliability. While the element-level technology (T/R modules, radiators) captures much attention, the array-level architecture is equally critical to system performance. This article examines the major array architecture families and their respective trade-offs.

Corporate Feed Networks

The corporate (parallel) feed network distributes the transmit signal from a single input to all elements through a tree of power dividers, and combines received signals through the reverse path. The network is typically implemented in microstrip, stripline, or waveguide. Corporate feeds provide equal path lengths to all elements (when properly designed), ensuring broadband operation without beam squint. However, the feed network grows in complexity with array size: a 1,024-element array requires a 10-level binary combining tree, and each level introduces insertion loss.

For AESA arrays, the corporate feed distributes only the low-power waveform signal (and LO for downconversion), while power amplification occurs at the element. This dramatically reduces feed loss compared to PESA architectures where the high-power transmit signal passes through the full combining network.

Space-Fed Arrays

Space-fed arrays eliminate the physical feed network by distributing signals through free space. In a lens array, a feed horn illuminates a collection of elements on one side of the array, each element receives the signal, processes it (phase shift, amplification), and re-radiates from elements on the opposite side. In a reflectarray, the feed illuminates the array face, and elements reflect the signal with controlled phase to form the beam. Space-fed architectures are simpler and lighter than corporate feeds for very large arrays but suffer from spillover loss and limited bandwidth due to the path-length differences across the aperture.

Subarray Partitioning

In large arrays, partitioning elements into subarrays balances beamforming flexibility against hardware complexity. Each subarray combines a group of elements (typically 4–64) behind a single digital channel, reducing the number of ADCs and processing channels by the subarray size. The analog beamforming within each subarray provides gain and coarse pointing, while digital beamforming across subarrays provides fine pointing, multiple simultaneous beams, and adaptive nulling.

The subarray configuration directly impacts array performance. Contiguous (non-overlapping) subarrays are simplest to implement but produce quantization lobes in the array factor due to the periodic subarray structure. Overlapped subarrays, where elements contribute to multiple subarrays through weighted combining, suppress these quantization lobes at the cost of increased combining network complexity. Randomized or aperiodic subarray layouts break up periodicity and spread quantization energy into a broader sidelobe floor rather than concentrated lobes.

Constrained vs. Space-Fed Beamforming

The choice between constrained and space-fed beamforming involves fundamental system-level trade-offs. Constrained (corporate feed) networks offer full control over amplitude and phase at every element, enabling low-sidelobe patterns with precisely controlled tapers. They support independent transmit and receive patterns (different tapers for transmit efficiency and receive sidelobe control) and are inherently broadband if path lengths are equalized.

Space-fed arrays trade control for simplicity. The amplitude distribution across the aperture is determined by the feed horn pattern, which is typically a fixed, approximately Gaussian taper with limited adjustability. Receive-only space-fed arrays can use digital post-processing to synthesize low-sidelobe patterns, but transmit patterns are limited to the feed horn illumination. For many applications — particularly satellite communications and long-range surveillance — the simplicity and cost advantages of space-fed architectures outweigh their reduced flexibility.