A DoE clarifies the best types, best available flux within each type, and process limits for each.

This article discusses an approach to selecting liquid wave solder fluxes. It includes the design of an internal company test vehicle with a range of component types and design characteristics, preliminary flux performance testing, and results verification on the top-performing fluxes.

Included is discussion of the gaps between the capability of existing flux chemistries and requirements of not only Pb-free assembly, but from industry standards and customer expectations with respect to flux residue and solder quality on more complex assembly designs.

Experimentation. The board used for flux testing was the internal company wave test vehicle (TV) (Figures 1 and 2). Table 1 lists general properties of the tested board. The board was available in thicknesses of 0.093", 0.125" and 0.180". Surface finishes could be selected based on the needs of the experiment. The most commonly used surface finishes have been organic solderability preservative (OSP) and immersion silver (Imm-Ag). The TV was for mechanical testing only and had no electrical functionality.


This board and related components (Table 2) were designed to demonstrate the limits of the wave soldering process.


PTH header connectors and bottom-side SOICs were positioned in optimum and non-optimum orientations relative to the direction of travel to encourage solder bridging.

For the given header with a 0.025" square pin, the barrels had finished diameters of 0.035", 0.039", 0.043" and 0.047", and were grouped by connection to two, four, six and eight copper layers. The lead-to-hole ratio and copper connections were designed to increase issues such as vertical barrel fill. DIPs and axial resistor barrels were also connected to between one and six copper planes to reduce vertical barrel fill.

All bottom-side components were attached with surface mount adhesive only, prior to wave solder.

There are several different classifications of liquid fluxes, according to J-STD-0041: combinations of 1) rosin or organic (RO or OR) flux; 2) high, moderate or low flux activity levels (H, M, L); 3) halide-bearing or halide-free (1 or 0) flux. Less formal terminology is also used to describe fluxes used for wave soldering. Liquid fluxes are described by the solvent or carrier used (VOC-based, low VOC, or VOC-free), by post-wave solder condition (water-soluble, low residue, no-clean), or by activator (rosin or organic acid).

Fluxes selected for the screening portion of this study came from most categories, and were a result of a combination of market research and efforts to meet industry standard and customer requirements. Table 3 lists fluxes included in this evaluation by category, carrier, percent solids, and J-STC-004 designation. A letter was assigned to each flux from the wave flux supplier for anonymity. The following abbreviations are used for each category of flux:

1. Low residue, organic acid, no-clean flux (OANC) = VOC-free, water-based no-clean flux.
   
   
2. Low residue, rosin, no-clean flux (LRNC) = VOC-based no-clean flux.
   
   
3. Organic acid, water-soluble flux (OAWS) = water-soluble flux (some VOC, some VOC-free depending on flux used).
   
   
4. High residue, rosin, no-clean flux (HRNC) = VOC no-clean flux.
   
   
5. Low VOC rosin emulsion, no-clean flux (LVRE) = low VOC emulsion no-clean flux.

As can be seen from Table 3, OAWS and HRNC fluxes have typically higher solids content and more flux activity compared with the other flux types indicated.


Two phases of experimentation led to the selection of the top performing fluxes. Phase 1 was a screening phase and Phase 2 was for verification. Phase 1 involved carrying out a designed experiment (DoE) on 12 fluxes (A through C, E through Q from Table 3) to determine which fluxes in each category performed best using the test vehicle described. Phase 2 involved repeating the test runs of Phase 1 using the best fluxes from the screening phases to verify the results.

Following screening, various properties were determined, where applicable, before final process verification and addition to the internal company approved material list (AML). These criteria included:

• Telecordia (Bellcore) SIR (surface insulation resistance) and EM (electromigration) compliance.
• IPC SIR compliance.
• IPC SIR compatibility testing with all other soldering materials used on the AML: SMT solder paste, cored rework solder wire and rework flux.

Certain characteristics on the solder assemblies or measured responses to the various process conditions were considered most critical to the quality of the product (CTQs). These were chosen according to each phase of the evaluation. Table 4 delineates the CTQs and their assigned impact levels.


The Phase 1 full factorial designed experiment employed three factors, two levels each, two repetitions each (Table 5), resulting in 16 TV assemblies per flux.


Each factor’s settings were intended to encompass the range of parameters expected for most products considered to be challenging with respect to solder quality at wave solder. It was expected that the best flux would exhibit good performance across a broad range of the process window. For example, preheat was selected to determine the activation and possible sublimation levels of the flux.

Phase 2 was conducted for verification. The top performing fluxes from each key category were evaluated again using the same DoE matrix. Two TV assemblies were processed for each flux. The verification run was repeated at two different manufacturing sites.

The Phase 2 full factorial designed experiment used the same experimental matrix from Phase 1, except with the chip wave set to “ON” as a constant, resulting in eight assemblies per flux, per manufacturing site.

An additional low-solids rosin flux (flux D) was added to the Phase 2 verification stage of the evaluation based on the promising performance of the flux in production conditions relative to the fluxes included in the initial screening phase.

The procedural steps carried out included pre-conditioning the test vehicles through the bottom-side surface mount component attachment (adhesive dispense, component placement, cure) and a forced pass through reflow to simulate a double-sided process. The glue cure and reflow profiles are shown in Appendix A (400KB PDF). After SMT placement, glue cure and reflow, assemblies were stored in nitrogen until processing at the wave solder machine.

A production wave solder machine platform was used for all experiments. The full configuration is described in Table 6. Wave solder profiles are shown in Appendix A. These corresponded to the parameter matrix described in Table 5.


Flux spray distribution and penetration were checked using paper compatible to the specific flux types. The paper was sandwiched between a bare TV board and a backing plate (Figure 3). Paper sensitive to low pH was used with the VOC-free fluxes. Heat-sensitive (fax) paper was used for VOC-based fluxes. Figure 4 shows a sample spray fluxer result.


Critical characteristics included:

• Full penetration through 0.020" vias.
  
  
• Even spray pattern over the surface of the board, without stripes of heavy spray mixed with stripes of light spray.
  
  
• Minimum flooding or excessive flux volume on the trailing edge of the test fixture, which appeared as large, saturated areas on the paper. This was confirmed by observing the bottom side of the test fixture to see if the flux sprayed evenly over the surface, or dripped off the trailing edge of the fixture.

As listed in Table 4, various responses were measured for correlation to the wave solder process parameters chosen for the experimental matrix.

Vertical barrel fill. In Phase 1 of the flux evaluation, vertical barrel fill was measured using production x-ray equipment. The x-ray system captured an image parallel to the top surface of the board at 85% of the board thickness. The grayscale of the image captured was automatically measured and then the slice was evaluated visually as to whether it was to be considered a “PASS” or “FAIL.” A threshold grayscale value was determined, and all solder joints were then tested based on the pass/fail threshold value. If the grayscale of the “slice” was greater than the pass/fail threshold value, the barrel was considered greater than 85% full and acceptable as a “pass.” Passes were recorded and subtracted from the total number of leads tested to determine failure rate. Barrel fill was given a high impact level of 5, as it is difficult to detect and can be difficult to correct if the original cause was thermally related.

Solder bridging. Bottom-side solder bridges were simply counted and each flux rated according to the total number of bridges. For consistency, SMT bridges were counted according to gaps that were shorted; PTH bridges were counted according to the number of leads connected to the bridge. One bridge between two leads was counted as two, but one bridge encompassing four pins was not counted as the permutations of P1-P2, P2-P3, etc. Bridging was given an impact level of 4; bridging violated IPC-A-610D2, but was detectable both visually and electrically, and rework relatively straightforward, with minimal impact to the assembly.

Bottom SMT opens (skips). Bottom-side SMT components (SOICs and SOTs) were examined visually for solder opens. This was facilitated by the use of OSP over bare copper board. If solder did not wet to the metallization surface, the orange-colored copper surface showed clearly in contrast to the silver-colored wetted surfaces. Skips were given an impact level of 2; opens violated IPC-A-610D2, but the likelihood of detection, either visually or electrically, was high, and most products would be designed so that SMT solder joints were not formed during wave soldering; therefore, this was deemed a relatively minor concern. Further, solder wetting relative to flux was indicated by how well the solder filled PTH barrels, which was already given a high impact level.

Solder balls. Solder balls were inspected on a basis of presence vs. absence across the original 16 test vehicle assemblies. Solder balls were given an impact level of 2; sufficient size and quantity of solder balls violated IPC- A-610D2; however, they were not relevant to water-soluble fluxes due to complete removal during the wash process and were typically encapsulated when using rosin-based fluxes.

Residue. Flux residue was assessed qualitatively during Phase 1. Each assembly was examined and ranked (Table 7).


Results and Discussion

Table 8 shows how all fluxes performed in Phase 1 screening. The weighted score is based on defects, so the lower the value, the better the flux performance. In general, the water-soluble (OAWS) and high-solid rosin (HRNC) no-clean fluxes performed best, along with the low-VOC emulsion flux. Low-solid rosin no-clean fluxes (LRNC) showed the next best results. Low-solid organic acid VOC-free (OANC) fluxes, along with the high-VOC rosin emulsion no-clean fluxes, ranked lowest. Table A1 in the Appendix shows detail with respect to specific CTQs/defects.


While the high-residue rosin VOC fluxes performed well, customer acceptance would be low because of cosmetic concerns and in-circuit test pin probability issues. Having identified the best of these fluxes, a possible solution for solder quality on the most difficult boards – particularly vertical fill – is identified if the customer is not concerned about the appearance of residue and has measures in place to remove flux residue for pin probeability.

One of the rosin emulsion fluxes (Flux P) showed promise; however, it did not pass Bellcore SIR testing and was removed from consideration. The other rosin emulsion flux, Flux Q, ranked among the lowest performers, and contained 70% VOC, in spite of being listed as a low-VOC flux. There was also evidence that some spray fluxers clogged as a result of the rosin solids of the LVRE type fluxes, which are suspended in the low-VOC solution and are sheared out of solution by the mechanical pump action of the spray fluxer.

The top fluxes in each of the remaining categories were used in the Phase 2 verification runs and are highlighted in Table 8 (Fluxes A, B, D, H).

As already mentioned, Phase 2 verification was performed according to the matrix shown in Table 5. Table 9 shows Phase 2 verification phase defect results. Each CTQ was again normalized and weighted, and the fluxes were ranked accordingly (Table 10).


As indicated, the lower the score, the better the flux performed. Given that the fluxes selected for verification in Phase 2 were among the best on the market, not only was the optimum flux determined for each category, but also the defect data were analyzed to determine process parameters having the greatest influence on the performance of the fluxes.

The following sections summarize many of the process conditions considered critical to product quality and the significant factors contributing to each.

Vertical Barrel Fill

Preheat effect. In general, the hotter the board, the better solder flows into the plated through-holes. The data show that x-ray defects decreased with higher preheat temperature when using water-soluble VOC (OAWS) or rosin low-VOC (LRNC) fluxes, while they increased with higher temperature when using the organic acid no-clean VOC-free fluxes (OANC). This indicated that OAWS and LRNC fluxes are more heat-tolerant than OANC fluxes.

Conveyor speed effect. The slower the conveyor speed, the greater the solder contact time so that the longer the solder has to wet to the barrel walls. This held true for the VOC OAWS and low-solid VOC LRNC fluxes. However, slower conveyor speed increased energy input to the board during preheating. As indicated previously, OANC VOC-free no-clean fluxes were less heat-tolerant. The data showed the lower conveyor speed was better for OANC fluxes, provided the preheat temperature was lower. Vertical barrel fill decreased significantly for OANC fluxes with higher preheat temperature on the 0.125"-thick OSP board for lead-free SnAgCu wave soldering.

Flux selection for improved Pb-free barrel fill. High activity fluxes (OAWS and LRNC) offered ~10% better barrel fill than VOC-free no-clean fluxes on the 0.125"-thick OSP boards. LRNC fill was slightly better than OAWS, but both were significantly better than OANC.

Bridges

Preheat effect. For the low VOC LRNC flux, lower board temperature resulted in lower solder bridging for Flux D. Preheat was a much more significant factor with respect to VOC-free OANC than OAWS VOC fluxes.

By the same reasoning applied with respect to vertical barrel fill, the OANC fluxes appeared to lose activity if overheated and could not perform their function of facilitating solder flow and smooth solder peel-off from the bottom surface of the PWB. This appeared to be a characteristic of organic acids, where the behavior was similar, but higher solid concentration in the OAWS fluxes survived longer than that of the OANC fluxes, leading to reduced bridging.

Conveyor speed effect. For the low VOC LRNC flux, higher conveyor speed resulted in less solder bridging for Flux D. High conveyor speed was better for the OANC VOC-free fluxes by similar argument with respect to vertical fill – more heat energy into the flux resulted in reduced effectiveness. Lower conveyor speed was better for the OAWS flux. Smooth peel-off from the board was more critical than thermal stability.

Flux selection for reduced solder bridges. On average, OAWS VOC fluxes exhibited less bridging than LRNC low VOC fluxes.

OANC VOC-free fluxes had significantly more solder bridging than the other categories under the high energy process conditions: low conveyor speed, low and high preheat temperature. It was only somewhat higher at low conveyor speed and low preheat temperature. LRNC low residue Flux E had an uncharacteristically high rate of bridging with the combination of low conveyor speed and high preheat temperature.

Solder Skips on Glued Bottom-Side Wave-Soldered SMDs

Preheat effect. Higher preheat increased solder skips when using a VOC-free OANC flux, while lowering the incidence of skips with low VOC LRNC and VOC OAWS fluxes. This again was likely due to the sublimation of flux solids during preheat. In general, preheat was a less significant factor.

Conveyor speed effect. As expected, solder skips were more numerous at higher conveyor speeds, but only with the no-clean fluxes, and not to a significant extent.

Flux selection for reduced solder skips. VOC-free OANC fluxes had the best results for solder skips. LRNC fluxes had very few skips. Higher preheat was worse for VOC-free OANC than VOC OAWS flux. By far, the OAWS fluxes generated more solder skips.

Solder Balls

Preheat and conveyor speed were not a factor in the presence or absence of solder balls with the organic acid fluxes. When using low-residue LRNC flux, there appeared to be fewer solder balls at higher conveyor speed.

Flux selection was the only critical factor. OAWS flux by its nature had zero solder balls due to washing of these boards. All boards with VOC-free OANC flux had solder balls. There were also significant differences between fluxes within each category.

Conclusions

As indicated, there are tradeoffs in characteristics within fluxes of the same type and among different flux type categories.

For Pb-free soldering 0.125"-thick OSP boards, the following conclusions can be drawn:

1. For increased Pb-free barrel fill, the water-soluble fluxes are the best to use, followed by low-VOC no-clean fluxes, followed by VOC-free no-clean fluxes.
   
   
2. For reduced Pb-free solder bridging for bottom-side wave-soldered SMDs, water-soluble fluxes are best, followed by low-VOC no-clean fluxes and VOC-free no-clean fluxes.
   
   
3. For reduced Pb-free solder balls on the bottom side of the wave-soldered board, water-soluble fluxes are best because solder balls are removed by washing, followed by low-VOC no-clean fluxes and VOC-free no-clean fluxes.
   
   
4. For reduced Pb-free solder skips on the board bottom-side, VOC-free no-clean fluxes are best, followed by low-VOC no-clean fluxes and water-soluble fluxes.

The data presented indicate much more work to do to improve Pb-free wave soldering results for organic acid VOC-free and low-residue rosin no-clean fluxes.

Future work should concentrate on the development of VOC-free and low-residue rosin no-clean fluxes for lead-free soldering, especially for boards with thicknesses of greater than 0.100". Also, pin probability evaluations need to be conducted more if low-residue, rosin no-clean fluxes are used, as the solids content may affect pin probe results.
 
References

1. J-STD-004A, Requirements for Soldering Fluxes, January 2004.
2. IPC-A-610D, Acceptability of Electronic Assemblies, February 2005.

Acknowledgments

The authors would like to acknowledge the flux suppliers for proving their wave soldering fluxes for this evaluation. They would also like to acknowledge their internal company colleagues who helped to support much of this work.

Ed.: This article was originally published at SMTA International in October 2007 and is used with permission.

Douglas Watson is process development engineer, Jasbir Bath is lead engineer and Pan Wei Chih is process engineer at Flextronics International (flextronics.com); douglas.watson@flextronics.com.