"The STAR Systems Toolset for Analyzing Reconfigurable System Cross-Section"
Heather Quinn, Debayan Bhaduri, Christof Teuscher, Paul Graham, and Maya Gokhale
Los Alamos National Laboratory
In recent years, space-based reconfigurable systems have become possible. Field-programmable gate arrays (FPGAs) are well-matched with many algorithms used to process satellite data sets. Processing data on-orbit will allow users to retrieve already analyzed data from the satellite. Since processed data sets can be significantly smaller than unprocessed data sets, there is an advantage to sending processed data sets. Unfortunately, reprogrammable FPGA devices are very soft and are prone to radiation-induced upsets. A majority of the radiation-induced errors can be mitigated by protecting the FPGA's design with triple-modular redundancy (TMR) schemes. Achieving "perfect" TMR is difficult in many cases, since triplicating input/output pins or the modules might not be possible. In these cases, partial TMR, such as Brigham Young University's BLTMR tool can be useful. Reliability analysis of these systems and their codes is a necessary step in designing, verifying, and using these devices in the harsh space radiation environment.
In this talk, we will present an overview of a reliability analysis toolset, called the Scalable Tool for the Analysis of Reliable Systems (STAR Systems) that determines the reliability of a reconfigurable payload, including a module for determining the reliability of FPGA designs (STAR-Circuits). The STAR Systems toolset was designed to provide circuit and system designers with a quick, worst-case reliability estimate so that reliability can be addressed early in the design process. This method is a good pre-step before fault injection or accelerated testing, since it is substantially faster and can be done before the system is completely realized. The STAR-C tool determines the reliability of the circuit's outputs from the circuit's EDIF. The reliability is calculated hierarchically, building the calculation from the architectural primitives up to the circuit elements. Reliability can be determined from either a probability of transistor failure or a sensitive cross-section for radiation-induced upsets. Reliability values are memoized for reuse. The STAR-C tool is used with the STAR Systems tool for determining the reliability of reconfigurable systems executing specific FPGA codes. The STAR Systems module allows the user to prototype payloads by allowing the user to change the system parameters so that a reliable system design can be found. Together these tools will allow system designers to explore potential problems with specific architectures before deployment.
2006 MAPLD International Conference Home Page