"Safe and Efficient One-Hot State Machine"
Jason Xin Zheng, Sunant Katanyoutanant, and Martin Le
Jet Propulsion Laboratory
Single Event Upsets (SEU) pose great threats to the reliability of the Finite State Machines (FSM) that make up the control logics of space avionics. Many mitigation methods, such as Triple Modular Redundancy (TMR), have been devised and used in the past. Although they provide good protection against synchronous faults, many do not cope well with asynchronous faults and carry ample speed and area overhead.
A conceptual analysis of the asynchronous nature of SEU suggests that although it is impossible to fix all asynchronous SEU faults, shortening the critical path can improve the probability of fixing an asynchronous SEU error. Hence, a state machine design approach with emphasis on both error-detection and speed is explored. The approach attacks the speed bottleneck of XNOR error-detection logic for one-hot state machines, improving performance to near the original one-hot state machines. Furthermore, to accelerate the adaptation of this approach, the detection logic and FSM design can be automated through the use of software tools written in Python scripting language by one of the authors. The tools parse text-based state transition tables to generate HDL codes that can be copy-and-pasted into designs. Consequent design modifications can either be made directly to the HDL code or the state transition table. Finally, to study the effects of such approach having on arbitrary state machines, 100 FSM are synthesized with existing and proposed error detection/correction methods. Comparisons are drawn on the logic levels and area consumption as reported by the place-and-route tools to show the effectiveness of the approach.
2005 MAPLD International Conference Home Page