Skip to Main content Skip to Navigation

Energy modeling and optimization of protograph-based LDPC codes

Mohamed Yaoumi 1, 2
Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance
Abstract : There are different types of error correction codes (CCE), each of which gives different trade-offs interms of decoding performanceand energy consumption. We propose to deal with this problem for Low-Density Parity Check (LDPC) codes. In this work, we considered LDPC codes constructed from protographs together with a quantized Min-Sum decoder, for their good performance and efficient hardware implementation. We used a method based on Density Evolution to evaluate the finite-length performance of the decoder for a given protograph.Then, we introduced two models to estimate the energy consumption of the quantized Min-Sum decoder. From these models, we developed an optimization method in order to select protographs that minimize the decoder energy consumption while satisfying a given performance criterion. The proposed optimization method was based on a genetic algorithm called differential evolution. In the second part of the thesis, we considered a faulty LDPC decoder, and we assumed that the circuit introduces some faults in the memory units used by the decoder. We then updated the memory energy model so as to take into account the noise in the decoder. Therefore, we proposed an alternate method in order to optimize the model parameters so as to minimize the decoder energy consumption for a given protograph.
Document type :
Complete list of metadatas
Contributor : Abes Star :  Contact
Submitted on : Tuesday, February 23, 2021 - 12:05:19 PM
Last modification on : Thursday, February 25, 2021 - 3:11:03 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03149756, version 1


Mohamed Yaoumi. Energy modeling and optimization of protograph-based LDPC codes. Information Theory [math.IT]. Ecole nationale supérieure Mines-Télécom Atlantique, 2020. English. ⟨NNT : 2020IMTA0224⟩. ⟨tel-03149756⟩



Record views