Hardening basic blocks in a mesh of clusters FPGA

Abstract : As feature sizes scale down to nano-design level, electronic devices have become smaller, more performant, less power-onsuming, but also less reliable. Indeed, reliability has arisen as a serious challenge in nowadays’ microelectronics industry and as an important design criterion, along with area, performance and power consumption. For instance, physical defects due to imperfections in the manufacturing process have been observed more frequently, impacting the yield. Besides, nanometric circuits have become more vulnerable during their lifetime to ionizing radiation which causes transient faults. Both manufacturing defects and transient faults contribute to decreasing reliability of integrated circuits. When moving to a new technology node, Field Programmable Gate Arrays (FPGAs) are the first coming into the market, thanks to their low development and Non-Recurring Engineering (NRE) costs and their flexibility to be used for any application. FPGAs have especially attractive characteristics for space and avionic applications, where reconfigurability, high performance and low-power consumption can be fruitfully used to develop innovative systems. However, missions take place in a harsh environment, rich in radiation, which can induce soft errors within electronic devices. This shows the importance of FPGA reliability as a design criterion in safety and critical applications. Most of commercial FPGAs have a mesh architecture and their logic blocks are gathered into clusters. Therefore, this thesis deals with the fault tolerance of basic blocks (clusters and switch boxes) in a mesh of clusters FPGA. These blocks are mainly made up of multiplexers. In order to improve their reliability, it is imperative to be able to assess it first, then select the proper hardening approach according to the available budget. So, this is the main outline in which this thesis is conceived. Its goals are twofold: (a) analyze the fault tolerance of the basic blocks in a mesh of clusters FPGA, and point out the most vulnerable components (b) propose hardening schemes at different granularity levels, depending on the hardening budget. As far as the first goal is concerned, a methodology to evaluate the reliability of the cluster is proposed. This methodology uses an existent analytical method for reliability computation of combinational circuits. The same method is employed to identify the worthiest components to be hardened. Regarding the second goal, hardening techniques are proposed at both multiplexer and transistor levels. At multiplexer level, two hardening solutions are presented. The first solution resorts to spacial redundancy and concerns the logic block structure. A novel Configurable Logic Block (CLB) architecture baptized Butterfly is introduced. It is compared with other hardened CLB architectures in terms of reliability and cost penalties. The second hardening solution is a redundanceless scheme. It is based on a “smart” synthesis that consists in seeking the most reliable design in a given founder library, instead of directly using a redundant solution. Then, at transistor level, new single-ended and dual-rail multiplexer architectures are proposed. They are compared to different other transistor structures, according to suitable design metrics.
Document type :
Theses
Complete list of metadatas

Cited literature [126 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/tel-01307489
Contributor : Abes Star <>
Submitted on : Tuesday, April 26, 2016 - 3:23:08 PM
Last modification on : Thursday, October 17, 2019 - 12:36:09 PM
Long-term archiving on : Wednesday, July 27, 2016 - 3:01:01 PM

File

theseBenDhiaArwaV2.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01307489, version 1

Collections

Citation

Arwa Ben Dhia. Hardening basic blocks in a mesh of clusters FPGA. Electronics. Télécom ParisTech, 2014. English. ⟨NNT : 2014ENST0068⟩. ⟨tel-01307489⟩

Share

Metrics

Record views

402

Files downloads

213