Precoding

Precoding fundamentally relies on accurate Channel State Information at the transmitter (CSIT). The precoding matrix is computed as a function of the estimated channel matrix H to pre-compensate for the propagation environment. In Frequency Division Duplex (FDD) systems, CSI must be fed back from the receiver via a limited feedback channel using quantized indices like the Precoding Matrix Indicator (PMI). In Time Division Duplex (TDD) systems, channel reciprocity is exploited to estimate the downlink channel directly from uplink pilots.

• Full CSIT: Enables optimal precoding designs like singular value decomposition (SVD)-based eigen-beamforming
• Partial/Limited CSIT: Requires codebook-based precoding where the transmitter selects from a finite set of predefined matrices
• Statistical CSIT: Uses long-term channel covariance information when instantaneous CSI is unavailable

Precoding schemes are broadly categorized by their computational complexity and interference management strategy. Linear precoding applies a simple matrix multiplication to the symbol vector, while nonlinear precoding employs more sophisticated symbol-level processing.

• Maximum Ratio Transmission (MRT): Maximizes received signal power at the intended user; optimal in noise-limited, single-user scenarios
• Zero-Forcing (ZF) Precoding: Completely eliminates multi-user interference by inverting the channel matrix, but may amplify noise in poorly conditioned channels
• Minimum Mean Square Error (MMSE) Precoding: Balances interference suppression and noise enhancement by regularizing the channel inversion
• Dirty Paper Coding (DPC): A theoretical nonlinear technique that achieves the MIMO broadcast channel capacity by pre-subtracting known interference without power penalty
• Tomlinson-Harashima Precoding (THP): A practical nonlinear approach using modulo arithmetic and successive pre-subtraction to approximate DPC performance

To reduce feedback overhead, modern standards like 5G NR and LTE define finite sets of precoding matrices known as codebooks. The receiver estimates the channel, searches the codebook for the matrix that maximizes a performance metric (e.g., throughput or signal-to-interference-plus-noise ratio), and reports the corresponding PMI index back to the transmitter.

• Type I Codebook: Standard-resolution codebook for single-user MIMO and basic multi-user scenarios, optimized for low feedback overhead
• Type II Codebook: High-resolution codebook using linear combination of multiple beams with amplitude and phase quantization, enabling advanced MU-MIMO with up to 12 orthogonal beams
• Port Selection Codebook: A subset of Type II where beams are selected from a reference signal port set rather than freely in space, reducing complexity
• eType II: Enhanced Type II codebook with frequency-domain compression for improved performance in wideband channels

In Multi-User MIMO (MU-MIMO) systems, precoding serves the dual purpose of maximizing signal power to each intended user while actively suppressing the interference caused to co-scheduled users. This spatial separation enables simultaneous transmission to multiple users on the same time-frequency resource.

• Block Diagonalization (BD): Constrains each user's precoding matrix to lie in the null space of all other users' channel matrices, completely eliminating inter-user interference when sufficient transmit antennas are available
• Signal-to-Leakage-plus-Noise Ratio (SLNR) Precoding: Maximizes the ratio of desired signal power to the interference leaked to other users plus noise, offering a computationally efficient alternative to BD
• Regularized ZF (RZF): Adds a regularization term to the ZF precoder to control the trade-off between interference suppression and desired signal power, converging to optimal DPC performance in massive MIMO regimes

In massive MIMO and millimeter-wave systems, fully digital precoding with one RF chain per antenna is prohibitively expensive and power-hungry. Hybrid beamforming splits the precoding operation between a low-dimensional digital baseband precoder and a high-dimensional analog beamformer implemented with phase shifters.

• Fully-Connected Architecture: Each RF chain connects to all antennas via a network of phase shifters, offering near-optimal performance at high hardware cost
• Partially-Connected (Sub-Array) Architecture: Each RF chain connects to a disjoint subset of antennas, reducing complexity and power consumption with some performance degradation
• Two-Stage Design: The analog precoder is first designed to maximize array gain using beamsteering codebooks, then the digital precoder manages multi-stream or multi-user interference
• Compressed Sensing for Beam Selection: Leverages the sparse nature of mmWave channels to select dominant beams with minimal training overhead

The 3GPP 5G New Radio (NR) specification defines a flexible precoding framework supporting diverse deployment scenarios from sub-1 GHz to millimeter-wave frequencies. Key features include advanced CSI reporting mechanisms and support for up to 8 layers in downlink and 4 layers in uplink.

• CSI-RS-Based Feedback: Channel State Information Reference Signals enable the UE to measure the channel and report Rank Indicator (RI), PMI, and Channel Quality Indicator (CQI)
• SRS-Based Uplink Precoding: Sounding Reference Signals allow the gNB to estimate the uplink channel and compute the precoding matrix for PUSCH transmission
• Non-Codebook Based Precoding: For TDD systems with channel reciprocity, the UE can determine its own precoder from downlink measurements without relying on a standardized codebook
• Coherent vs. Non-Coherent Transmission: Defines codebook subsets for fully coherent, partially coherent, and non-coherent antenna configurations to accommodate different UE capabilities

The critical enabler of any closed-loop precoding system. CSI represents the known channel properties—scattering, fading, and power decay—that a transmitter uses to adapt its signal. Without accurate CSI feedback, precoding cannot spatially steer energy. Key aspects include:

• Implicit Feedback: Exploits channel reciprocity in TDD systems
• Explicit Feedback: Quantized CSI reports from the receiver in FDD systems
• CSI Aging: The degradation of precoding performance due to channel variation between estimation and transmission

A feedback index from the User Equipment (UE) that recommends a specific precoding matrix from a predefined codebook. The PMI allows the transmitter to select the optimal beamforming weights without requiring full channel quantization. Critical for 5G NR and LTE standards:

• Type I Codebook: Standard resolution for single-user MIMO
• Type II Codebook: High-resolution linear combination for multi-user MIMO
• Port Selection: Extensions for FDD massive MIMO systems

The mathematical backbone of optimal linear precoding. SVD decomposes the MIMO channel matrix into parallel, non-interfering eigenmodes. By precoding with the right singular vectors, the transmitter creates independent spatial pipes—a process known as eigen-beamforming:

• Water-filling: Optimal power allocation across eigenmodes based on their gain
• Rank Adaptation: Selecting the number of active streams based on the singular value spread
• Condition Number: A metric derived from singular values indicating channel suitability for spatial multiplexing

A theoretical precoding strategy that achieves the capacity region of a MIMO broadcast channel. DPC pre-subtracts known interference at the transmitter without increasing transmit power, as if the interference were not present. While computationally prohibitive in practice, it serves as the theoretical benchmark:

• Costa's Writing on Dirty Paper: The foundational information-theoretic result
• Tomlinson-Harashima Precoding (THP): A practical, non-linear approximation using modulo arithmetic
• Vector Perturbation: Another suboptimal approach that searches for a perturbation vector to minimize transmit power

A linear precoding technique for Multi-User MIMO (MU-MIMO) that completely eliminates inter-user interference. BD constrains each user's precoding matrix to lie in the null space of all other users' channel matrices:

• Zero-Forcing MU-MIMO: A special case where each user has a single antenna
• Regularized BD: Introduces a regularization term to balance interference suppression with noise enhancement
• Signal-to-Leakage-plus-Noise Ratio (SLNR): An alternative optimization criterion that maximizes desired signal power relative to interference leaked to other users

An architecture for massive antenna arrays that splits precoding between digital and analog domains. Hybrid beamforming reduces the prohibitive cost and power consumption of a dedicated RF chain per antenna:

• Digital Precoder: Low-dimensional baseband processing for multi-stream multiplexing
• Analog Beamformer: High-dimensional phase-shifter network for directional gain
• Fully-Connected vs. Sub-Array: Architectures trading off beamforming flexibility against hardware complexity