Pre-silicon FEC decoding verification on SoC FPGAs

Forward error correction (FEC) decoding hardware modules are challenging to verify at pre-silicon stage, when they are usually described at register-transfer (RT)/logic level with a hardware description language (HDL). They tend to hide faults due to their inherent tendency to correct errors and the...

Descripción completa

Detalles Bibliográficos
Autores: Fernández Solórzano, Víctor Manuel|||0000-0003-0614-151X, Abad García, Carlos, Álvarez Ruiz, Ángel, Ugarte Olano, Íñigo, Sánchez Espeso, Pablo Pedro
Tipo de recurso: artículo
Fecha de publicación:2021
País:España
Institución:Universidad de Cantabria (UC)
Repositorio:UCrea Repositorio Abierto de la Universidad de Cantabria
Idioma:inglés
OAI Identifier:oai:repositorio.unican.es:10902/20523
Acceso en línea:http://hdl.handle.net/10902/20523
Access Level:acceso abierto
Palabra clave:Verification
Platform FPGAs
Prototyping
Emulation
BER/CER testing
Descripción
Sumario:Forward error correction (FEC) decoding hardware modules are challenging to verify at pre-silicon stage, when they are usually described at register-transfer (RT)/logic level with a hardware description language (HDL). They tend to hide faults due to their inherent tendency to correct errors and the required simulations with a massive insertion of inputs are too slow. In this work, two verification techniques based on FPGA-prototyping are applied in order to complement the mentioned simulations: golden model vs implementation matching with thousands of random codewords and codeword/bit error rate (CER/BER) curve computation. For this purpose, a system on chip (SoC) field-programmable gate array (FPGA) is used, implementing in the programmable hardware part several replicas of the decoder (exploiting the parallel capabilities of hardware) and managing the verification by parallel programming the software part of the SoC (exploiting the presence of multiple processing cores). The presented approach allows a seamless integration with high-level models, does not need expensive testing/emulation platforms and obtains the results in a reasonable amount of time.