GHz testing and its fuzzy targets

Summary form only given. GHz ICs happen to be more sensitive to interactive effects than their ancestors because 100 nm geometries have more process variance, and the clock period is so small at 200-500 ps. These small clock periods are less forgiving. A parameter may have a statistical variance that by itself may not cause failure, but can cause failures when combined with the statistical values of other parameters. These are called interactive effects. These variables have implications for testing GHz ICs. The statistical parameters that we now see demand statistical analysis of test variables, such as F_max, temperature, I_DDQ, memory access time, and V_DD. The test procedure will fail ICs, not on a single test, but on deviations from expected behavior over the full range of data. Multi-parameter testing will become the dominant test paradigm.