AN1231 Motorola / Freescale Semiconductor, AN1231 Datasheet - Page 16

• Manufacturer Part Number: AN1231
• Description: Plastic Ball Grid Array (PBGA)
• Manufacturer: Motorola / Freescale Semiconductor

AN1231
16
leaded devices reveals that the PQFP and PLCC tend to last
longer. For example, N 50% for the 68 PLCC is a factor of
three greater (6561 versus 2171 cycles) than the 72 PBGA
for identical cycling conditions. But, it can be seen in Table 7
that the ’s associated with leaded device data are signifi-
cantly lower, on average, than those of the PBGA. This
greater spread in the leaded device data can be attributed to
a greater variation in the factors that influence solder joint
reliability such as solder volume and device coplanarity. This
data and when subsequent extrapolations are made down to
N 0.1% or lower, they tend to become comparable to the same
estimates for PBGA. Extrapolating N 0.1% for the 72 PBGA
and 68 PLCC mentioned above, yields 1058 and 904 cycles,
respectively. The same thing can be done when comparing
the 225 PBGA to the 208 PQFP although the thermal cycle
on the leaded device was the more severe – 40 to 125 C.
The 208 PQFP had a higher
are compared there is a much different situation (4685 cycles
PBGA versus 639 cycles PQFP).
under and proximate to the die perimeter tend to fail first.
This can be seen in Figure 14 which compares the outer, die
perimeter, middle, and inner rows of the standard 361 pin
PBGA. This gets back to the previously mentioned mismatch
between silicon and the other PBGA and FR4 PCB materi-
als. To further prove this point, tests on 72 pin packages with-
out die went for many more cycles without failure than
identical packages with die. Another conclusion drawn from
Motorola testing is that there is a relative cycle basis accel-
eration factor of about 3.5 between 0 to 100 C and – 40 to
125 C PBGA testing. This actually results in no relative
disparity means that there is a greater spread in the leaded
Company
Comparing the cycles to 50% failure for the PBGA and
Another observation from Motorola testing is that rows
Motorola
Motorola
Motorola
Motorola
Motorola
Motorola
Motorola
Compaq
Compaq
Compaq
AT&T
AT&T
72 PBGA (270x270)
72 PBGA (270x270)
119 PBGA (280x437)
225 PBGA (400x400)
361 PBGA (450x450)
68 PLCC, Cu Leadframe
208 PQFP, Cu Leadframe
169 PBGA (364x359)
225 PBGA (389x398)
72 PBGA (270x270)
165 PBGA (437x437)
225 PBGA (389x398)
Device (Die in mils)
Table 6. Reliability Predictions Using the Lower 95% Confidence Limits
From Best Fit Line
95% Lower Confidence Limit
Table 7. PBGA Accelerated Thermal Cycling Data from Motorola
and N 50% , but when N 0.1% ’s
and Others(with Comparisons to PQFP/PLCC)
for the Weibull and Lognormal Distributions
– 40 to 125, 1cph
0 to 100, 3cph
0 to 100, 2cph
0 to 100, 2cph
0 to 100, 2cph
– 40 to 125, 1cph
– 40 to 125, 1cph
0 to 100, 2cph
0 to 100, 2cph
– 25 to 100, 2cph
– 25 to 100, 2cph
– 25 to 100, 2cph
Cycle ( C)
N 0.1%
1058
3013
4459
4685
6319
2103
1412
1734
1010
1252
904
639
Predicted N 0.1% (in Cycles) Using:
Weibull
acceleration on a time basis since the more severe cycle is
approximately three times longer than the 0 to 100 C cycle.
Also observed from the testing was that underfill (although it
causes the devices to be unreworkable) can give up to a 4X
increase in cycles to failure and that increased device stand-
off (obtained through using larger diameter solder balls)
improves the fatigue life as would be expected.
Failure Analysis
cycling or in an application can provide extremely useful
information with regard to determining the failure character-
istics and mode. Generally, the failure analysis of PBGA
devices consists of resistance probing to locate specific
failed joints when possible, cross-sectioning and also die
penetrant analysis of fractures. Probing of individual joints
(actually joint pairs) usually only pertains to daisy-chain
devices on test boards that have vias integral to each solder
pad which drop down to the bottomside of the board and can
be probed. Knowing the schematic of the daisy-chain net,
specific via pairs can be probed until a failing pair with atypi-
cal or infinite resistance is found.
PBGAs with solder joint failures. Vibrations and flexure
caused by the cutting of a failed device out of a test board, if
not performed properly, can further propagate or initiate
fractures. High speed diamond blades, abrasive wheels or
routers are recommended over band saws. Once the device
is removed, setting of the potting compound should optimally
take place in a vacuum. Slower curing potting epoxies are
also usually better than the quick setting variety to ensure
that the PBGA is completely underfilled. After grinding,
polishing and etching to reveal the solder structure, fractures
4685
4121
Analysis of devices that have failed in accelerated thermal
Care must be taken when preparing cross-sections of
N 1%
1363
4034
5848
5592
7804
1818
1553
2741
1993
2194
1309
1664
N 50%
11495
2171
6895
9683
7737
6561
7912
4459
3749
3379
2106
2807
Log Normal
5659
5239
10113
11886
(N 63% )
2260
7222
7958
7332
9111
4651
3960
3508
2195
2937
MOTOROLA FAST SRAM
13.0
10.9
9.1
7.9
8.4
3.3
2.6
8.7
6.7
9.8
8.9
8.1
Internal
Internal
Internal
Internal
Internal
Internal
Internal
Semi/HDP-10/93
Semi/HDP-10/93
1993 IEPS
1993 IEPS
1993 IEPS
Source