Manufacturing with 0201s: the latest developments: placing and inspecting these tiny components are still major challenges. But our authors reveal a process to drop in 0201 with near Six Sigma results.

The use of 0201 components is an enabling technology to further the miniaturization process of electronic products. Many consumer product development teams have been interested in using 0201 components but have not yet implemented them on a large scale. Issues with the supply chain, factory equipment capability, the soldering process and high-volume production robustness are all factors in slow implementation.

One main concern is the capability of placement equipment to place 0201 components. Most factories currently use a process window of [+ or -] 0.15 mm for their machine capability studies. However, prior process research on 0201 components has shown that this window needs to be reduced to [+ or -]0.10 mm to avoid excessive defect rates. As a result, all the major equipment suppliers need to improve their equipment's placement control systems, as compared to those systems used for 0402 components.

This paper will discuss some of the challenges required for placing and inspecting 0201 components. The study also concentrates on developing a process to drop in 0201s by developing the correct padstack to yield a Six Sigma process without altering the stencil thickness or paste type.

Placement Equipment

The capability of chipshooters has steadily improved with exposure to 0201 production experience over the last several years. Placement is still the most critical of the 0201 front-end processes with a multitude of important variables. The variables can be comprised from various subsystems such as: camera (positioning, calibration, resolution and scaling); vision recognition; nozzles; vacuum; x, y, theta, and z axes for placement and pickup; feeders and printed circuit board (PCB) support.

A chipshooter may be specified to place 0201s, but it may not have the subsystem accuracies to support a process window of 100 microns. As one example, we found in our research that camera pixel resolution and scaling of a major vendor were not addressed or not sufficient to support the anticipated requirements. The accuracy results with the scaling issue are shown in Table 1.

To address this deficiency in the short term, the component library data had to be altered from true data while the pixel scaling calibration issue was addressed. Table 2 shows the accuracy results after altering the library data to account for the incorrect camera scaling. The long-term approach was resolved by integrating improved placement head/camera systems.

Cost Adders for 0201 Manufacturing

Added placement equipment cost related to the support of 0201 components can be divided into capital, consumables and resources. Resources include the access to the equipment for increased critical preventative maintenance, process development and troubleshooting, and the added technical support. The related capital costs are new capital purchases, upgrades, feeders and advanced calibration tools. Related consumable costs would be nozzles, increased and hypersensitive wear components, filters and grease. Hypersensitive components are items that would otherwise function to acceptable levels for non-0201 production. Related capital items would have to be assessed on an older machine for possible replacement or rebuild, while new machines need to be monitored for wear degradation verses cycles or production hours.

In our experiment, when upgrading a production chipshooter from normal non-0201 placement to support the 0201 process, the machine's subsystem status needed to be assessed. This process was performed on a machine that had approximately 230 million placement cycles and that was 4.75 years old at the time. The machine was originally specified to be 0201 capable.

The machine's accuracy was quantified, and a review process was set up with the original chipshooter manufacturer to define critical measurable settings that were required before 0201 placement would begin. The overall original accuracy was quantified to be 1.41 Cp/1.35 [C.sub.pk] for the x-axis and 1.35 Cp/0.97 [C.sub.pk] for the y-axis with limits of [+ or -] 100 microns, With this result, further measurements were completed to measure head and x-y table subsystems. One head and the y-axis ballscrew were found to need replacement.

Advance calibration tools and jigs were used to verify the camera subsystem and dialing in of the total system. The resulting accuracy measurement was 2.09 Cp/1.95 [C.sub.pk] for the x-axis and 1.96 Cp/1.58 [C.sub.pk] for the y-axis with limits of [+ or -] 100 microns. Wear items were replaced in the heads and vacuum solenoids as a precautionary measure due to the hypersensitivity of the process. The total cost to refurbish and outfit the machine to 0201 capability was approximately $25,000 in parts and labor. This amount covered nozzles but did not cover the purchase of any cameras or 0201 feeders.

Overall, the cost to retrofit an older chipshooter for 0201 capability can vary widely depending on which components need to be replaced or upgraded. Cameras and vision systems alone might drive the cost to about $50,000. Cost is greatly dependent on what needs to be upgraded and what wear items need to be replaced.

Inline Inspection Equipment

Going forward, the level of sophistication to improve and take the 0201 process to the next level will require more data. That data will come from the metrology capable in-line inspection machines, both solder paste and component placement inspection. With these high-end systems, variable data can be obtained verses the low-end system, which only logs pass/fail criteria. This high-end data rich information can be used for more in-depth statistical analysis but comes at a price: High-end systems typically cost $180,000 for a base machine, excluding repair loops and off-line programming software.

Putting solder paste and component placement inspection in each line can represent significant additional capital. Factories that cannot put inspection in every line may want to use an offline, highly accurate measurement system as a minimum requirement. For the factory with multiple inspection machines, the offline optical measurement system is a useful, independent tool to ensure the proper operation of inspection machines. The cost of the offline system is about $100,000 depending on the table size, advanced accuracy requirements and other options.

Inspection machines will immediately highlight the changes to equipment, materials and processes with statistical process control (SPC) and trend analysis. Ensuring a more stable environment will be the basis for improvement. The 0201 production will require inspection and measurements to quantify all process improvement work, while ensuring a predictable production environment. These inspection machines will give advance warning to a hypersensitive 0201 process. The inspection machines will quantify and separate the process indicators for solder paste and placement, allowing simplistic troubleshooting of any catastrophic failure that may occur.

Selection of the in-line machines for both solder paste and component placement inspection will be critical to the effectiveness of the data collection. In the past, inspection companies have had difficulties delivering on capabilities for anything but a dedicated line and product. With the 0201 process window anticipated to be between 75 to 100 microns, the measurement system should be 10X improved compared to the process requirements. The inspection machines to be metrology capable will have to be accurate to 7 to 10 microns.

Pixel resolution will also be an important factor with decreased component and solder deposit size. Many other system factors are important for the viability of a metrology capable inspection machine. Those factors may not be readily apparent after the machine is installed but may be realized after a prolonged exposure to a high-volume/high-mix environment. Programming, robustness, balance of low false accepts and low false failures are just a few factors.

Padstack Development

The process variables governing the mass placement of 0201s do not differ from other components such as 0402s. However, process variables for 0201s become more sensitive to causing defects. The study outlined here concentrates on developing a drop-in process for 0201s by developing the correct padstack to yield a near Six Sigma process without altering the standard stencil thickness of 125 microns.

Experimental Procedure

Five padstacks were developed using best-in-class practices--varying from metal defined, mask defined and mask/metal defined--all of which had to accommodate a high-density interconnect (HDI) microvia in the metal pad. Figure 1 shows the different padstacks used for this experiment.

[FIGURE 1 OMITTED]

Three solder pastes were used:

* Paste 1: Flux A type III, type V blend

* Paste 2: Flux A type III

* Paste 3: Flux B type III, type V blend.

Test boards were created with several hundred of each pad type with varying spacing options: 0.38 mm, 0.3 mm, 0.25 mm and 0.2 mm. Spacing was not studied as a factor of the padstack but as a design/process limitation.

Parts were placed with varying offsets in ram: (0, 0), (0.5, 0.5), (0.1, 0.1). The effect of the offset is to simulate an out-of-control process. With the increasing offset, more defects are expected, thus allowing for statistical analysis of the defect data. The machine placing the components was running at a Cp of 2.75 and [C.sub.pk] of 2.5 to [+ or -]0.1mm window throughout the experiment. The Cp and CpK measurements were taken for every board; the placement machine did not waiver from these values throughout the experiment.

The defect level for each offset was noted. The lower the defect levels at the larger offset, the more robust the padstack was. The location of the parts post-reflow was measured using laser inspection machine. The better the self-centering effect during reflow, the more robust and more effective the padstack was. For this particular experiment over 300,000 components were placed, and experiments for both resistors and capacitors were carried out. Only data for capacitors are reported; however, the results were similar for resistors.

Gage Repeatability and Reproducibility

Gage repeatability and reproducibility (gage R&R) was performed on the measurements for both placement in paste and post reflow. The pasted board was measured immediately after placement three times, removed from the machine and placed back by a different operator, and the measurements repeated. The time between solder print and final measurement was three hours. The same procedure was carried out on reflowed boards, but the operator variable was time. The boards were measured on day one, and then after three days, during which time the inspection machine was used for other measurements.

The results were acceptable for both components in paste and post reflow. Table 3 shows the output for the post-reflow part inspection. The gage R&R is 6.45%.

Results

The theory behind the post-reflow data is that the better the padstack, the better the post-reflow centering data. The measurement program outputs the distance from the computer-aided design (CAD) centroid. Note that, where components are missing from the board, the line item is ignored, potentially skewing the analysis. However, these data become insignificant on three of the padstacks due to the vast quantity of data. The three padstacks are shown in the defect charts in Figures 2 and 3.

[FIGURES 2-3 OMITTED]

The data analysis is a simple crossed Design of Experiments (DOE), including the interaction of paste and padstack as the factors and post-reflow location of the part as the output. Figure 4 shows the prediction model for the various padstacks and solder pastes.

[FIGURE 4 OMITTED]

The self-centering results show paste type to have little effect. However, the padstack plays a critical role in the self-centering effect. Of the pads shown, PanM_0201 performs self-centering the best. Mmod and Lib2 performed the worst.

Defect Rates

Defects were counted and are shown in Figures 2 and 3. The data for two of the pad types were ignored in the chart due to the large number of defects that occurred, which obscures the chart.

The defect data only half confirm the prediction model for self-centering. The defects from Lib t and Lib2 were so high that the parts were not on the boards after reflow and could not be measured. In many cases the parts had drawbridged so that the part rested on the reflowed solder at one end but did not form a joint. As a result, the inspection machine was not able to record a location for the part.

Blended pastes were also noted not to be a significant variable in producing defects. Both pastes #2 and #3 produced very few defects at the 0.05 mm placement offset. Paste #1 was from a solder paste supplier where the flux vehicle was not optimized for the Type III/Type V blended powder.

Conclusion

Mot_020l is the ideal padstack and should be used with paste #1; however, the pad design prevented component spacings closer than 0.38 ram. The desire to drive spacings closer and realize the full benefits of the 0201 process negates this option. The PanM_0201 padstack and Mmod_0201 resulted in two options for the choice of padstacks. PanM_0201 had zero defects with a 0.05 mm offset and is a good choice for implementation. Since the defect rates increased significantly from the 0.05 mm placement offset to the 0.1 mm placement offset, this highlights the need to use tighter process margins.

The experiment proved that a process can be developed that allows 0201 placement with acceptable yields. However a true sigma level cannot be determined until the process is run in full production. The data suggest that, with correct machine and feeder maintenance and a stencil wipe every print, defect levels close to Six Sigma should be possible.

Also recommended is inline inspection to monitor the solder print deposition and the placement accuracy in the use of 0201 s in manufacturing. This inspection will be required until placement' machines can truly have Six Sigma process with a 0.075[micro]m placement window.

           X             Y
Cp        2.09          1.97
Cpk       1.79          1.66
Mean      14.8[micro]m  16.0[micro]m
[sigma]   16.2[micro]m  17.2[micro]m

TABLE 1: Initial data with inaccurate camera scaling for 0201s.

           X             Y
Cp        3.02          2.47
Cpk       2.65          2.01
Mean      12.6[micro]m  18.9[micro]m
[sigma]   11.2[micro]m  13.7[micro]m

TABLE 2: Altered 0201 library information correcting
Camera pixel scaling factors.

                              Measurement Unit    %Tolerance
                                Analysis
Repeatability (EV)              0.0058167            5.8167
Reproducibility (AV)            0.0027652            2.7652
Operator Part Variation (IV)    0.0003561            0.3561
Gage R&R (RR)                   0.0064504            6.4504
Part Variation (PV)             0.0748410           74.8410
Total Variation (TV)            0.0751185
Sigma Multiple                  6.0000000
Tolerance                       0.1000000

TABLE 3: Gage R&R.

This article is a shortened version of a paper presented at 2002 SMTA International, Chicago, IL. For the full version, access www.circuitsassembly.com/online/0305/0305index.shtml.

Reference

(1.) Butterfield, Andrew & Kevin Pieper, "0201 Solder Process Study", Motorola 1999 Hermes Symposium, Austin, TX, July 1999.

Andrew Butterfield is a senior staff engineer with Motorola's Infrastructure Group, Fort Worth, TX; [e-mail]. Mark Guilford is a principal staff engineer with Motorola's Personal Communications Sector, Libertyville IL; [e-mail]. Kevin Pieper is an engineering section manager with Motorola Advanced Product Technology Center, Plantation, FL; [e-mail].