Source-linked AI summary
Iterative Detection and Decoding Algorithms using LDPC Codes for MIMO Systems in Block-Fading Channels
André Uchoa, Cornelius T. Healy, Rodrigo C. de Lamare
TL;DR
The paper addresses IDD for MIMO systems in block-fading and fast Rayleigh fading channels, where few related studies exist and antenna diversity must be exploited in receiver design. It develops soft-information IDD and LDPC scheduling techniques, achieving gains of up to 2 dB in block-fading and 1.5 dB in fast-fading settings.
Problem
Few studies address IDD for MIMO systems in Rayleigh and fast Rayleigh fading channels, motivating receiver designs that exploit multiple-antenna freedom.
Method
The paper develops IDD algorithms that manipulate decoder-output LLRs, exploit LDPC code structure, and include the ROLBP scheduling algorithm for block-fading channels.
Results
The proposed algorithms achieve gains of up to 2 dB for 2×2 MIMO and 1.5 dB for 4×4 MIMO in block-fading, while LLR-SP-BF IDD gains up to 1.5 dB for 2×2 fast-fading MIMO.
Takeaways & Limitations
The proposed IDD and scheduling techniques improve BER performance for MIMO systems in both block-fading and fast-fading channels.
Abstract
from arXiv · showhide
We propose iterative detection and decoding (IDD) algorithms with Low-Density Parity-Check (LDPC) codes for Multiple Input Multiple Output (MIMO) systems operating in block-fading and fast Rayleigh fading channels. Soft-input soft-output minimum mean-square error receivers with successive interference cancellation are considered. In particular, we devise a novel strategy to improve the bit error rate (BER) performance of IDD schemes, which takes into account the soft \textit{a posteriori} output of the decoder in a block-fading channel when Root-Check LDPC codes are used. A MIMO IDD receiver with soft information processing that exploits the code structure and the behavior of the log likelihood ratios is also developed. Moreover, we present a scheduling algorithm for decoding LDPC codes in block-fading channels. Simulations show that the proposed techniques result in significant gains in terms of BER for both block-fading and fast-fading channels.
I. INTRODUCTION
The paper develops MIMO IDD techniques for block-fading channels, addressing the limited prior work in this setting. Its contributions manipulate decoder LLRs, exploit LDPC code structure, and introduce sequential scheduling without significant extra computation or memory.
- The work builds on MIMO diversity and multiplexing gains, cost-effective SIC-based detection, and prior IDD strategies.
- The paper addresses the limited prior study of MIMO systems operating over block-fading channels.
- The proposed IDD scheme manipulates decoder LLRs while exploiting LDPC code structure to improve block-fading performance.
- A new sequential scheduling method is introduced to further improve MIMO IDD performance in block-fading channels.
- The proposed algorithms improve performance without significant extra computational effort or additional memory storage.
II. SYSTEM MODEL
The paper models Root-Check LDPC-coded MIMO transmission over block-fading and fast-fading channels, using iterative soft-information exchange between an MMSE-SIC detector and an LDPC decoder.
- Transmission model: The system uses a Root-Check LDPC encoder with rate 1/F across n_tx transmit antennas and F independent fading blocks.Encoded symbols are organized into n_tx-dimensional vectors and transmitted over the block-fading channel.
- Transmission model: The received vector follows a linear MIMO model with channel matrix H and complex Gaussian noise of covariance σ_v^2I.The channel coefficients represent gains between transmit and receive antennas, and the received signal is demodulated and matched filtered.
- Channel conditions: Fast fading is represented by assigning each received symbol a distinct fading coefficient, corresponding to F = L.The fading-block index is related to the time index through a ceiling-based mapping.
- IDD receiver: Iterative detection and decoding exchanges extrinsic and soft information between the soft MIMO detector and the LDPC decoder.Decoder-only updates are inner iterations, while detector–decoder exchanges are outer iterations.
- IDD receiver: The receiver uses linear MMSE filtering with successive interference cancellation, while decoding includes a block-fading scheduling method combining LBP and RBP benefits.The detector and decoder exchange log-likelihood information for each coded bit.
III. PROPOSED LLR COMPENSATION SCHEME
The proposed LLR-PS-BF scheme compensates for unequal parity-check LLR magnitudes in Root-Check LDPC codes, preventing unreliable feedback from degrading iterative detection.
- Motivation: Simulations show that parity-check nodes in Root-Check LDPC codes may fail to converge, with the deepest-faded connections having lower LLR magnitudes than others.This magnitude difference was not observed for standard LDPC codes.
- Motivation: Unequal and low parity-check LLR magnitudes can approach the non-reliable decision region and cause incorrect symbol demapping during IDD.The exchanged LLRs can therefore degrade overall performance.
- LLR-PS-BF procedure: LLR-PS-BF organizes detector a posteriori LLRs into systematic and parity-check components before compensating the parity-check values.The method assumes systematic-node LLR magnitudes remain greater than zero and computes γ = α − β > 0.
- LLR-PS-BF procedure: The compensation constructs parity-check magnitudes and signs, combines them through a Hadamard-product operation, and replaces the parity-check portion of the detector output.The optimized vector retains systematic LLRs and substitutes compensated parity-check LLRs.
- Effect in IDD: Applying LLR-PS-BF after each detector a posteriori output keeps exchanged LLRs suitable for iteration and yields better BER performance.The scheme is required whenever Root-Check LDPC codes are used in the described IDD process.
IV. PROPOSED IDD SCHEME BASED ON SCHEDULING
The paper develops scheduling strategies for LDPC decoding in block-fading channels, culminating in ROLBP, which alternates residual-based and ordered updates to preserve root-connection information.
- Scheduling comparisons: LBP outperforms standard BP after applying the proposed LLR-PS-BF scheme, whereas RBP and NWBP are outperformed by standard BP.The reported scheduling behavior depends on compensating the LLR effects introduced by block fading.
- Scheduling comparisons: Residual-based methods can prioritize messages from variable nodes lacking channel information, producing performance degradation in block-fading channels.Small residuals in Tanner-graph subgraphs do not necessarily indicate convergence.
- ROLBP scheduling: The proposed schedule forms a descending residual queue and sequentially updates variable-to-check and check-to-variable messages according to the selected order.The method is initialized with channel-information LLRs and repeated until the stopping rule is satisfied.
- ROLBP scheduling: ROLBP alternates residual ordering with a predefined check-node order to overcome block-fading scheduling problems.It computes residual metrics and ordering on alternating iterations, while retaining ordered updates on the others.
- ROLBP scheduling: ROLBP outperforms standard BP and RLBP because its alternating schedule supplies root connections with sufficient information.The strategy avoids residual values associated with degradation in Root-Check decoding performance.
- Complexity: In complex multiplications, NWBP is most expensive, followed by RLBP, ROLBP, BP, and LBP.The comparison is expressed in terms of variable-node and check-node degrees and the number of Tanner-graph edges.
V. SIMULATIONS
Simulations evaluate the proposed IDD techniques in block-fading and fast-fading MIMO channels, showing BER gains over standard decoding strategies and manageable complexity trade-offs.
- Simulation setup: Root-Check LDPC codes require fewer inner decoder iterations than standard LDPC codes, while the simulations used up to 20 inner and 5 outer iterations.The evaluated code rates were 1/2 and 1/4, with block length N = 1024.
- Complexity: The ROLBP algorithm has higher complexity than BP and LBP but lower complexity than RLPB and NWBP.Complexity was analyzed in terms of complex multiplications.
- Block-fading results: ROLBP with LLR-PS-BF outperformed standard LDPC codes with BP by up to 1.5 dB in 4×4 block-fading MIMO.ROLBP alone outperformed BP by about 1.25 dB in this setting.
- Fast-fading results: In fast-fading 2×2 MIMO, LLR-PS-BF with ROLBP gained about 1 dB over standard LDPC codes and about 1.5 dB over LBP at low SNR.The standard LDPC curves for BP, LBP, and ROLBP had the same performance in the reported comparison.
- Code applicability: Root-Check codes with F = 2 and the proposed compensation scheme can address scenarios with F = L/2 or F = L/4 without redesigning the code.The F = 2 code captures the reported advantages across a wider range of fading numbers.
VI. CONCLUSION
The paper presents an IDD scheme and ROLBP scheduling algorithm for MIMO systems in block-fading channels. The proposed algorithms provide reported SNR gains in both block- and fast-fading settings, especially for high-throughput users with slowly changing channels.
- Contributions: The work presents an IDD scheme, proposes ROLBP scheduling, and studies different scheduling strategies.
- Block-fading conclusion: The proposed algorithms achieved up to 2 dB gain for 2×2 and up to 1.5 dB for 4×4 MIMO in block-fading channels with F = 2.The reported gains are for point-to-point systems.
- Scope: The proposed algorithms are suitable for MIMO systems whose users experience high throughput rates and slowly changing propagation channels.In these scenarios, the symbol period is much smaller than the coherence time.
LLR-PS-BF MATHEMATICAL ANALYSIS
The LLR-PS-BF compensation scheme transforms decoder LLRs so that unreliable parity-check decisions are moved farther from the non-reliable region. It uses functions f and g to compute and apply the compensation.
- Compensation mechanism: LLR-PS-BF modifies the decoder LLR vector by computing a positive compensation value and generating a compensated vector.The transformation is represented by functions f[lC] and g[lC].
- Mathematical formulation: Given an input vector lC, f[lC] obtains a real positive value ∆, and g[lC] generates the compensated output ˜lC.
- Example: For the illustrated case, the input vector has length N = 1024 and the first K = 512 components are treated separately from the remaining components.
- Effect on LLRs: The compensation moves parity-check LLRs farther from the region associated with non-reliable decisions.The non-optimized vector contains parity-check LLRs near that region, whereas the compensated vector places them farther away.
Algorithm 1 Proposed LLR-SP-BF Scheduling IDD Scheme
The proposed scheduling IDD procedure combines soft MIMO detection, LDPC decoding, and block-fading-aware extrinsic LLR processing. In block-fading channels, it selects the best channel realization before forming the extrinsic information.
- Detection and LLR processing: The receiver performs MMSE successive-interference-cancellation detection for each transmit layer and then computes extrinsic bit LLRs.
- Block-fading scheduling: For block-fading channels, the algorithm selects index δf corresponding to the maximum |det(hk,f)| across fading realizations.The extrinsic LLR ǫk is calculated at the fading index selected by this criterion.
- Scheduling condition: The scheduling branch is optional, so channels other than block-fading do not require the additional best-realization selection steps.
- Iterative decoding: The procedure decodes with the specified LDPC equations, applies standard belief propagation, and obtains the decoder's a posteriori LLR output for the soft detector.
16. Apply the proposed LLR-PS-BF scheme equations (3) up to (7)
The scheme computes extrinsic information for both decoder and detector exchanges using the code-related LLR lC[xj].
- Extrinsic information lE[xj] is calculated from lC[xj] for transmission to the decoder.
- Extrinsic information lE[xj] is calculated from lC[xj] for transmission to the detector.