Built-in testing requires a method of checking the output responses of the circuit under test that is simpler and less storage-intensive than the conventional bit-by-bit comparison of the actual output values with the previously computed correct values. The usual method is to capture and compare some statistic, called the signature, of the circuit output responses rather than to compare the individual bits themselves. The concept is illustrated in Figure 4.1. Fault detection occurs if the signature realized by a circuit differs from the signature of a fault-free version of the circuit. This chapter provides a brief introduction to six different compression techniques: ones counting, transition counting, parity checking, cyclic codes, syndrome test, and compression using Walsh spectra.
The signature and its collection algorithm should meet four qualitative guidelines: