Title
Transient Fault Detection in State-Automata
Abstract
State automata are implemented in numerous ways and technologies - from simple traffic light controls to high-performance microprocessors comprising thousands of different states. Highlyintegrated microprocessors get more and more susceptible to transient faults induced by radiation, extreme clocking, temperature and decreasing voltage supplies. A transient fault in form of a single event-upset (SEUs) can change the current state of an automaton to another valid state, thus causing a control-flow error. From control-flow based simulations of a microprogrammable automaton we determine the number of effective, overwritten and latent faults. Faults can be detected by counting the number of transitions to the ending state and the comparison with a precomputed value being part of the microcode and the number of counted cycles. Faults cannot be detected if the original state is transferred to another valid state, reaching the ending state in the same number of transitions. We further determine the number of faults which can be detected by using this simple scheme and propose to encode these states in a way that a bit-flip will result in a state with a different distance from the ending state without any additional space consumption for the code.
Year
DOI
Venue
2007
10.1109/DEPCOS-RELCOMEX.2007.51
DepCoS-RELCOMEX
Keywords
Field
DocType
highlyintegrated microprocessors,original state,different state,valid state,transient fault detection,high-performance microprocessors,state automaton,transient fault,control-flow error,different distance,current state,finite state machines,error correction,automatic control,control flow,temperature,automata,radiation,fault detection,voltage
Microcode,Fault detection and isolation,Computer science,Automaton,Voltage,Algorithm,Finite-state machine,Real-time computing,Automatic control,Error detection and correction,Single event upset
Conference
ISBN
Citations 
PageRank 
0-7695-2850-3
0
0.34
References 
Authors
3
2
Name
Order
Citations
PageRank
Bernhard Fechner17812.18
Andre Osterloh2104.49