Tiled Hybrid Beamforming Architecture

• Tiled hybrid beamforming architecture is a design that employs beam broadening techniques to counteract beam squint in wideband, sub-THz MIMO systems.
• Quadratic-phase profiles and partitioned subarray methods enable robust mainlobe synthesis, ensuring near-optimal spectral efficiency and multiuser support.
• Movable tile configurations and phase-only designs balance hardware constraints and power budgets, facilitating dual radar-communication functionalities.

Beam broadening approaches refer to a class of methods in large-scale antenna array systems designed to deliberately increase the angular width of the main beam—i.e., to synthesize a beam pattern with increased spatial/frequency coverage—relative to standard, narrow pencil beams. This is a critical technique in the context of hybrid analog/digital beamforming, especially for wideband and sub-THz massive MIMO architectures, where hardware constraints preclude fully digital per-antenna control and wide instantaneous bandwidths result in deleterious frequency dependence ("beam squint") in analog beam patterns. Beam broadening enables robust, frequency-insensitive transmission and reception by mitigating gain loss across the operational bandwidth and supporting multiuser access or radar/communication coexistence under hardware power and cost constraints.

1. Physical and System-Level Motivations for Beam Broadening

Beam broadening arises as a necessity in wideband phased arrays and tiled hybrid beamforming architectures. In such systems—exemplified by sub-THz massive MIMO—non-negligible fractional bandwidth (e.g., 20% or more) causes the spatial frequency associated with a fixed physical direction to become frequency-dependent: $$\Omega_k(f) = \frac{f}{f_c} \Omega_{c,k}, \quad \Omega_{c,k} = \pi \sin \theta_k$$ A beamformer synthesizing a narrow mainlobe at the center frequency becomes misaligned for subcarriers away from $f_c$, incurring "beam squint": an angular mispointing and corresponding gain loss that increases around edges of the bandwidth and at large steering angles. This effect is pronounced in analog-only beamformers and hybrid architectures where each tile/subarray applies a per-tile phase shift vector identical across the entire band. Beam broadening approaches—synthesizing beams with larger angular support—counteract this frequency misalignment, trading spatial selectivity for robust gain across the band and, in multi-user MIMO, enabling tile sharing between streams/users (Haddad et al., 6 Dec 2025).

2. Quadratic-Phase and Partitioned Broad Beam Synthesis

The canonical beam broadening approach is the application of a quadratic phase profile across the elements of a tile/subarray: $$\phi[n] = \Omega_c n + \frac{\Delta\Omega}{2 N_a} n^2$$ where $n$ indexes the element within the tile of size $N_a$, $\Omega_c$ is the center spatial frequency (matched to the desired steering direction at carrier), and $\Delta\Omega = |\Omega_c| (B/f_c)$ is the instantaneous spatial frequency shift over the full bandwidth $B$. This yields a broad RF beam with its mainlobe spanning $\Omega_c \pm \Delta\Omega/2$—precisely overlaying the spread of spatial frequencies that arise at fixed steering angle due to beam squint (Haddad et al., 6 Dec 2025).

Alternative strategies include partitioning tiles into sub-bands, each engineered to cover a fraction of the spatial/frequency domain with a broad mainlobe, or synthesizing a tile beam as a phase-only approximation to the frequency-averaged dominant eigenmode of $$\frac{1}{B}\int \mathbf{h}_{mk}(f)\mathbf{h}_{mk}^H(f)\,df$$. The selection among these strategies is determined by the trade-off between analog hardware simplicity (phase-only control), the requirement for amplitude or true time-delay elements (which add complexity and power overhead), and the need to support multiple spatial directions or multiuser access.

3. Impact on Spectral Efficiency and Multiuser Support

Beam broadening has a direct impact on user SINR and system spectral efficiency in wideband MU-MIMO regimes. When user directions are moderately separated in angle, disjoint tile allocation with per-subcarrier digital combining (e.g., LMMSE) and phase-only broad RF beams suffices to achieve near-optimal spectral efficiency: for a typical scenario (e.g., $N=256$, $N_a=32$, $N_d=8$), performance under quadratic-phase beam broadening closely approaches the information-theoretic benchmark of an unconstrained, fully digital MIMO-OFDM receiver, recovering nearly 80–95% of ideal sum rate even at challenging steering angles ($\theta=55^\circ$) (Haddad et al., 6 Dec 2025).

When user angles are highly clustered (interference-limited), full multiuser digital processing per subcarrier becomes essential, and the advantage of broadening (versus narrowband focusing) increases, with digital combining yielding 30–50% rate gain over RF-only decoding. However, tile-level beam broadening alone is generally sufficient to mitigate most of the destructive effects of beam squint on individual user channels.

4. Implementation and Design Trade-Offs

Beam broadening techniques are tightly coupled to hardware and power constraints in scalable tiled hybrid architectures. The phase-only, quadratic broad beam can be realized entirely with a bank of $N_a$ analog phase shifters per tile, without requiring amplitude control or high-complexity true time-delay elements. Typical configurations (e.g., $N_a\approx16$–$32$ antennas per tile, yielding 8–16 tiles for $N=256$) achieve low RF insertion loss (2–3 dB) and front-end power budgets ($\approx$8–11 W for 256 elements) compatible with fixed wireless or small cell deployments, while retaining manageable system complexity (Haddad et al., 6 Dec 2025).

For reference, the per-element cost is dominantly that of the LNA and phase shifter ($\sim$20 mW), while each tile's mixer, downconverter, and ADC add approximately 400 mW to system power. The total analog front-end power is minimized for larger tiles, but this reduces spatial degrees of freedom and may worsen beam squint if not compensated by broadening (Haddad et al., 6 Dec 2025).

A comparison of disjoint tile allocation (phase-only broad beams per user) versus full tile sharing (multiple beams per tile, amplitude+phase control) indicates that disjoint allocation with digital combining almost always matches or outperforms full sharing in both sum-rate and worst-user rate, except in rare scenarios with extremely well-separated user angles. Full sharing increases hardware complexity and is generally not justified by modest rate gains.

5. Extensions: Movable Tiles and True-Time-Delay Emulation

A distinct approach to beam broadening—targeted at next-generation near-field wideband arrays—introduces slow, mechanical reallocations of tile position, where only the centers of subarrays are moved within mechanical constraints to emulate the broadband "focusing" effect of true time delay (TTD) hardware, but at substantially reduced calibration and complexity burdens (Zhang et al., 11 Nov 2025). This "movable antenna–aided" hierarchical sub-connected hybrid beamforming ("MA-HSC-HBF"—editor's term) architecture leverages tile center translations to spatially realign the array's steering vectors across subcarriers, mitigating beam squint by matching array geometry to frequency dispersion. Simulations show that the MA-HSC-HBF can match or outperform TTD-equipped fixed arrays, achieving $+2$–$5\%$ higher sum rate at $B=30$ GHz bandwidths, and up to $+144\%$ improvement over conventional fixed arrays for specific large-N scenarios (Zhang et al., 11 Nov 2025). The architecture requires only a mechanical actuator per tile and preserves low per-tile wiring and computational overhead.

6. Applications in Radar and Joint Radar-Communication Systems

In wideband massive MIMO radar, beam broadening supports scalable beamspace processing—reducing system dimensionality by windowing the 2D spatial-DFT outputs per tile to cover the mainlobe expected over the frequency band (Noroozi et al., 6 Dec 2025). Tiles apply angle-of-arrival dependent beamspace windows, ensuring that each target's mainlobe is retained across all relevant subbands/beams, and concatenating the windowed signals enables centralized MVDR processing at drastically reduced complexity, computation, and training requirement, with no loss in detection SINR or null depth compared to full-dimensional synthesis (Noroozi et al., 6 Dec 2025).

Dual-function radar-communication (DFRC) systems also leverage reconfigurable/tiled subarray architectures for joint optimization of communication sum-rate and prescribed radar SCNR; here, the sparsity and isolation of tiled beamformers allow beam broadening to achieve both robust wideband coverage and deep clutter nulls in the radar beampattern (Jin et al., 24 Apr 2024).

7. Summary Table: Main Beam Broadening Approaches

Approach	Physical Mechanism	Key Advantages
Quadratic-phase broad beam	Phase-only, tile-local, quadratic phase profile	Suppresses beam squint, low hardware cost
Partitioned broad beams	Tiles cover distinct spatial bands	Multiuser flexibility, per-subcarrier assignment
Frequency-averaged eigenmode	Approximate dominant eig. over band	Best average gain, higher complexity
Movable tile centers	Slow mechanical reconfiguration	Emulates TTD, near-optimal broadband focusing

Beam broadening design provides a critical lever for reconciling hardware efficiency, computational tractability, and reliability in scalable wideband antenna arrays. The quadratic-phase approach emerges as a practical solution for robust sub-THz hybrid beamforming; physical reconfiguration (movable tiles) and dynamic subarray assignability enable further improvements, particularly at the system scale demanded by next-generation MIMO, radar, and dual-function networks (Haddad et al., 6 Dec 2025, Zhang et al., 11 Nov 2025, Noroozi et al., 6 Dec 2025, Jin et al., 24 Apr 2024).