Fault Simulation
Fault simulation is a method in which we try to insert actual fault value at the fault site, i.e., 1 for a SA1 or 0 for SA0 etc.
Then the effect of the fault activation is carried out at the primary output of the design
At the last step the primary output is compared with the actual response and once found a miss match, indicates the fault is actually present at the fault site
This is not a test pattern generation (TPG), rather a pattern will be applied first & its response will be compared
Consider the following example,
This is a method which is easy to implement (regular logic simulator), Note: Simulator calculates f(x) value where x is an input; where a TPG calculates x for a given f(x) value
Fault Simulation Complexity O(F*P*G),
F: number of faults,P: number of test patterns,G: number of gates; so is a very time consuming process
Application: Fault Grading, ATPG, etc.
The ATPG process is speed up by the use of fault simulation; otherwise it is used to calculate coverage.
Primarily fault simulation is of two types:
- Serial fault simulation
- Parallel fault simulation
In Series type, simulation of only one fault is carried out at a time, just like the example we have considered earlier in this module
In parallel type, by exploiting parallelism of logic instructions (bit-wise operations) simulation of multiple faults are carried out,
For an example if we consider 2 faults to be checked, we have to consider a word length of w=3, where the MSB bit is reserved for storing fault free response for the circuit for comparison
Parallel Fault Simulation
Fault injection
For injecting faults at simulation level in fault simulation one of the methods is the usage of a transformation method, where the resultant behaviour of a fault injection is shown by:
Summary
- We have seen that parallel fault simulation is much more faster than serial one, the other forms of fault simulations are: PPSFP, Deductive fault simulation, Concurrent fault simulation, Differential fault simulation
- Fault simulation is able to simulate many fault models (stuck-at fault, delay fault, Bridging fault, …) you name it
- But, it is a lengthy process as it has a time complexity of O(m*n), m faults & n vectors,
- Sometimes it is used in calculating DFT Coverages