February_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 20 NO. 1 12 The LVP technique [3] integrates voltage waveforms acquired on each transistor of the flip-flop. It requires a looping test sequence with all the measured signals synchronized with the trigger signal of the loop. Un- fortunately, the scan chain failure is on the capture test sequence, which utilizes a high-speed clock generated by an internal phase-locked loop, not synchronized with the trigger signal. The failing test sequence is not visible using LVP. To address the LVP limitations, the shift test sequence (showing no functional failures) is used because it oper- ateswith an external clock. Even if this sequence is always “pass,” the signals acquired on the clock-gated transistor highlighted in SDL show an abnormally slow transition on the rising edge (PMOS1), which generates delays after reformatting of the post-transistor stages (PMOS2 and PMOS3), as shown in Fig. 3. (continued on page 14) Fig. 3 (a) Layout viewof the suspected failing transistor probing locations. LVP image (left) and CAD view (right). (b) Schematic of the failing stage of the flip-flop. Gated clock PMOS1 is affected by a slow transition, which generates an extra delay in stage 2 and at output 3 (Q). (c) Waveforms measured by LVP on a passing reference and the failing flip-flop with the same test conditions. The CK signal is acquired as the reference signal. The failing flip-flop shows strong degradation of the rising edge, creating a CK-to-Q delay 10 × higher than the output signal of the failing flip-flop. These results are acquired with a “passing” shift test loop. Signals are stuck and flat with a “failing” capture test loop. (a) (b) (c)

RkJQdWJsaXNoZXIy MjA4MTAy