It is well established in the semiconductor I/C industry that the proportio
n of customer field returns attributed to damage resulting from electrical
over-stress (EOS) and electro-static discharge (ESD) can amount to 40% to 5
0% (Cook C, Daniel S. Characteristics and failure analysis of advanced CMOS
submicron ESD protection structures. EOS/ESD symposium proceedings #14, Da
llas, TX, 1992. p. 147; Densen WK, Green TJ. A review:of EOS/ESD field fail
ures in military equipment. EOS/ESO symposium proceedings-10, 1988. p. 7. S
traub RJ. Automotive Electronics IC Reliability. CICC Proceedings, 1990; Eu
zent BL, Maloney TJ, Donner II R. Reducing field failure rate within proven
EOS/ESO design. EOS/ESO Symposium Proceedings #13, Los Vegas, NV, 1991. p.
59). ESD events are the subset of EOS events caused by high voltages that
are associated with electrostatic charge. Although additional hard and soft
failures can occur in the factory, these are normally screened by effectiv
e test programs. It is therefore necessary to determine the probable becaus
e of failure before cost effective corrective action can be initiated.
Distinguishing between EOS and ESD failures and differentiating the subtle
differences between damage due to the several distinct ESD models continues
to challenge failure analysis capabilities as dimensions shrink and critic
al defect sizes are reduced. Many of the damage sites are not visible with
optical microscopy. De-processing together with very high magnification exa
mination using the scanning electron microscope (SEM) is most often necessa
ry. However, the use of test model simulators to replicate the ESD events c
an most often replicate a failure signature, i.e. a unique die location and
morphology associated with the specific model (Morgan IH. ESO Failure Anal
ysis Signatures. Proceedings of the 3rd ESO Forum, Grain, Germany, 1993. p.
275).
This paper summarizes the evaluation performed on a standard programmable l
ogic complimentary metal-oxide silican (CMOS) product to ascertain the ESD
immunity. The study entailed ESD simulation using a variety of ESD models,
conducting detailed physical failure analysis and then comparing the result
s with documented analyses performed on customer field returns and factory
failures. As a result of the differences in current stress magnitude and ov
er-stress time domain, the location, type and severity of damage at the fai
lure site is known to show considerable variation (Morgan IH. A Handbook of
ESO models. AMD Internal Publication, 1992 (available from AMD literature
department upon request)). The purpose of the study was to develop a catalo
gue of failure signatures, and to determine to what extent this catalogue c
ould be used to relate a signature to electrical failure for a particular d
ie and pin function. (C) 1998 Elsevier Science Ltd. All rights reserved.