Stefan Weithoffer received his Ph.D. in 2018 from the University of Kaiserslautern, Germany, where he worked with his doctoral advisor Prof. Norbert Wehn on implementation issues of high-throughput Turbo decoding. He has designed several Turbo Decoder IP in the course of collaborations with academia as well as industry which have been successfully integrated and manufactured in Systems-on-Chip (SoC) on 28 nm. Since September 2019, he is an associate professor with the MEE department at IMT Atlantique.
TurboLEAP (https://www.imt-atlantique.fr/en/research-innovation/projects/turboleap) is a JCJC project financed through the ANR AAPG 2020 CE 25 « Réseaux de communication multi-usages, infrastructures de hautes performances, Sciences et technologies logicielles ».
The efficiency of very high throughput LDPC and Polar decoders in terms of throughput/area is higher than that of Turbo decoders. On the other hand, it comes at the cost of a lack of flexibility with regard to the code rate and the frame size. Most Turbo decoders are either very flexible regarding frame sizes or can be made flexible. However, when supported with the classic BCJR decoding algorithm, this flexibility represents one of the main factors that lead to the reduced efficiency of Turbo decoders. Consequently:
- There is great potential for reducing Turbo decoder complexity for applications with limited flexibility requirements and
- second, for applications requiring high flexibility such as mobile broadband, Turbo codes can be significantly superior to LDPC or Polar codes if the rate-limiting factors achievable for hardware implementations can be overcome.
The path to Turbo decoders supporting Tb/s goes through the joint introduction of deep modifications of their decoding algorithm and their code design while taking into account implementation and application constraints.
TurboLEAP aims at decoding Turbo codes at Tb/s with current silicon technologies and closing the throughput-gap to LDPC and Polar codes.