In the intricate architecture of 5G New Radio (NR), ensuring data integrity during high-speed transmission  is paramount. This is the critical role of Forward Error Correction (FEC), a sophisticated set of channel  coding techniques designed to detect and correct transmission errors on the fly. By proactively adding redundant  “parity” bits to the data before it’s sent over the airwaves, 5G networks can overcome signal noise and  interference, drastically reducing the need for retransmissions and ensuring the low latency and high  reliability that define the standard.

Unlike previous generations of wireless technology that primarily relied on Turbo and Convolutional codes,  5G NR adopts a strategic split approach, utilizing two different state-of-the-art FEC schemes for  distinct purposes: Low-Density Parity-Check (LDPC) codes and Polar codes.

LDPC Codes: The Workhorse for High-Speed Data

Low-Density Parity-Check (LDPC) codes are the primary FEC mechanism for all user data in 5G.

They are applied to the data-carrying physical channels:

1. Physical Downlink Shared Channel (PDSCH): Carries all user data from the base station (gNB) to the user device (UE).
2. Physical Uplink Shared Channel (PUSCH): Carries all user data from the UEs to the gNB.

Why LDPC was chosen for data:

The selection of LDPC for the Enhanced Mobile Broadband (eMBB) use case was driven by its exceptional performance characteristics for large data blocks.

High Throughput: The structure of LDPC codes allows for highly parallelizable decoding algorithms.
This means that a large block of data can be processed simultaneously by multiple hardware units,
making it ideal for the high-throughput, multi-gigabit speeds promised by 5G.

Capacity-Approaching Performance: LDPC codes are renowned for their ability to perform very
close to the Shannon Limit—the theoretical maximum rate of error-free data transmission over a noisy channel.
This translates to highly efficient use of the available radio spectrum.

Scalability: The 3GPP standard defines a flexible and rate-compatible LDPC framework.
It uses two base graphs—one optimized for smaller transport blocks and another for larger
ones—that can be adapted to a wide variety of code rates and block sizes, providing scalability
for different network conditions and service requirements.

5G NR LDPC Codec and Rate Matching Suite

Our solution delivers a complete implementation of the LDPC codec (encoder and decoder) and rate matching/recovery functions as defined in 3GPP TS 38.212, fully optimized for the demands of 5G NR physical layer processing.

Key Features

• Standards-Compliant LDPC Codec
  1. Full LDPC Codec Implementation
     1. Compliant with 3GPP TS 38.212
     2. Supports Base Graphs BG1 and BG2
     3. Supports all Code rates and Block sizes corresponding to possible lifting sizes
  2. Rate Matching and Recovery
     1. Complete support for puncturing, repetition, and shortening
     2. Fully integrated with HARQ buffer management
     3. Supports up to 16 HARQ processes and 64 user pers HARQ ID (configurable)

• CUDA-C Implementation for GPUs
  Designed and optimized for NVIDIA GPUs, including low-cost RTX gaming GPUs. Scalable and highly configurable across GPU generations.

• Superior BLER Performance
  Achieves improved BLER performance over generic decoder implementations. Demonstrated via detailed simulation plots and throughput analysis as per 38.214  – Table 5.1.3.1-1: MCS index table.

• Fully Integrated HARQ Management
  Includes complete HARQ soft buffer management internal to the GPU, minimizing PCIe overhead and maximizing throughput in real-time systems.

• Flexible and Easy Integration
  Provides a standards-aligned FEC API for smooth integration into existing 5G L1. Designed for modularity and rapid deployment.

• Comprehensive FEC Stats and Logging, Memory Usage Stats and Safety Memguard Mechanism for host and device memories.