Long Polar vs. LDPC Codes under Complexity-Constrained Decoding This work is supported by the German Federal Ministry of Research, Technology and Space (BMFTR) within the project Open6GHub (grant no. 16KISK019).
Felix Krieg, Marvin Rübenacke, Andreas Zunker, and Stephan ten Brink · 2025
Read Long Polar vs. LDPC Codes under Complexity-Constrained Decoding This work is supported by the German Federal Ministry of Research, Technology and Space (BMFTR) within the project Open6GHub (grant no. 16KISK019). by Felix Krieg, Marvin Rübenacke, Andreas Zunker, and Stephan ten Brink (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-05485.txt.
Introduction
Felix Krieg, Marvin Rübenacke, Andreas Zunker, and Stephan ten Brink challenge the received wisdom that polar codes only shine for short messages, in a paper published in two thousand twenty-five. Industry and academia alike assume low-density parity-check codes pull ahead as blocks grow longer, an assumption built on giving the LDPC decoder many iterations.
Real hardware cannot always afford those iterations, since latency and complexity budgets bite. The authors compare the families at equal computational effort, counting fixed-point operations. Under that fair budget, long polar codes under successive cancellation decoding beat their LDPC counterparts. A simplified decoder for polar codes scales better than the naive bound suggests. It even needs fewer operations than a single belief propagation round of the rival code. The result reframes a settled debate: the winner depends on the decoding budget, not just the code. Future wireless standards may care.
Catalog record
| CREATOR | Felix Krieg, Marvin Rübenacke, Andreas Zunker, and Stephan ten Brink |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-05485
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Long Polar vs. LDPC Codes under Complexity-Constrained Decoding This work is supported by the German Federal Ministry of Research, Technology and Space (BMFTR) within the project Open6GHub (grant no. 16KISK019). Felix Krieg, Marvin Rübenacke, Andreas Zunker, and Stephan ten Brink Abstract The prevailing opinion in industry and academia is that polar codes are competitive for short code lengths, but can no longer keep up with LDPC ( LDPC ) codes as block length increases. This view is typically based on the assumption that LDPC codes can be decoded with a large number of BP ( BP ) iterations. However, in practice, the number of iterations may be rather limited due to latency and complexity constraints. In this paper, we show that for a similar number of fixed-point LLR ( LLR ) operations, long polar codes under SC ( SC ) decoding outperform their LDPC counterparts. In particular, SSC ( SSC ) decoding of polar codes exhibits a better complexity scaling than N log N and requires fewer operations than a single BP iteration of an LDPC code with the same parameters. ML maximum likelihood BP belief propagation BPL belief propagation list LDPC low-density parity-check BER bit error rate SNR signal-to-noise-ratio BPSK binary phase shift keying AWGN additive white Gaussian noise LLR log-likelihood ratio MAP maximum a posteriori FER frame error rate BLER block error rate SCL successive cancellation list SCAL successive cancellation automorphism list SC successive cancellation BI-DMC Binary Input Discrete Memoryless Channel CRC cyclic redundancy check CA-SCL CRC-aided successive cancellation list PAC polarization-adjusted convolutional BEC Binary Erasure Channel BSC Binary Symmetric Channel BCH Bose–Ray-Chaudhuri–Hocquenghem RM Reed–Muller RS Reed-Solomon SISO soft-in/soft-out 3GPP 3rd Generation Partnership Project eMBB enhanced Mobile Broadband CN check node VN variable node GenAlg Genetic Algorithm CSI Channel State Information OSD ordered statistic decoding MWPC-BP minimum-weight parity-check BP FFG Forney-style factor graph MBBP multiple-bases belief propagation URLLC ultra-reliable low-latency communications mMTC massive machine-type communications DMC discrete memoryless channel SGD stochastic gradient descent QC quasi-cyclic 5G fifth generation mobile telecommunication SCAN soft cancellation LSB least significant bit MSB most significant bit AED automorphism ensemble decoding AE-SC automorphism ensemble successive cancellation PPV Polyanskyi–Poor–Verdú RREF reduced row echelon form PL permutation linear LTA lower triangular affine BLTA block lower triangular affine PTPC pre-transformed polar codes UPO universal partial order SSC simplified successive cancellation SSCL simplified successive cancellation list ASIC application-specific integrated circuit LMS layered min-sum HARQ hybrid automatic repeat request DE density evolution I Introduction LDPC codes are today’s workhorse of error correction [ 1 ] . Their well-understood, iterative BP decoder and near-Shannon-limit performance [ 2 ] made LDPC codes the preferred choice for numerous communication standards, including IEEE 802.3 10G Ethernet, IEEE 802.11 WiFi, IEEE 802.16 WiMAX, DVB-S/T/C2, CCSDS, and the 5G NR data channel. While LDPC codes closely approach the Shannon limit in practice, polar codes were the first to be provably capacity-achieving as the block length tends to infinity [ 3 ] . Nonetheless, it was the development of short-length polar coding techniques—such as the incorporation of an outer CRC ( CRC ) code and the introduction of the SCL ( SCL ) decoding algorithm [ 4 ] —that established polar codes as a leading solution in the short block length regime and finally led to their inclusion in the 5G NR control channel [ 5 ] . Since their introduction in the standard, further progress has been made in academia. Different pre-transformations [ 6 , 7 ] enable further performance gains, while SSC [ 8 ] and SSCL ( SSCL ) [ 9 ] allow for even more efficient decoder implementations. In consideration of the upcoming 6G standard, the already considerable performance in 5G ( 5G ), and the potential incorporation of those recent advances, polar codes emerge as a promising candidate for a second standardization. In addition to “raw” error-rate performance, however, another design objective of a new mobile communications standard could be the unification of decoding schemes [ 10 ] . Rather than employing a variety of codes for each use case, it is beneficial to implement a single code with a standardized, simplified description. This approach would conserve resources, both during the implementation phase and subsequently, in the deployed hardware. However, for such a goal to be realized, the code employed must exhibit sufficient performance and energy efficiency over a broad spectrum of lengths and rates. In the context of 6G as the successor to 5G, the question therefore arises: is it possible to design LDPC codes that achieve the desired performance even at short lengths, or can satisfactory results also be achieved with polar codes at long lengths? In recent years, numerous efforts have focused on improving the performance of short-length LDPC codes (see, e.g., [ 11 ] and references therein; also [ 12 ] , [ 13 ] ), whereas the alternative approach of employing long polar codes has received relatively little attention, with only a few exceptions such as [ 14 ] , [ 15 ] . In our opinion, there are three prevalent misconceptions about polar codes which have led to the perception that LDPC codes are superior in the long block length regime: 1. The larger scaling exponent of polar codes [ 16 , 17 ] compared to LDPC codes make them operate far from the channel capacity for finite lengths; 2. For polar codes to be competitive at all, highly complex decoding algorithms such as list or ensemble decoding are required; 3. The \mathcal{O}(N\log N) complexity scaling of polar encoding and decoding is prohibitively large compared to the linear scaling of LDPC codes. However, the evolution of mobile communication standards has given rise to challenges in energy efficiency, and as a result, BP decoders with large number of iterations—despite their superior error correction capabilities—may no longer be considered optimal [ 18 ] . To the best of our knowledge, the literature lacks a comprehensive comparison between long polar codes and LDPC codes. In this paper, we want to show that the choice between polar and LDPC codes is less clear-cut than commonly assumed. To this end, we show using both theoretic analysis and Monte Carlo simulation that polar codes are competitive against LDPC codes in the long block length regime, especially under decoding complexity constraints. Our key findings are summarized as follows and directly address the misconceptions above: 1. While asymptotically the error rate of polar codes does not decrease as fast as that of LDPC codes, their performance under finite decoding complexity is similar; 2. This competitive performance of polar codes can be achieved using low-complexity SSC decoding; 3. SSC decoding of polar codes scales better than \mathcal{O}(N\log N) . For any practical codeword length, the additional factor of \log N is smaller than the multiplicative constants of BP -based LDPC decoding. Most striking, a full polar decoding is significantly less complex than a single iteration of BP . II Preliminaries II-A Polar Codes & Successive Cancellation Decoding Polar codes are constructed using the polar transform, defined by the N\times N Hadamard matrix \boldsymbol{G}_{N}=\left[\begin{smallmatrix}1&0\\ 1&1\end{smallmatrix}\right]^{\otimes n} , where N=2^{n} and (\cdot)^{\otimes n} denotes the n -th Kronecker power. This transform polarizes a set of identical channels into synthetic channels that are either highly reliable or