Column grid array High-reliability option for packaging
08/01/2000
An enhancement of BGA technology meets the challenges of long life, extreme thermal and mechanical conditions, the use of large components, and the need for ultra-high reliability.
By Keith Sturcken
 Figure 1. CGA packaging. |
With the ever-increasing capabilities of integrated circuits have come the requirements on electronic packaging for providing higher density, higher I/O and higher performance. For most commercial applications, these requirements have been addressed through the implementation of ball grid array (BGA) packaging. Area array components provide many times more the number of I/O than peripheral lead packages, such as quad flat packs, leadless chip carriers and small outline J-leaded packages.
Consisting of an array of solder ball connections across the underside of the component, BGA can provide more interconnection in less area than traditional peripheral packages while providing improved electrical characteristics.
Unfortunately, because of the low height between component and circuit board and the use of eutectic solder material, BGA components often do not meet the reliability requirement of some applications. BGA technology is not suitable in cases where long product life, extreme temperature and mechanical conditions, use of large components or ultra reliability define the environment. Such conditions could be characterized by automotive, industrial, avionic, wireless communications or satellite applications. For these situations, column grid array (CGA), an enhancement of BGA, is an appropriate solution. CGA packaging uses taller solder columns that are often made of more flexible or compliant low-tin solder, rather than solder balls.1 An example of CGA packaging is shown in Figure 1. A tall solder column is flexible and can absorb the different thermal expansion rates between a package and a circuit board during temperature excursions. Because a package and circuit board consist of different material (such as silicon or ceramic) vs. organics (such as epoxy), the two elements expand and contract at different rates over temperature changes. This difference in expansion rates must be absorbed by the solder connection. Over a number of such temperature cycles, the effect of such induced fatigue stress on a solder joint can be very damaging.
 Figure 3. Thermal expansion inducing stress of module/circuit board interconnect. |
Figure 2 shows the difference in construction between a solder ball and a solder column. The tall solder column absorbs the stress induced from thermal expansion mismatch between an electronic package and the circuit board to which it is mounted (Figure 3).2 Testing also has shown that the low-tin material usually used in a solder column is more resistant to fatigue stress than the eutectic solder generally used in solder balls. Lead, which is a ductile material, makes the joint more flexible when it constitutes a higher percentage of the joint composition.
 Figure 4. Lognormal probability plot of accelerated temperature cycling results for BGA and CGA modules. Note from N.01 and first fail points, CGA reliability is 3x BGA. |
Actual temperature cycle testing performed on BGA and CGA components mounted to circuit boards of the same material show that CGA components have more than three times the life of BGA components. The results of accelerated temperature cycle testing (temperature cycle testing using high temperature extremes and fast cycle times to get results in a reasonable time frame) are plotted in Figure 4. For the test temperature cycle, the first BGA component had a solder connection failure after 222 cycles, while the first CGA component did not experience any failure until after 622 cycles. Figure 5 shows a BGA connection fractured after 300 cycles and a CGA connection flexed, but not fractured, after 800 cycles. The accelerated temperature cycling test results can be related to actual application environments such as telecommunications satellites. While the temperature extremes of such applications are usually much less severe than those in the experimental testing, the program life is many times longer (greater than 10 years). Using an equation like the Coffin-Manson relationship3 the test results can be used to predict field life. The plots in Figure 6 show how CGA packaging technology can extend the life of a product from a few months to many years.
 Figure 5. Actual CGA and BGA temperature cycle test samples showing the dramatic difference in reliability between BGA and CGA. |
Additional mechanical vibration and shock testing show that the CGA technology also is able to survive this type of stress at least as well as BGA. Examples of this type of stress include launch profiles for telecommunication satellites and engine-induced vibrations in automotive applications.4
Columns are being used in place of balls on modules with case outlines ranging in size from less than 20 mm to greater than 40 mm. While the smaller modules using columns are typically used in high-reliability applications, such as satellites, it is generally thought that modules 32 x 32 mm and larger require columns even in commercial applications. Typical CGA module sizes and corresponding I/O are shown in Table 1. It can be seen that even with surface array packages, the I/O requirements of processors and interface devices in excess of 1,000 will drive component size to 30 x 30 mm and beyond. Like BGA, the commonly used I/O pitch at this time is 1.27 mm (.050 inch); however, a 1.0-mm (.040-in.) column pitch has been qualified by some suppliers. The 1.0-mm column pitch will likely become the norm in the near future, particularly for higher I/O applications. There is no limiting factor to achieving tighter column pitch at the package level. The main technical consideration is the ability of circuit board technology to support this level of density.
CGA Production
Recognizing the need for CGA packaging technology, a number of techniques have been developed for their production. As with any product, for CGA packaging technology to be successful, CGA packages must be manufactured using fully established and repeatable processes. Key elements include:
- Good, repeatable attachment of the columns to the package
- Uniformity of the joined columns in terms of straightness, pitch (periodicity) and column height
- Adaptability of the CGA process to components of different size and I/O counts. In particular, the CGA process should not be a limiting factor in the total number of connections that can be made for the module.
These elements can be achieved by a process that has the capability to:
- Screen solder paste onto the connection pads of the package to provide column joining. Consistent, uniform depositions of the solder paste can be achieved through controlling the weight of solder paste screened onto the package I/O pads.
- Align a full array of columns to the corresponding pads of the package
- Join the columns to the connection pads of the package through a solder reflow process
- Trim columns to ensure uniform height, which is essential to creating proper solder joints at the board level.
A proper tool set and fixtures can ensure that such a process be attained in a high-quality, repeatable, time-effective manner. Low-tin solder columns that do not reflow during the joining operations can be procured from a reputable metals-processing supplier. Properly designed fixturing can allow for different package sizes and ensure that good solder column joining is achieved on every connection. Depending on the volume required, various levels of automation can be achieved for the CGA process. Process controls and tests, such as periodic column pull and final visual inspection, further ensure that high quality is maintained. By using a robust solder-joining process, if necessary, attached columns can be removed and replaced, further enhancing a high-yield process. Figure 7 shows a flow of the CGA joining to package process.
 Figure 6. Life expectancy of BGA vs. CGA modules in actual satellite environments. Predictions are based on actual experimental test data. |
CGA was introduced to provide a high-reliability alternative to BGA. At present, there are many applications using BGA that could benefit from this attribute of CGA. For this reason, it is desirable to be able to convert BGA components into CGA components - in effect, providing a plug-compatible module at the circuit board level with vastly improved reliability. Such a capability is now available using existing valid processes for ball removal and pad dressing. Once such processes have been completed, column joining can be undertaken using the process described above.
Alignment is Key
The keys to successful, reliable column assembly are good alignment of the columns to the corresponding pads on the component and a well-formed solder fillet. For CGA components with 0.05-in. (1.27-mm) pitch, the columns should be 0.0225-in. in diameter. Mounting these columns to component pads 0.030-in. in diameter offers the best opportunity for proper alignment and fillet formation. There are more than 30 specific inspection criteria a CGA joint should meet. Essentially, the fillet should extend around the perimeter of the column, and exhibit good wetting to both the column and pad with a concave profile. The solder in the fillet itself should be shiny, not grainy.
 Figure 7. Module assembly processes for column-joining operations. |
Like BGA components, if necessary, CGA modules can be tested at the land grid array (LGA) level prior to the joining of the columns. The module at the LGA level is a very robust package virtually immune to handling damage, making it a highly manufacturable product.
Transport and Mounting
Once the CGA component has been assembled, a fully developed set of protective carriers will ensure that column integrity is maintained through handling, shipping and storage until the components are to be mounted to circuit boards. Such transport containers, often designed to a JEDEC standard, usually accommodate 10 to 12 components in a multi-component tray configuration that may be procured from a molded plastics supplier specializing in such carriers.
Once the columns have been joined and trimmed, the component can either be immediately mounted to a circuit board or stored for future use. CGA technology is compatible with standard surface mount technology (SMT) processing for a circuit board. A standard eutectic solder paste should be screened onto the circuit board I/O with a 0.007-in. thickness above the copper pads. This extra thickness will ensure a good fillet and compensate for any variation in column height. If the columns were originally joined to the component using standard eutectic solder, the SMT reflow process will not adversely affect these joints because the surface tension of the solder will maintain the joint's integrity. For 0.0225-in. diameter columns on 0.050-in. pitch, circuit board I/O pads should be 0.032-in. in finished diameter. If the circuit board assembly is to be coated with protective organics, care should be taken to make sure the encapsulant does not build up around the columns of a component. This will reduce the free length of the columns and reduce overall reliability. A bead of organic, such as silicon, around the edge of the CGA components will prevent this problem from occurring.
If a CGA component needs to be reworked or replaced, the component can be removed using techniques such as localized forced hot-air reflow. The site then can be dressed, and a new component can be placed again following SMT procedures. The board-mounting process for CGA is similar to that for BGA.
Application Considerations
CGA technology is essentially a packaging platform that can accommodate either flip-chip or wirebond mounting of devices. Its full potential as a high-density interconnect (HDI) vehicle can be fully realized when the device it contains is flip-chip mounted. The commonly used method for this type of interconnect is "C-4" solder balls placed in a high-density array between the chip device and the package. It provides for very high counts of signal connections along with power and ground I/O that can, in turn, be supported by the density of CGA while not forcing the device itself to be enlarged. In addition, CGA also will support even more advanced designs, such as multichip module (MCM) layout and other alternate forms of high-density interconnect. Because of its ability to extend the size of chip level, it not only offers high reliability but also a new stage of integration at both the board and component level.
While CGA offers improved reliability over BGA, the electrical characteristics are almost equivalent. Some basic parameters, such as inductance, increase less than 10 percent but are more than made up for by the additional number of power and ground I/O that become available with CGA.
For larger components, it is appropriate to consider CGA interconnect on plastic packages. This can be especially true if standard epoxy glass board material is used, and the component has a low coefficient of thermal expansion (CTE) because of a large silicon die contained in the package that would affect the component's expansion rate during temperature cycling.
Summary
By using a properly installed and qualified tool set, CGA technology is a highly manufacturable, high-yielding, cost-effective electronic packaging option. When columns are joined to the package to the specified criteria and controlled board-level processing is used, CGA packaging is fully compatible with standard SMT assembly techniques. CGA packaging is a high-density, fully productionized platform that supports MCM and HDI technologies. CGA provides a viable electronic packaging solution for any high-reliability, high-performance requirement.
References:
- R. David Gerke, "Ceramic Solder Grid Array Interconnection Reliability Over a Wide Temperature Range," Motorola C4 Product Design Center, Austin, Texas.
- Donald R. Banks and R. David Gerke, "Assembly and Reliability of Ceramic Column Grid Array," Surface Mount International Conference & Exposition, San Jose, Calif.,1994, p. 271-276.
- Raj N. Master, Marie Cole, Greg B. Martin and Alain Caron, "Ceramic Column Grid Array for Flip Chip Applications," ECTC, Hopewell Junction, N.Y., 1995.
- Arvind Sinha, Phil Isaacs and Tim Tofil, "Mechanical Reliability Analysis of Ceramic Column Grid Array Solder Joints," IBM Journal of Research and Development, April 1997, p. 133-141.
KEITH STURCKEN, senior electronic packaging engineer, can be contacted at Lockheed Martin, 9500 Godwin Drive, Fl. 3, Manassas, VA 20110-4166; 703-367-4948; Fax: 703-367-3540; E-mail: [email protected].
null
Determining reliability of module solder connections during temperature excursions
 Table 1. Current CGA product offerings. |
To provide reliability results in a timely manner, experimental temperature cycling employs an acceleration process. This temperature cycling acceleration is achieved by maximizing temperature extrema and accelerating cyclic frequency. Whereas an actual application might operate in a temperature range of about 20°C with a maximum temperature of 60 to 80°C, the effect of such temperature excursions on the life of a component's solder joints can be determined by using a much more aggressive temperature range. Likewise, a product might experience such a temperature excursion one to six times a day. The effect of the frequency of these temperature excursions can be determined in a timely fashion by using a test frequency of 12 to 48 cycles a day. Even in product applications requiring that temperature extremes from below 0°C to above 100°C be met, the likelihood of such a temperature cycle being experienced is remote. A typical experimental test cycle for a temperature range from -55 to 105°C with a frequency of 12 cycles per day can be highly effective in predicting the life of component solder connections in product applications. An example of such an accelerated test temperature cycle is shown in Figure 8.
 Figure 8. Monitored temperature for an experimental accelerated-temperature cycle. |
The results of such testing can be tracked through either continuous monitoring of the solder connects or periodic inspection. This data may be best understood by plotting a probability distribution on a lognormal scale. What is plotted is the cumulative module fails as a function of cycles (Figure 9). Experimental temperature cycling is typically run until 50 percent or N50% of the test sample has failed. Upon completion, the cumulative fail data points are plotted against their cycle readout points and a linear regression method used to fit a curve to the data. Two sets of data are shown on the plot presented here. The first plot shows early data with an aggressive temperature cycle. The second set of data uses a temperature cycle more typical of board-level testing. The test hardware in the set of data also consists of a non-woven Kevlar epoxy board technology designated as BI by JEDEC that has a lower coefficient of thermal expansion than most circuit boards at about 10 ppm/°C. The second set of data used hardware built with a more refined assembly process. Both sets of data are valid and may be used in determining CGA reliability, but the factors mentioned above should be taken into account when establishing this reliability.
 Figure 9. Lognormal probability distribution plot of experimental temperature-cycling of CGA modules. |
Of particular interest is the N.01% point, which equates to a fail rate in a field population of one part in 10,000. At this rate, the mechanisms being evaluated are not considered a determinant in the overall reliability of the module. The cycle count at which this point occurs is a key result of the experimental temperature cycling.
Relating Test Results to Field Conditions
 Figure 10. Coffin-Manson relation correlating test cycles to field cycles. |
The experimental temperature cycling results can be related to a field condition through the Coffin-Manson relation (Figure 10). Empirically derived, this relation has been historically proven to relate experimental results to field conditions. By relating key elements of temperature, frequency and applied stress because of thermal expansion, the experimental cycle count of the N.01% can be related to the field cycles, which in turn can be related to field life in terms of days or years.
An example of relating test results to field conditions is given in Figure 11. As can be seen in this example, depending on field conditions, the test cycle results can relate to many years of life (in this case, 95 years). This is because solder fatigue is very dependent on cycling rates and temperature ranges. Using the latest test data presented in this article, the field life on the CGA technology could be significantly greater. Similarly, the experimental test results can be related to other field conditions using the Coffin-Manson relation in a like manner. A more severe field condition would obviously have a shorter predicted life. This reliability calculation is based on a statistical analysis and provides an approximation of predicted field life; however, these predictions are based on actual experimental test cycling.