Hybrid Beamforming

The core architectural principle dividing precoding into digital baseband and analog RF domains.

• Digital Stage: Low-dimensional (e.g., 4-8 RF chains) performing multi-user MIMO, interference cancellation, and frequency-selective processing
• Analog Stage: High-dimensional phase-shifter network (e.g., 64-256 antennas) implementing beam-steering via constant-modulus weights
• Key Constraint: Analog precoder applies the same phase rotation across all subcarriers, unlike fully digital architectures
• Result: Dramatic reduction in RF chains—from hundreds to single digits—while preserving most array gain

Two competing analog network topologies defining the mapping between RF chains and antenna elements.

Fully-Connected Architecture:

• Each RF chain connects to all antennas via a dedicated phase shifter
• Maximizes beamforming gain but requires N_RF × N_ANT phase shifters
• Higher insertion loss and power consumption

Sub-Connected (Partially-Connected) Architecture:

• Each RF chain drives a disjoint subset of antennas
• Lower hardware complexity and power draw
• Reduced beamforming flexibility and array gain
• Trade-off: spectral efficiency vs. energy efficiency

Practical analog components impose discrete phase resolution, fundamentally limiting beamforming precision.

• Ideal Assumption: Continuous phase control (infinite resolution)
• Practical Reality: 2-6 bit phase shifters (e.g., 4-bit = 22.5° granularity)
• Impact: Quantized phases create residual interference between data streams, degrading spectral efficiency
• Mitigation: Joint optimization algorithms that incorporate quantization constraints during precoder design
• Emerging Alternative: True-time-delay elements replacing phase shifters for wideband operation, eliminating beam squint

Hybrid beamforming leverages the inherent angular sparsity of mmWave channels to reduce dimensionality.

• Physical Basis: mmWave propagation dominated by few dominant paths (typically 2-5 clusters)
• Consequence: Channel matrix is low-rank in angular domain despite high antenna count
• Architectural Implication: Few RF chains can capture dominant spatial modes when analog beams align with physical angles of arrival/departure
• Compressed Sensing: Hybrid architectures naturally implement compressive measurement via analog combining, enabling efficient channel estimation
• Limitation: Performance degrades in rich scattering environments where channel rank exceeds RF chain count

An alternative architecture replacing phase-shifter networks with a discrete lens array performing spatial Fourier transform.

• Mechanism: Lens focuses energy from different angles onto distinct antenna feeds, creating a beamspace representation
• Beam Selection: Digital precoder selects a subset of beamspace ports corresponding to dominant channel paths
• Advantage: Eliminates phase shifters entirely—only requires RF switches
• Energy Efficiency: Significantly lower power consumption than phase-shifter-based architectures
• Application: Particularly suited for fixed wireless access and backhaul where angular spread is limited

A technique using a single RF chain connected to an array of antennas via phase shifters to steer a high-gain beam. While energy-efficient, it is limited to single-stream transmission and offers only coarse directional control. In a hybrid architecture, the analog stage handles the wideband beam steering to overcome high path loss, while the digital stage manages fine-grained precoding.

An architecture where each antenna element is driven by a dedicated RF chain, enabling full control over amplitude and phase in the baseband. This allows for highly flexible multi-user MIMO and interference nulling but is prohibitively expensive and power-hungry at mmWave frequencies due to the high number of antennas. Hybrid beamforming aims to approximate this performance at a fraction of the hardware cost.

The frequency band typically between 30 GHz and 300 GHz, where abundant spectrum enables multi-gigabit-per-second data rates. The short wavelength allows for dense antenna packaging in a small form factor, making massive MIMO feasible. However, signals suffer from severe path loss and blockage, necessitating highly directional beamforming, which is the primary driver for hybrid architectures.

A mathematical matrix applied to data streams before transmission to map them onto antenna elements. In hybrid beamforming, this matrix is decomposed into a product of a high-dimensional analog precoder (implemented with phase shifters) and a low-dimensional digital precoder (implemented in baseband). The joint optimization of these two matrices is a non-convex problem central to hybrid beamforming design.

A MIMO technique that transmits multiple independent data streams simultaneously over the same time-frequency resource to increase spectral efficiency. The number of streams is limited by the minimum of transmit antennas and RF chains. Hybrid beamforming enables spatial multiplexing with fewer RF chains than antennas, trading off some multiplexing gain for drastically reduced hardware cost and power consumption.

The known channel matrix H describing the complex gain between each transmit and receive antenna pair. Accurate CSI is critical for computing the optimal hybrid precoder. In mmWave systems, the channel exhibits sparsity in the angular domain, which is exploited by compressed sensing algorithms to estimate the channel with reduced pilot overhead, a key enabler for practical hybrid beamforming.