By Dr. Reza.Ghaffarian, Guest Editor

Research has shown that packages with high I/Os and fine pitches, especially under 0.8 mm, may require the use of costly microvia PWBs, which may perform poorly when assembled onto boards. From 1997-1999, the JPL MicrotypeBGA Consortium built a test vehicle (TV-1) employing 11 package types and pitches. Lessons learned were published as a guidelines document for industry use.¹

Finer pitch CSPs, which recently become available, were included in the next test vehicle of the JPL CSP Consortium.² The Consortium team jointly concentrated its efforts on building the second test vehicle (TV-2) with 15 packages of low to high I/O counts (48-784) and pitches of 0.5 mm to 1.27 mm.

Figure 1. The assembled TV-H with numerous CSP and fine pitch BGA packages is shown. The package/PWB daisy chain for the 280 I/O FPBGA is at bottom right.

In addition to the TV-2 test vehicle, other test vehicles were designed and built by individual team members to meet their needs. At least one common package was included as a control in each of these test vehicles, which enabled the team to compare the environmental test results. The use of a common package also allowed the team to understand the effects of PWB build and its manufacturing variables.

One test vehicle was designed and assembled by Hughes Network Systems, using its internal resources, and is identified as TV-H. This paper presents the most recent thermal cycling test results to 900 cycles, currently being performed under -55 to 125°C conditions for packages with various die sizes. This paper also presents mechanical fatigue test results for the 280 I/O count packages under various deflections with and without local heating.

The TV-H consisted of eight packages ranging from 48 to 280 I/Os with pitches of 0.8 mm as shown in Figure 1. The four-layer PWB consisted of two resin coated copper (RCC) layers and an FR-4 core (1+2+1) with a total thickness of 0.43 mm. Microvia technology was used.

The pad consisted of a 0.1 mm (4 mil) microvia hole at the center of pad. A non-solder-mask-design (NSMD) pad with a diameter of 0.3 mm and 0.05 mm clearance was used. The surface finish of the PWB was Ni/Au immersion with about 2-8 micro inch of gold over 100-200 micro inch Ni. No-clean solder paste for assembly was applied with a 5-mm thickness laser-cut stencil. The test vehicle was 11.9 cm by 4.6 cm (4.75″ by 1.85″) with one connector attached for continuous thermal cycling monitoring. The width of the test vehicle was cut into 1 inch for the three-point bend fatigue testing.

Figure 1 shows a fully populated test vehicle (TV-H) with two sites for the 280 I/O fine pitch ball grid array (U4 and U2 sites).

All packages were daisy-chained, and they were divided into several internal chain patterns. The daisy chain pattern on the PWB completes the chain loop into the package through solder joints. Several probing pads connected to daisy chain loops were added for failure site diagnostic testing. The package and PWB daisy chains for the 280 I/O package is shown at the bottom of Figure 2. All location sites (including U4 and U2) were populated for the thermal cycling assembly testing. Only the U4 location was populated for mechanical cycling testing. All packages were pre-baked at 125°C for 2.5 hours prior to assembly.

Figure 2. Three-point bend test specimen with deflection mechanism.

Thermal cycling was performed in the range of –55°C to 125°C. Chamber setting and thermocouple readings are shown in Figure 3. The heating and cooling rates were 2° to 5° C/min with a dwell time at maximum temperature of more than 10 minutes and a shorter dwell time duration at the minimum temperature. Each cycle lasted 159 minutes.

The test vehicles were monitored continuously during the thermal cycles for electrical interruptions and opens. The criteria for an open solder joint specified in IPC-SM-785, §7.8, were used as guidelines to interpret electrical interruptions. Generally, once the first interruption was observed, there were many additional interruptions within 10% of the cycle life. In addition, daisy chain opens were verified manually at room temperature after weekly removal of test vehicles with failed assemblies from the chamber.

A mechanical cycling test was performed using a three-point bending test set up as shown schematically in Figure 2. This test vehicle coupon had one package at its center. The center of package along the PWB was set on a stationary ram with 3.13 mm (1/8″) radius. The coupon was in contact along its edges with two moving bars, spaced at 80 mm apart, each having a 3.13 mm (1/8″) radius. These bars pushed the coupons down onto the stationary ram with specified maximum deflections for mechanical cycling.

To determine the maximum deflection of assembly under static conditions, a test coupon was tested to failure by continuously increasing deflection. Failure was automatically detected by the application of current to the daisy chain with 3.2 Ohms resistance to achieve a 4V output.

Monitoring was accomplished by interfacing a PC into the output signal from the mechanical testing system and recording data. When the daisy chain was opened, the output voltage from the system increased to 4.16 volts. Voltage output and deflection levels were continuously recorded in a data base for retrieval and data analysis.

The first three test results were inconsistent and failed unexpectedly at very low values. We made two observations: (1) The test coupon was moved slightly during cycling from its zero deflection state; and (2) the application of high current into the daisy chain introduced heat into the package.

The first observation was corrected by using a double-sided tape to secure the coupon to the bars, minimizing the coupon’s movement. The second observation corrected for, it was allowed to use current for a set of test coupons which further accelerated cycles-to-failure relative to a no-current condition.

To find the effects of current on package temperature, the package surface temperature was monitored by thermocouple before the start and during mechanical cycling. The temperature rises were in the range of 95°C-105°C.

When no current was applied, the daisy chain resistance was monitored using a multimeter. The first increase in resistance (considered initiation of failure) was reported. Test was continued until complete daisy chain open condition and this value was also recorded.

Only a limited number of test vehicles were subjected to accelerated thermal cycling condition (-55 to 125°C). The most recent thermal cycle test results (to 900 cycles) are presented here.

Cycles-to-failure (CTF) results for the 280 I/O package with 3 die sizes are compared and discussed. Preliminary test results for several other packages are presented elsewhere[2]. CTF for other packages and under other thermal conditions are being gathered and will be analyzed and presented in the future.

Twenty test coupons were used for accelerated mechanical cycling testing under deflection levels from 0.125 mm to 1.875 mm (0.005" to 0.075"). The maximum cycling deflection value represents approximately 30% of maximum deflection under static test.

MCTF for assemblies with various die sizes for single and a combination of two to three deflection levels to 300,000 cycles are presented. The effects of local thermal heating on MCTF degradation were also determined and are discussed. Degradation due to assembly exposure to high temperature representative of curing of an underfill is also reported.

Figure 3. Thermal cycle profile in the range of -55°C to 125°C

Figure 4 shows cycles to first failures for the 280 I/O FPBGA with three die sizes. It also includes the CTFs for the 208 I/O package with 11.4 mm die with an identical package technology. To generate plots, the CTFs were ranked from low to high, and failure distribution percentiles were approximated using median plotting position F_i = (i-0.3)/(n+0.4).

For this package technology, the relative die size had the most significant effect on CTF. The 208 I/O package with an 11.4 mm die size in a 15 mm package showed CTF in the range of 176 to 573 cycles. The 280 I/O package with an 11.8 mm die in 16 mm packages failed in the range of 303 to 824 cycles.

CTF data for assemblies at the center of PWB (U4 site) and the edge were also distinguished by plotting a square and a triangle. Even though most of U4 packages failed at higher cycles, it is not clear if this is due to the location of the package on the PWB or a manufacturing anomaly. Only two out of 10 assemblies with 9.5 mm die failed at 690 and 836 cycles.