Page 146 - Wire Bonding in Microelectronics

124 Cha pte r F o u r

normally be detrimental to the subsequent operating life of the device.
A reliability problem would arise only if severe high-frequency vibra-
tions (such as ultrasonic cleaning) were encountered or if the bonds had
low loops and were subject to temperature cycling (see Secs. 8.3 and 8.4).
Forty years after the invention of the nondestructive pull test at
Autonetics (Rockwell), the idea of the nondestructive pull is still con-
troversial. Some people worry about possible metallurgical damage to
the neck or heel of the bond, and others are concerned that the hook might
hit and damage an adjacent wire as it is being positioned. (Note dis-
cussion of the substitution of SPC during bonding for the NDPT on
high pincount packages at the beginning of this section and in Sec. 4B.1.)
At the time of this writing (2008) there have been hundreds of mil-
lions of nondestructive wire pull tests performed, [4-57], which have
been a requirement for some military and space (K) devices. All of the
evidence available indicates that the test is nondestructive. In addi-
tion, it has been shown that the NDP test does not lower the bond-
force distribution of devices that later undergo the usual military qualifi-
cation tests of temperature cycle, burn-in, shock, and vibration [4-58].
With regard to damage to adjacent wire bonds (on a single-tier pack-
age), a trained operator is less apt to damage a wire with the hook
while positioning for a pull than an equivalently trained operator is
to misplace or otherwise damage a wire while actually making a bond
with a manual bonder. Automatic nondestructive pull testers have been
made to specifically avoid touching adjacent wires. In such cases, the
hook turns parallel to the wire for placement and then perpendicular
to it for pulling. (However, even this is not adequate for fine pitch or
packages with bonds in several tiers—where SPC must be used.) The
NDPT is often used for automatically testing large diameter wedge
bonds in high volume assembly.
The nature of immature and otherwise poorly bonded interfaces
has been fully described in Chap. 2. They consist of a series of uncon-
nected microwelds. When an appropriate force is applied, the interface
begins to separate, first breaking the microwelds nearest to the bond
heel, resulting in a crack. This crack propagates rapidly along the
microwelded interface with characteristics similar to those of a (modi-
fied) “Griffith crack” and completely breaks the interface within a few
milliseconds. If the force is below a threshold value (too low to break
the first few microwelds at the heel), then no break or damage occurs
to the interface. Thus, the NDP test is largely a go, no-go test, and any
possible marginal damage can be assessed by the statistical methods
outlined above.

4B.4 Limitations of the NDP Test
Regardless of all of the comments above, the user of NDP-tested
devices must be aware of the limitations of this test. The test will only
perform one function. It will remove weak, poorly made bonds with pull