Deriving sequential error propagation probabilities from circuit descriptions

Transient faults, faults caused by temporary disturbance of the circuit, are considered to be a major problem in future circuit designs that build upon nanoelectronic technology. Once a transient fault has occured inside the sequential element (such as a Flip-Flop) of a circuit, it can either be canceled out because the current state of the circuit does not create an active path to another memory elements, or it is propagated into other sequential elements eventually leading to an faulty system output.

The probability of a transient fault being propagated to a system output is highly dependent on the topology and state of the circuit and accurate computation of the probabilities is not feasible except for the smallest of circuits. Statistical methods have proven to be more promising in order to estimate error propagation probabilities for large circuit and can be used to find hot-spots with high likelihood for causing erroneous system outputs and will consequently be used to focus error correction effort.

The goals of the thesis are:

Conducting a survey of existing proposals for computing error propagation probabilities and evaluating these methods regarding their feasibility for large industrial circuits
• Implementing a method from the survey and conducting a study, that compares error probabilities under the assumption, that the error propagation probabilities can be changed by some form of error correction

Contact: Christian Zöllin ( Email: christian.zoellin@informatik.uni-stuttgart.de )