Figure 4.1: Global flow of the delay tomography.
Figure 4.2: A sensitizable path from an input to the output. Inputs to the circuit are
set such that a rising (falling) transition in input a can propagate to the
output n.
(SLE) with gate variations as its unknown parameters. Then, we estimate varia-
tions by two methods (4a and 4b). The first method is based on the traditional
^-minimization (4a.) In the second method, we show sparsity of variations in
wavelet domain and use compressed sensing (^ɪ-regularization) to estimate vari-
ation more efficiently.
40