The solder masks were processed according to standard conditions specified in their respective processing guides. All masks were given an UV bump of 2 J/sq cm prior to the final thermal cure. An exception was made for mask A, which required a UV bump of 4 J/sq cm after thermal cure.

Ultraviolet bumping minimises the ionics, as measured after hot air solder leveling post clean. Mask A required a longer exposure to ultraviolet light because of its dependence on radiation for complete crosslinking prior to hot air solder leveling.

In the tack drying process, all of the solvent must be removed to leave a tackfree film on the board for contact imaging. Processing the boards at the appropriate turnover rate for the oven is critical when operating at temperatures sufficient to remove the solvent without curing the solder mask.

Alpha Metals found that currently, available ovens do not provide consistent tack dry curing of the same liquid photoimageable solder mask from shop to shop.

Appropriate turnover rate is imperative if the solvents used are the environmentally safe types, which tend to boil at higher temperatures. Excessive solvent entrapped in the mask after tack drying will cause problems farther down the line. Not only will the phototool stick to the solder mask, the mask will blister after hot air solder leveling, wrinkle and lose adhesion to the laminate.

Some ovens have a lower turnover rate than others. Turnover rate can be calculated by dividing airflow through the stack in cubic feet per minute by the capacity of the oven chamber. An independent study of the airflow of ovens used in various fabricators' facilities indicated turnover rate should be greater than three times per minute.

All boards were processed through hot air solder leveling using a standard hot air solder leveling flux. 'They were then cleaned using a saponifier with a deionized water rinse at 65°C (150°F) and subjected to a final water rinse.

No-clean flux was applied using spray, bottles containing the individual fluxes.

A sufficient quantity of flux was sprayed on the boards to protect their surfaces but not inundate them to an extent that would defeat the purpose of the test.

The boards were then populated with surface mount components on the bottom side and through-hole components on the top side, and wave soldered. wave solder parameters were as follows: solder temperature. 260°C (500°F); conveyor speed 4 fpm; top preheat temperature, 93°C - 99°C (200°F - 210°F) measurement at surface with thermal labels.

Figure 1.	Solder masks used in both studies are combined in this figure. B and C were not used in the Alpha Metals study; F through H were not used in the Allen Bradley study.

Gloss numbers of the six solder masks are shown in Figure 1
[solder masks B and C were not evaluated in this stud).

A high degree of gloss was measured on all the masks except mask A.

Figure 2.	Solder masks used in both studies are combined in this figure. B and C were not used in the Alpha Metals study; F through H were not used in the Allen Bradley study.

Test results, shown in Figure 2, indicate that solder mask A consistently yields less than five solder balls per square inch regardless of the no-clean flux used.

1. One hypothesis for this phenomenon is that, although the hills and valley's of the solder mask's matte surface greatly increase the total surface area. the area solder balls can attach to is reduced.
   Solder balls can only attach themselves to the tops of the hills.
2. Another theory is that the matte surface acts as a reservoir where the flux is entrapped. The trapped flux provides a heat sink that prevents the mask's surface temperature from rising above the glass transition temperature, the point at which the mask softens and provides a tacky surface solder balls can adhere to.
   
   
3. Still another possible explanation is that the flux entrapped in tile matte surface vaporises when passing over the wave. As the flux vaporises, it blows solder balls breaking from the wave's trailing edge off the board's surface. Alpha metals is conducting further studies to determine the actual mechanism that reduces the number of solder balls left on a board.

Allen-Bradlev's study was performed on site using a production assembly line with specially designed test boards.
Component spacing, component orientation and pad sizes on the boards were designed specifically to provoke solder shorts.

• six axial lead components mounted next to vias,
• twelve 1206 components mounted next to clinched leads,
• twenty 1206 components clustered transverse to board travel,
• twenty four 1206 components clustered parallel to board travel,
• twenty 1210 components clustered transverse to board travel,
  
  
• twenty-two 1210 components clustered parallel to board travel,
• ten 29-pin transverse-mounted SOIC components and
• eight transverse-mounted square-pin leaded connectors.

Five different liquid photoimageable solder masks were selected to represent a wide spectrum of surface textures as measured by percent gloss (mask F through H were not used in this study)

The test boards were prepared at the solder mask manufacturer's facility and submitted for assembly.
Some boards with solder mask A were given a UV bump of 4'J/sq cm after thermal cure; others were not.

Surface mount components placed on the bottom sides of the boards, and through-hole components were inserted on the top sides.
Following surface mount component placement, the boards passed through an adhesive cure infrared oven. After through-hole component insertion, the boards were sent through 1,1,1 trichloroethane solvent for two minutes at 35°C (95°F), spray only.

Ten boards of each solder mask were wave soldered with nitrogen cover gas using the following parameters:
preheat temperature, 127°C (260°F) ± 10 degrees: solder

contact, two seconds; oxygen, less than 10 percent; flux, 2 percent solids no clean.
The surface texture of each solder mask was then evaluated by taking gloss measurements (Figure 1).

Meseran studies were performed on mask A using under-cure, standard-cure and over-cure conditions to see if there were changes in the number of solder balls (Figure 3).

Figure 3.	No decrease in the number of solder balls was shown with solder mask A once the recommended standard optimum cure was achieved

Figure 4.	Fewer solder balls and solder shorts are produced when liquid photoimageable solder masks with matte surface are used.

Test results demonstrated that the surface texture of a liquid photoimageable solder mask has a significant effect on the incidence of solder balls and solder shorts (Figure 4).
Masks with rougher surfaces produced fewer solder shorts and solder balls than liquid photo-imageable solder masks with smoother surfaces.

Source unknown.