A mass measurement system for screen printing: several experiments have proven mass measurement's viability for monitoring the printing process.

Measuring screen-printed deposits on a printed circuit board (PCB) has been a struggle since the first board was printed. Many systems have tried to accurately measure the deposited material, and current measurement systems rely on complex camera or laser systems for this measurement. A major issue for these systems is the reference. Whether the reference is a non-printed pad, conductor trace or the surrounding material, it is forever changing and, therefore, a moving target.

Defining a reliable reference and the systems' complexity means adding cycle time to ensure accurate readings are made. Inspecting only the critical areas of the board is common, which may lead to poor product being manufactured due to the wrong area inspected. Most inspection systems require coordinate programming to determine the area being inspected. Whether this programming is done manually or with the aid of Gerber data, it is a time consuming and sometimes frustrating task.

A mass measurement system (MMS) technique utilizes technology that has been employed for years in the scientific and manufacturing sectors to control processes. By using the actual weight of the raw board and comparing it to the post-print weight, accurate data may be used to monitor the control of the process. This simple, yet effective, system yields: improved cycle time (potentially zero effect); low cost; simple operator involvement; faster changeover and flexibility for different applications.

The concept of this system is easy. Simply weigh the raw board prior to printing; then weigh the board again after paste has been deposited. Subtracting the difference yields the actual paste weight.

The mechanics of testing the concept are also simple. A bench-top scale can be used to accurately weigh the pre- and post-board weights, and a simple spreadsheet can be used to collect and collate the data. The data can then be imported into statistical process control (SPC) software for analysis.

This article reviews the steps taken to prove the concept of mass measurement as a viable process monitoring system. Additionally, the MMS concept was implemented at three electronics manufacturing services (EMS) providers: Benchmark Electronics International (BEI, Winona, MN); Celestica (Rochester, MM) and Pemstar (Rochester, MN). These companies contributed real time production data, which was collected using MMS, end-of-line defect data and other data collection systems currently in use. This implementation's goal was not to interfere with current production throughput or procedures, and the data collection was invisible to the current systems in place.

Mass Measurement Experiment

This mass measurement experiment involved an in-depth analysis of the system using sophisticated equipment. A comparison test was developed to compare the results from the MMS and a 3-D laser scanning system. Utilizing a common memory-type product, the laser system was programmed for 100 percent of the pads on the board.

Once the board was programmed, the boards were printed. Screen printing issues were thrown in to document each system's reaction. The following print parameters were used:

* print speed: 20 mm/sec.

* squeegees: 60[degrees] metal

* stencil: stainless 0.006 in. thick, laser cut

* paste: common no clean

* print gap: 0

* separation speed: 2 mm/sec.

* squeegee pressure: 7 kg

* tooling: magnetic block.

Collection of the data was fairly simple, with the only cumbersome issue being the waiting time of 8 min. for the laser system to inspect the boards. The scale has a stabilization time of 6 sec. or less. Once all the data were collected, the real work began. The laser system's reams of data had to be compiled so that the laser system compared to the MMS. The results were compared to tolerance limits, and this information was displayed in a traffic light format. Limits were determined by data collected during the "best run."

The results are shown graphically represented by a traffic light and a color-coded representation of the board. Additionally, the laser data was separated into 100 percent inspected and a regions-of-interest (ROI) area. Since most production runs cannot wait 8 min. to inspect 100 percent, the ROI area consists of typical critical sites selected to represent the entire board. The theoretical time for ROI is approximately 1 min.

The first traffic light(top) represents mass measurement. A + sign indicates if the weight of the paste was high, and a - sign indicates a low weight. The second traffic light (middle) represents 100 percent 3-D measurement. The third traffic light (bottom) represents nine ROIs typical of sites chosen to maximize inspection coverage while maintaining throughput. The sites chosen are the top, bottom and middle component groupings on the left, right and middle boards.

Most stable print deposit

Figure 1 shows the boards chosen to represent the most stable print deposit. Tolerance and control limits are calculated from these hoards. Note that a green color indicates the data are within the control limits; yellow indicates the data are between the control limits and the tolerance limits; and red indicates the data are outside of the tolerance limits. The following was observed:

[FIGURE 1 OMITTED]

* 100 percent 3-D: As expected, the first board printed was typically below volume due to priming of the stencil apertures required for optimum repeatability.

The laser system clearly picked this defect up; however, the following boards would most likely be passed as the out-of-control situation was on non-critical areas.

* ROI: The ROI did not pick up any out-of-control situations.

* MMS: The low volume of the first board was picked up.

Purposely offset from stencil

Figure 2 represents boards that have purposely been offset from the stencil. The offset is a theta rotation of -100 arc-sec, an x offset of 0.072 mm, and a y off-set of 0.124 mm. The following was observed:

* 100 percent 3-D: Note that the 3-D system was only set up for volume variation, not alignment. With full 3-D, out-of-tolerance conditions occurred on the first board printed, possibly caused by a poor localized gasket due to the misalignment.

* ROI: The ROI indicated out of control at the fifth board when an out of tolerance was picked up.

* MMS: Initially, a low mass was indicated, likely due to the priming of the first print. Volume increased as the number of prints increased due to smearing of the deposit. This smear was caused by a poor gasket due to the alignment offset.

[FIGURE 2 OMITTED]

Increased print speed

Figure 3 represents boards that were printed with an increased print speed from 20 mm/sec, to 75 mm/sec. Eventually, the paste stuck to the squeegees on the left side, and no paste was deposited on that side of the panel. The following observations were made:

* 100 percent 3-D: Immediately, the 3-D system picked up the reduction in volume created by the increased print speed.

* ROI: This system also picked up the reduced volume immediately.

* MMS: Since print speed has a global effect, mass measurement picked up the reduced volume.

[FIGURE 3 OMITTED]

Print gap

A print gap was created for the next five boards, shown in Figure 4, by placing a strip of tape beneath U4. This gap was designed to represent a component level defect. The following was observed:

* 100 percent 3-D: Immediately, the 3-D system picked up an increase in volume on component U4 created by the print gap.

* ROI: This test showed the limitations of a ROI-type system as it passed all boards.

* MMS: This type of component-level defect may also indicate the limitations of a MMS. However, the MMS did show some indication of error on the last board.

[FIGURE 4 OMITTED]

Underside tooling replaced

For the last set of boards, as shown in Figure 5, the full underside tooling was replaced with a single pillar per board. This setup represented improper tooling support. The following was observed:

* 100 percent 3-D: The 3-D system picked up an increase in volume due to the poor support.

* ROI: This system also picked up an increase in volume.

* MMS: MMS gradually picked up the increase in volume and failed the print by the third board.

[FIGURE 5 OMITTED]

Mass Measurement Field Trials

Test Procedure and Equipment

For the mass measurement field trials at the three EMS providers--BEI, Celestica and Pemstar--the test procedure and equipment basics were the same for all facilities. The equipment consisted of:

* Sartorious LP620S desktop Balance with 620 g capacity at 0.001 g readability and 0.001 g repeatability--scale set up for normal vibration and half-digit stability

* laptop computer running Microsoft Excel 

* QC-Calc statistical software.

The procedure consisted of collecting data by weighing the board both prior to printing and after printing. Due to its sensitivity to air currents, the scale was placed in a static dissipative box. After the board was placed on the scale, a cover was placed over the box to eliminate air currents. After the post-printed board was weighed, it was placed onto the line conveyor and traveled to the next production stage.

Additional test and post-reflow inspection data were collected. All data were correlated to the boards by individual serial numbers.

Once the data were collected, they were imported into a SPC software package. These tests' goal was to determine if a MMS can be used as a viable process monitoring system. Using the data collected, a control chart was created. The limits were calculated automatically by selecting data collected during the most stable process period. The tolerance limits were calculated using [+ or -] 3 sigma with control limits set at 80 percent of the tolerance limits. Once the limits were calculated, the data were displayed with the limits in place. If the system was working, any out-of-tolerance plots should correlate with other measurement data such as end-of-line defect data.

BEI's Results

BEI's basic machine parameters for the product evaluated included flexible metal squeegees at a 60[degrees] angle and magnetic pillar support tooling.

Figure 6 is a control chart showing actual production data, which reveals an issue with the volume picked up by MMS and not by the 2-D system running at the same time. End-of-line data showed bridging on fine-pitch components. The operator adjusted squeegee pressure to compensate for the amount of paste being deposited, but this adjustment did not help. The tooling support was found to be insufficient to maintain the correct volume of paste. The last 10 plots show the system in control after the tooling was improved.

[FIGURE 6 OMITTED]

On a second product run the machine parameters were flexible metal squeegees at a 60 [degrees] angle and compliant support tooling.

For this set of boards, the paste roll was intentionally allowed to diminish. This product also had the onboard 2-D system programmed and running. As Board 17 was being inspected, the 2-D system flagged an insufficient solder error. As can be seen in Figure 7, a MMS would have also picked up the low paste deposit. The board was reprinted and passed both 2-D and MMS as indicated by Plot 18.

[FIGURE 7 OMITTED]

Celestica's Results

The machine parameters for the Celestica product consisted of flexible metal squeegees at a 60[degrees] angle and magnetic pillar support tooling.

No correlating end-of-line defect data were found that would indicate an issue with the data collected by the MMS; however, some interesting points were uncovered. During the production run, the stencil was removed and cleaned according to normal procedure. To make the cleaning time more efficient, a back-up stencil was used to continue production. Noticing the difference in the paste mass, the original stencil was replaced during the next stencil cleaning period and swapped once again during the last cycle. With the MMS the two stencils deposited different amounts of paste. As can be seen in Figure 8, each stencil resulted in a different plateau.

[FIGURE 8 OMITTED]

Pemstar's Results

The machine parameters for the Pemstar product consisted of metal squeegees at a 45 [degrees] angle and compliant support tooling. The Pemstar product started out with raw boards that have gold plating for the pad finish and switched midway to boards with organic solder protective (OSP) coating. The OSP boards have a different design stencil than the gold boards, so the data were separated.

Figure 9 has limits calculated on the most stable of the first 33 data points. The large dip represents Boards 16, 17 and 18. The paste roll diameter deteriorated, and paste was added prior to Board 19. If the limits were set up as in Figure 9, an out-of-control situation would have been indicated, prompting action.

[FIGURE 9 OMITTED]

Figure 10 uses Points 42 to 59 to calculate tolerances for the OSP-coated boards. Note the large upward spike is Board 38, which corresponds with excessive solder at end of line. If the tolerances were set up as in Figure 10, an out-of-control situation would have been indicated. If corrective action were in place, the defects could have been fixed before additional assembly.

[FIGURE 10 OMITTED]

Conclusions

Data from in-house experimentation and from the SPC program from the simple tests performed at the three sites show that mass measurement is a viable control monitoring system. With MMS better control can be realized in:

* Setup Verification: As seen by the first data sample, if a MMS were in place, production would not have begun because the first board would have failed, prompting the operator to verify setup more closely.

* Paste Roll Monitoring: As the paste roll diminishes, the volume of paste deposited is affected. Eventually, the paste roll becomes so low that more paste sticks to the squeegee, causing insufficient deposits on the board.

* Most Out-of-Control Situations Related to Deposition Volume: Unintentional debris on boards and other miscellaneous hazards will cause volume displacements due to blocking apertures and/or a poor gasket. These hazards have a direct effect on the deposition volume.

If set up as a process monitoring tool, the MMS would have triggered an alarm condition in correlation with end-of-line defects and all other paste inspection systems in place. From the results of this exercise, benefits of implementing an automatic system may include: improved cycle time (potentially zero effect); low cost; simple operator involvement; faster changeover; flexibility for different applications and virtually no programming time.

Acknowledgments: Special thanks to the companies involved in the data collection, especially to the engineers directly involved: Steve Beck, BEI; Scot Hanson, Celestica and Scott Severson, Pemstar. Jeff Schake and Rick), Bennett of DEK were instrumental in running the experiments and making sense of all of the data.

Richard Lieske is a senior applied product development engineer with DEK, Flemington, NJ; [phone]; e-mail: [e-mail].