A Unique Heating/Reflow Technique to Minimize Solder Paste-Induced Voiding Under LEDs
BY DERRICK HERRON, INDIUM CORPORATION AND JOHN MATHURIN, CHARLIE WILKINSON, ANSEN CORPORATION
When soldering LEDs to a circuit board, voiding can be a critical issue. Just as with other bottom terminated components, the thermal pad is used to transfer heat away from the part, allowing it to function at a cooler temperature. Voiding within the solder joint will interfere in heat transfer, causing the component to run hotter and shortening the working life of the LED. As is known in the industry, solder paste-induced voiding on these thermal pads is a common problem.
A novel process has been developed where solder paste is printed, the paste is then dried, fresh paste is printed over the original deposits, and the assembly is then reflowed using standard reflow practices. Aspects of the printing and reflow process considered in this paper include drying temperature, stencil thickness, aperture size, pitch, slump of the paste, and amount of void reduction.
Traditional electric lighting has used the incandescent bulb in which electrical current is passed through a metal filament to heat it until it glows to produce light. This type of technology was first attempted in the early 1800s and continued until the end of that century when Thomas Edison and his partners achieved a commercially successful version. Although improvements were made on this technology, even today the bulbs only produce at small efficiency percentages with the majority of the power input being converted to heat instead of light. The lifespan of the bulbs is also a drawback to this technology due to the fragile nature of the filaments. For these reasons, recent trends and regulations have pushed for more efficient lighting technologies to come into the mainstream.
One popular solution is to use light-emitting diodes (LEDs) in place of traditional light bulbs. LEDs are less fragile than filament-containing light bulbs and have longer service lives. LEDs are also more efficient, using less energy to heat than a traditional bulb. However, it is also heat creation that is a major concern for LED lighting manufacturers—the heat produced by the die must be able to escape, keeping the die cool, or the lifespan of the LED will be shortened.
Per the Bergquist Company, it is very common today for three- or five-watt LEDs to be used in the lighting industry. The trend is progressing towards 10-watt LEDs for high-power lighting solutions. With these higher power components, more heat is generated. It is common for any LED over one watt of power to be surface mounted on a PCB to allow for better heat transfer than would be gained simply through the leads of a leaded device.1
With bottom terminated components, such as high-power LEDs, heat is carried away from the component through thermal pads on its underside. The solder joint that connects the pad to the board allows for heat transfer into the board and away from the component. Voids found within this solder joint interfere with the straight-line heat transfer into the board and away from the component, creating hot spots. In high-power LEDs, this can lead to a change in the LED output color and also shorten the component life.
A novel method of limiting voiding on these thermal pads has been developed. First, paste is printed on the pad in the normal way. The board is then sent through a low temperature oven to dry the paste. The same board then gets printed in the printer a second time, with the paste applied directly on top of the first print deposit.
One of the main concerns found with this experiment is that solder paste is printed and then dried out on the board, with a second printing done afterwards on top of the original print. If the same stencil is being used for both prints, then the paste must not slump at all during the drying process or the stencil will not be able to gasket well. Financially, it would be a best case scenario if the same stencil could be used for both the first and second prints, with no damage being done to the stencil in the process.
Originally published in the Proceedings of SMTA International, Rosemont, IL, September 25‐September 29, 2016.
In theory, two possibilities could happen:
1. The first print does not slump during
the drying process (Figures 1 and 2). The stencil gaskets to the board nicely during the second print and more paste is forced into the deposit (Figure 3), making a denser paste deposit prior to the actual reflow. This deposit will have more metal volume than the initial print. Due to the drying of the initial paste deposit, the flux volume will not be significantly increased. The volatile components of the flux should, in fact, be greatly reduced as those from the first print were driven out during drying, and fewer volatiles will have been added during the second print.
2. The first print slumps during the drying process (Figure 4). This causes solder particles to be outside of the printed area, preventing the stencil from creating a good gasket (Figure 5). The new paste is fully printed on top of the first paste deposit (Figure 6).
In theory, either one of these could be beneficial to reduce voiding in the final solder joint. In the first scenario, more metal volume is added to the joint before reflow without significantly adding more flux. This can be compared to the use of a flux-coated solder preform added to the joint.2
The second scenario, in which the gasket between the stencil and board is poor, provides an increase in solder volume to the joint. This provides higher standoff of the component and more room for outgassing of the solder paste.3
The following experiments were done to further study this method of drying paste and then printing a second layer in order to reduce voiding.
Part 1 – Examination of slump
Two solder pastes were used for this experiment. One was a low-voiding halogen-free paste (A) and the other was a low-voiding halogen-containing paste (B).
Before printing, it was noted that paste A had a higher viscosity than Paste B, and felt firmer when it was gently stirred and hand-transferred to the stencil.
A test board with OSP treated pads of size 0.010” x 0.050” and a stencil with 1:1 aperture to pad ratios in this area were used. The stencil was 0.004” thick.
Paste was printed and separate boards were then baked in a convection reflow oven with temperatures of 125°C, 150°C, and 180°C with a duration of four minutes each. It was theorized that a higher temperature bake would allow for more flux outgassing, which would be preferred. However, it was also theorized that a higher temperature bake would lead to more slump of the solder paste during the baking.
The lowest temperature bake condition was with an oven temperature of 125°C, due to limitations of the available oven. It can be seen that even at this lowest baking condition, there was still some spread of both pastes. It should be noted that both of these pastes pass J-STD-005 slump test requirements per IPC-TM-650. The slump seen was not enough that bridging would have been a concern. However, the paste spread was enough that the same stencil would not have formed a proper gasket if used to print a second time on one of the baked boards.
Measurements were taken on the paste width before and after baking. Due to data scattering, it was not clear if one baking condition worked better than another at preventing slump; all showed very similar results. At first glance, the 125°C baked condition in the photos looks worse than the others. Using the pad size as a reference point and normalizing the numbers, the slump was no worse with this condition than it was on the others.
Part 2 – Double printing experiment
Solder Pastes A and B were each printed on a QFN test board using available laboratory resources. Due to the small pitch I/O apertures on the QFN design, this would be a worst-case scenario with regard to slump and bridging.
Each of the pastes was printed onto a fresh test board and a photo was taken of the fresh paste. Each of these boards was then baked at the 125°C condition and the boards were analyzed. Boards were then printed a second time.
It can be seen in Figures 9 and 10 that the solder paste after the baking procedure has lost its shiny appearance. This is most likely due to volatile components of the flux being driven out during the baking process. The paste at this step has lost its tackiness and would not, in most cases, hold a component in place as a standard solder paste would in the SMT process.
The larger photos in Figures 9 and 10 show the second layer of paste printed on top of the first. It is clear that there are now two layers of paste, thus representing Scenario 2.
Part 3 – An examination of voiding under leds experiment
With the information garnered in the first two parts of the experiment, it was known that the second print would place a second layer of paste on top of the first. A customer-designed LED-containing board for an industrial lighting application was used to evaluate the void reduction properties of the print–bake–print method. Based on customer requirements for a halide-free paste, only Paste A was used in this part of the evaluation.
The LED components used were commercially available 0.120” x 0.120” LEDs. The LEDs were arranged symmetrically around a board designed for an industrial lighting application.
Taking into consideration that all of the bake temperatures in Part 1 led to some slump of the solder paste, this final test was limited to only the 125°C bake.
After initial printing using a 0.004” thick stencil, it was determined that the paste had an average thickness of 0.0041”. The paste deposits had a typical brick shape to them, somewhat tapered from the board level up to the top of the paste deposit. The paste had its usual wet appearance. This can be seen in Figure 12.
After drying at 125°C for ten minutes, the paste obtained a dried appearance. It was also observed that the paste had spread to fill more of the printed area, as evidenced by the green box in Figure 13, which was the same size as the green box in Figure 12. The average height remained 0.0041”.
After drying, the second printing was attempted at two separate print pressures to determine whether more pressure would force more paste into the solder joint and give different results than a second print at a lower pressure. A second print with a print pressure of 11 PSI gave the type of deposit seen in Figure 14 (left), while a print pressure of 20 PSI gave the deposit seen in Figure 14 (right).
Of note in Figure 14 is that some paste was scooped out of the upper right corner of the 11 PSI print during the print. The two separate layers of printed paste can clearly be seen.
In both cases of print pressure, the average thickness of the resulting solder joint was 0.0056”.
After placement of the LED components and the final reflow using typical SMT reflow conditions, X-ray images were taken of the solder joints. Figure 15 shows representative voiding images after the standard SMT process, while Figure 16 shows representative voiding images after the print–bake–print process.
An examination of the voiding images produced by the X-ray does appear to show that the print–bake–print method has reduced the amount of voiding within the solder joint. Taking the actual percentages of voiding and charting them gives the following numbers.
In Figure 17, it is seen that the results from Paste A using a traditional SMT process averaged less than 20% voiding and were not that unacceptable. However, after implementing the print–bake–print process before reflow, the voiding was reduced to an average of less than 10%.
It is theorized that the reason for the decrease in voiding comes from the fact that the paste printed in the first step had most of its volatile components removed during the 125°C bake. During the second print, less paste was placed on the board than during the first print; therefore, there were less volatiles present to escape during the final reflow than during the traditional SMT process. The smaller amount of flux that was added during the second print, along with the flux solids left behind after the bake, were enough to allow for good wetting and flow of the solder joint. Of note, the paste deposits depicted in Figures 12, 13, and 14 are all square in appearance, whereas the X-ray images in Figures 15 and 16 depict rectangles. This is due to the board pad being larger than the component pad and the solder flowing during reflow is mostly beneath the LED pad.
It was also noted that the variations in pressure on the second print in the print–bake–print method had no impact on the resultant voiding. These results were combined to simply have results using the print–bake–print method, independent of second print pressure.
This method is recommended for LEDs, especially more so than other bottom terminated components. The reason for this is that in many cases, LEDs will have larger stencil apertures and larger distances between the apertures than components with finer pitches or many I/O pads. The theory is that the stencil will be better able to support having a small amount of dried paste trapped beneath the metal during the second print than a thinner web of the stencil would on a finer pitch component. Although it has not been tested in the present work, there is concern that finer pitch stencils may be damaged over time using this method.
Some questions arose during the writing of this paper and point toward possible future work. This includes the investigation of the following:
• Testing this method using various other sizes of LED components to verify that the results still hold up.
• Testing various stencil apertures and distances of the metal between them to see if finer pitch really will lead to more stencil damage when using this method.
• Using a stencil with a smaller aperture in the first print so that, even with some slump of the paste during the bake, the stencil of the second print will fully gasket to the board.
• If only specific pads are printed in the first print, does the bake step have any impact on the exposed pads prior to the second print and reflow?
A new method of attaching LED components while still using common technology has been developed. By printing the paste on the thermal pad of the LED board, then baking the board at 125°C followed by a second printing and reflow, voiding has been reduced significantly.
This method adds two additional steps to the LED attachment process—a bake step and a second print. However, if voiding is of great concern, this method may be beneficial and should be taken into consideration.
1. The Bergquist Company. “Thermal Management for LED Applications – Solutions Guide.” http://www.bergquistcompany.com/pdfs/LED_496KB.pdf
2. Homer, Seth and Ron Lasky. “Minimizing Voiding in QFN Packages Using Solder Preforms.” SMTAI, October 2011.
3. Herron, Derrick et al. “Voiding Control for QFN Assembly.” SMTA Pan Pacific, 2011.
Derrick Herron, Indium Corporation, Clinton, New York, United States of America, firstname.lastname@example.org
John Mathurin, Charlie Wilkinson
Ansen Corporation, Ogdensburg, New York, United States of America, email@example.com; firstname.lastname@example.org