Previous Page  8 / 66 Next Page
Information
Show Menu
Previous Page 8 / 66 Next Page
Page Background

edfas.org

ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 18 NO. 4

8

checkerboard row, pinpoints the latch callout (Fig. 5a).

In Fig. 5(b), the six scan chains in a pack are failing at the

same latch position, indicating a clock-type failure.

CASE STUDIES

Because the in-line scan chain logicmacro is diagnos-

able and single-latch fallout canbe identifiedbasedon the

preload test, ahard single-latch failuremaybe successfully

analyzedwith routine physical failure analysis (PFA) using

top-down scanning electron microscopy (SEM) inspec-

tion. However, as defined previously, the root cause of a

clock-type failure could be any failure in the buffers and

the interconnects of the preload data and clock lines, or a

bridging fault in the preload data and clock input circuits

inany individual latches. This involves a large area. Clearly,

it is more challenging to perform PFA on a clock-type

failure. Thus, for clock-type failures, LVI and LVPdiagnoses

are necessary for localizing the failure to a small area prior

to PFA. In addition, for soft single-latch failures, although

the possible defective area is the same as for hard single-

latch failures, the defect causing a soft single-latch failure

ismuch smaller andmore subtle. Due to the subtle nature

of the defect and the large number of transistors (asmany

as 50 per latch) that must be inspected, the success rate

with routine PFA for soft single-latch failures is very low.

With the help of LVI and LVP, the success rate has been

improved to approximately 85%. This section describes

three cases that demonstrate how to apply LVI and LVP in

diagnosing scan chain clock-type failures and soft single-

latch failures. The root causes of these failureswere found

by subsequent PFA.

SKIP TEST SCAN CHAIN FAILURE

As previously mentioned, the in-line scan chain logic

macrohas beendivided into 16 blocks, withdifferent latch

types or layout variations in each block. A skip test scan

chain is implemented to enable each block to be tested

individually. Before testing the 16 blocks of main scan

chains, a skip test is performed to check the integrity of the

skip test scan chain. In the first case study, a chip suffered

fromskip test failure. First, LVI was performedwith the skip

test scan chain under flush test, and the LVI image showed

that the data could pass through the skip test scan chain

successfully (Fig. 6a). However, the data signal stopped

at the sixth latch, aligning with the last pack of block 1,

when the skip test scan chain was run under clock mode

test (Fig. 6b). This implied that the skip test scan chain

failure was due to a clock signal issue. Both L clock and D

clock signals weremappedwith LVI, and it was found that

the L clock signal propagation was fine, while the D clock

Fig. 5

Examplesof (a) single-latchand (b) clock-type failures

Fig. 6

LVI image for the skip test scan chain. (a) Flush test.

(b) Clock mode test

Fig. 7

LVI images at the clock buffers (D clock) for block 1 (reference), block 2, and block 3

(continued on page 10)

LVI AND LVP APPLICATIONS IN IN-LINE SCAN CHAIN FAILURE ANALYSIS

(continued from page 6)