Squeegee Blade Wear Evaluation

Document Sample
Squeegee Blade Wear Evaluation Powered By Docstoc
					DK-99-83
LIESKE-BECK - 1 of 7
EDIT 1 - 04/02/11


                       Reducing Defects and Costs With Adhesive Printing
                                by Steve Beck and Richard Lieske


       Adhesive dispensing for SMT components is a process that many manufacturing engineers
wish would disappear, but somehow it never does. Problems with adhesive dispensing include com-
ponents falling off in the wave, adhesive in solder joints and, sometimes, adhesive in places it
shouldn’t be. Just about the time an engineer thinks the process is under control, something goes as-
tray. Very often, manufacturability design review teams tell designers to “get as many of those SMT
parts to the top as possible,” only to hear the famous comeback, “the circuit does not allow it.”
       In an attempt to resolve this dilemma, a study of the cost and quality of adhesive printing vs.
adhesive dispensing was conducted on a high-count bottom side SMT printed circuit board assembly
at the Winona, Minnesota, facility of EMS provider Benchmark Electronics. For the study, a board
with “high count adhesive sites” was defined as an assembly that required more than two minutes in
a typical adhesive dispensing machine. This length of time calculates out to approximately 1,000
components on the bottom side, given the dispensing equipment used at Benchmark Electronics.
       The particular assembly used for the study consisted of an 18x20-inch PCBA with more than
2,500 bottom side SMT placements, requiring only minor allowances for wave soldering. The board
was populated with 0603, 0805, SOT23, Tantalum Capacitors and 0.020” pitch QFP bottom-side
SMT components. It also featured a high mix of pin grid arrays, large connectors and various other
leaded parts.

Process Development

       Because this assembly was of the mixed technology type (i.e. surface mount and leaded com-
ponents on the same PCB), wave soldering was necessary. Also, since there were some 0.020-inch
pitch QFP devices on the bottom side, printing of solder paste was required. A complete assembly
solution was needed to address all these component types without adding hand solder as an opera-
tion. The first attempt at processing this PCB followed the steps shown in Figure 11.

                                              [Figure 1]




                                                                                                   dkl-bart
DK-99-83
LIESKE-BECK - 2 of 7
EDIT 1 - 04/02/11


       After running a number of jobs using this process flow, it was apparent that improvements
needed to be made. The largest contributors to the defect list included:

                                       Total assemblies run = 200
                                      Total parts placed = 522,600

                                       Open solder joints = 71%*
                                       Insufficient solder = 23%*
                                       Missing components = 6%*

                            *Defect percentages are based only on the defects listed

       Further analysis showed that the insufficient and open solder joints resulted from a layout that
had not been properly optimized for the wave soldering process, as most of the defects were confined
to particular areas of the assembly. Rotation of the PCB through the wave process was considered as
a possible solution, but further investigation showed that no other direction would put all the compo-
nents into a preferred wave orientation. A new process was needed to reduce these defects.
       Since solder paste was already being printed for fine pitch components, it was decided to
print solder paste on all SMT pads and place epoxy dots between the paste as required. This ap-
proach would create solder joints before the assembly entered the wave, greatly reducing the possi-
bility of insufficient and open solder joints. A selective wave soldering process would be necessary
(it was also required in the first process), as the fine pitch devices could not be exposed during wave
soldering. Since the fine pitch area of the board was the only place where there was enough clearance
to design the selective wave pallet properly, it was decided to continue wave soldering the entire as-
sembly and mask off the fine pitch area with the pallet. This dictated the use of epoxy and solder
paste for all discrete components.
       An added bonus gained by wave soldering all discrete components was that no new tooling
was required; the wave pallet that had already been designed and was being used in the current
process could continue to be used. This saved Benchmark the expense of re-tooling for a non-proven
process during the development stage.
       After making the necessary changes, the revised process followed the flow chart shown in
Figure 2.




                                                                                                dkl-bart
DK-99-83
LIESKE-BECK - 3 of 7
EDIT 1 - 04/02/11


                                                   [Figure 2]

                                       Total assemblies run = 260
                                    Total SMT parts placed = 679,380

                               Open solder joints = 169x defect reduction*
                               Insufficient solder = 53x defect reduction*
                               Missing components = 1.4x defect increase*

         *Defect reduction/increase is based on comparison between the original and above production runs

       As can be seen in Figure 2, the defect levels for insufficient and open solder joints were sig-
nificantly reduced, as was expected. There was an unexpected increase in defects, however, in miss-
ing components. This increase was attributed to the solder paste. When a component is placed direct-
ly on the PCB without any solder paste, the component is approximately flush with the PCB. When
solder paste is added first, the component sits 2-3 mils higher, decreasing the contact area between
the adhesive dot and the component, as shown in Figure 3.
                                                   [Figure 3]

       When the dot size drifts to the smaller end of the process window, there is no longer suffi-
cient contact area with the component, increasing the likelihood that it will fall off during the wave
process. With the addition of the solder paste, the acceptable process window for the dot size be-
came much smaller, and was likely to be smaller than the machine process window. With the limited
amount of time available for the study, a capability study could not be run on the glue dispenser to
verify this assumption. Instead, observations were made of the remaining glue dots at several loca-
tions on the production boards where components were missing. In these cases, a dot was present,
but it was slightly undersized and had been compressed very little.
       With the high missing component count being a cause for concern, other process and equip-
ment changes were considered. After many hours of brainstorming, a new process was developed:
solder paste was printed, as in the current process, and then epoxy dots were printed between the
solder paste, using the same DEK 265 screen printer for both solder paste and epoxy dot deposition.
It was anticipated that this process would provide several significant quality benefits:




                                                                                                            dkl-bart
DK-99-83
LIESKE-BECK - 4 of 7
EDIT 1 - 04/02/11


        1. Much tighter process control of dot size. Printing of epoxy can be thought of as a pseudo
positive displacement system, in which printing continually fills an aperture with a fixed amount and
releases a fixed amount.
        2. Ability to engineer the dot shape to best fit the application. A dispenser allows round dots
only, when sometimes an oval or other shape would be preferred. An oval engineered glue dot mini-
mizes the chance of missing components by allowing additional glue-to-component surface contact,
as shown in Figure 4.
                                                        [Figure 4]
        3. Dot placement accuracy is thought to be better with a printer than with a dispenser, since
the dispensing nozzle (the stencil) is aligned directly to the board. A dispenser aligns the glue head,
not the dispensing nozzle, to the board.
        4. The dispensing system Benchmark used in this testing was shown to have elevated epoxy
chamber temperatures. During heavy production runs, epoxy begins to cure in the tube during dis-
pensing. This temperature change causes a change in the viscosity and, as a result, the dispensing
characteristics of the adhesive cause the dot size to drift.
        The process flow chart for this proposed process change can be seen in Figure 5.

                                                    [Figure 5]

                                         Total assemblies run = 70
                                     Total SMT parts placed = 182,910

                                Open solder joints = 165x defect reduction*
                                Insufficient solder = 49x defect reduction*
                                Missing components = 9x defect reduction*

          *Defect reduction/increase is based on comparison between the original and above production runs

        Unfortunately, a quantity of boards comparable to those used in the previous process trials
could not be run, due to schedule shifts. However, as can be seen in Figure 5, the results were very
encouraging. The defect levels for insufficient and open solder joints were comparable to the pre-
vious process, while the missing component defect levels were significantly reduced.




                                                                                                             dkl-bart
DK-99-83
LIESKE-BECK - 5 of 7
EDIT 1 - 04/02/11


Cost Savings Realized

         During process development, a cost analysis was conducted to see how the process of dis-
pensing epoxy and printing paste compared to the process of printing epoxy and printing paste.

                                             Dispensing                          Printing
Throughput cost1                             $10,000/yr                          -
Tip replacement cost2                        $1,000/yr                           -
Stencil replacement cost3                    -                                   $400/yr
Estimated adhesive cost4                     $1,469/yr                           $2,613/yr
PCB repair cost5                             $15,000/yr                          $5,000/yr
TOTAL                                        $27,469/yr                          $8,013/yr
1
  Cost is based on shifting the gating process to adhesive dispensing from component placement
2
  Tips have been found to have a limited life span due to the stand-off wearing down.
3
  Stencil replacement is for the estimated single ECO on this assembly per year.
4
  Cost of adhesive includes projected waste. If a sealed squeegee head is used, waste from the printing process would be
reduced significantly and the costs would likely be comparable.
5
  Repair cost is based on the time necessary to repair process related defects.

         On an annual basis, the total cost savings realized, for a single line on a single shift, would
amount to $19,456. In a large facility, these savings would be multiplied by the number of lines and
shifts being run. In addition to cost issues, a number of other considerations are noteworthy when
comparing dispensing vs. printing adhesive.
         One of the benefits of adhesive printing is that spare parts inventory can be reduced, as one
machine type (adhesive dispenser) can potentially be eliminated. The addition of the printer would
not be expected to add any additional spare parts, as it is assumed the spare parts inventory already in
place for the existing printers would be sufficient. The same elimination of a machine type would
reduce training time for operators, engineering and maintenance personnel. Finally, dispensers are
limited to the single task of dispensing adhesive. Printers have a much wider range of process possi-
bilities, so the machine could be utilized in other fashions in the future if needed.
         The benefits of dispensing include the ability to change the glue dot locations with simple
programming changes rather than with a new stencil. While this can be a concern on boards with
continual ECOs, it should be noted that an internal cost of generating these programs does exist and
should be considered. In addition, dispensing allows on-the-fly dot size changes to be made. Howev-




                                                                                                                 dkl-bart
DK-99-83
LIESKE-BECK - 6 of 7
EDIT 1 - 04/02/11


er, having the ability to change processing parameters easily can have drawbacks as well, specifically
if unnecessary changes are made frequently.

When to Consider the Process

       As with any process, there are products that will benefit greatly from it, and others that will
not see a change. When analyzing the potential use of this process on a particular assembly, the is-
sues to be considered include:
          If the board’s SMT layout is marginal for wave soldering. Creating solder joints prior to
           wave soldering by the solder paste reflow method will almost eliminate open solder
           joints.
          If there is a high number of 0603 or smaller bottom side SMT components. When going
           below an 0805 size component, the dot size repeatability appears to deteriorate rapidly.
           From experience, this is the most common part to be lost in the wave.
          If there is a high SMT count on the bottom side of the board. Throughput can be signifi-
           cantly increased by printing adhesive when counts exceed 1,000 components.
          If revisions to the PCB are low. When revisions are done on the PCB, a new stencil will
           likely be required.

Conclusion

       The development of an epoxy printing process has been shown to have great potential for re-
ducing defects for some products. Further data needs to be collected to verify the results, but all indi-
cations show that significant process improvements have been made. In addition, there is a potential
for cost savings to be realized. As with any product/process relationship, a complete engineering
analysis needs to be performed and all considerations need to be factored into determining the best
process for a particular product.


                                                      ###




                                                                                                 dkl-bart
DK-99-83
LIESKE-BECK - 7 of 7
EDIT 1 - 04/02/11


       For more information, contact Richard Lieske at DEK USA, 8 Bartles Corner Road, Fleming-
ton NJ 08822; Tel: 908-782-4140; Fax: 908-782-4774; website: www.dek.com; e-mail:
rlieske@dek.com, or Steve Beck at Benchmark Electronics, 4065 Theurer Blvd., Winona MN
55987; Tel: 507-453-4749; website: www.bench.com; e-mail Steve.Beck@bench.com.




References
1
  Figure 1 and all subsequent figures and tables are taken from: Beck, S. and Lieske, R., “A Cost
and Quality Analysis for High Count Adhesive Sites,” presented at Nepcon West 1999.




                                                                                             dkl-bart

				
DOCUMENT INFO