BOARD LEVEL ASSEMBLY AND RELIABILITY CONSIDERATIONS FOR QFN TYPE PACKAGES

Transcription

1 BOARD LEVEL ASSEMBLY AND RELIABILITY CONSIDERATIONS FOR QFN TYPE PACKAGES Ahmer Syed and WonJoon Kang Amkor Technology, Inc S. Price Road Chandler, Arizona ABSTRACT There is a strong interest in understanding the surface mount assembly requirements of QFN (Quad Flat No-Lead) type packages due to their rapid industry acceptance. Board level reliability is also of great concern as this is a package without compliant leads. This paper provides guidelines in board design and surface mount of this package based on extensive surface mount experiments. Board level reliability data has also been generated for accelerated temperature cycling test conditions and is presented here. The data is generated for different material sets, various body/die sizes, temperature cycle conditions and board thickness. The data shows reliable surface mount process is achievable and the package is very reliable for most applications. Key words: QFN, MLF, Surface mount assembly, board level reliability, solder joint reliability. INTRODUCTION During the last couple of years the introduction of QFN (Quad Flat No -lead) package has taken the industry by storm due to its superb electrical and thermal performance characteristics. Amkor Technology markets this package as MicroLeadFrame (MLF) and has recently shipped 1 billionth package of this type. The MicroLeadFrame package (MLF) is a near CSP plastic encapsulated package with a copper leadframe substrate. This is a leadless package where electrical contact to the PCB is made by soldering the lands on the bottom surface of the package to the PCB, instead of the conventional formed perimeter leads. Amkor s ExposedPad technology enhances the thermal and electrical properties of the package. The exposed die attach paddle on the bottom efficiently conducts heat to the PCB and provides a stable ground through down bonds and electrical connections through conductive die attach material. The design of the MLF package also allows for flexibility. Its enhanced electrical performance enables the standard 2 GHz operating frequency to be increased up to 10 GHz with some design considerations. The small size and weight with excellent thermal and electrical performance make the MLF package an ideal choice for handheld portable applications such as cell phones and PDAs or any other application where size, weight and package performance are required. (a) (b) (c) Figure 1. MLF Package Photo and Cross Section Drawings. (a) Punch Singulated, (b) Saw Singulated, Full Lead (c) Saw Singulated, Lead Pullback The MLF package comes in two formats: punch singulated, and saw singulated. While punch singulated packages are individually punched from molded strip during final assembly, the saw singulated package are assembled in array format and separated into individual components during the final sawing operation. The saw singulated package is further divided into two options: Full Lead package, and Lead Pullback package. The full lead package has the whole thickness of the lead exposed on the package sides. On the other hand, the lead pullback package has a bottom half etch leadframe, resulting in only the top half of the lead thickness exposed to the sides of the package. Figure 1 shows the differences in these package configurations. As shown in Figure 1, the lands on the package bottom side are rectangular in shape with rounded edge on the inside. Since the package does not have any solder balls, package to board electrical connections are achieved by printing the solder paste on the board and reflow soldering after component placement. In order to form reliable solder joint, special attention is needed in designing the board pad pattern. Similarly, the surface mount process for this package is complicated due to presence of large exposed pad on bottom side of the package. Attention is required for proper stencil design, paste printing, and reflow profiling. The package design as well as board and test conditions also impact the board level reliability of this package. This paper deals with all of these aspects and provides guidelines and data to achieve a reliable solution for a particular application.

2 BOARD DESIGN CONSIDERATIONS Typically the PCB pad pattern for an existing package is designed based on guidelines developed within a company or by following industry standards such as IPC-SM-782. Since MLF is a new package type and the industry guidelines have not yet been developed for PCB pad pattern design, the development of proper design may require some experimental trials. Amkor has used IPC s methodology [1], with constraints added to accommodate the exposed die paddle, in designing the pad pattern. The pad pattern developed includes considerations for lead and package tolerances. Figure 2 shows the cross-sectional views of full lead and lead pullback package options for saw singulated packages. These options are also commonly referred to as Full Connecting Bar (FCB) and Half Etch Connecting Bar (HECB) designs. Notice that in full lead option the peripheral leads are extended all the way to package edges on the bottom side of the package. In case of lead pullback, the end of the leads are etched resulting in lands that are embedded in the mold compound except for the bottom side. To increase the length of the exposed leads on bottom, the leads are also ectended toward the center by 0.1mm nominal. The cross-sectional view is shown in Figure 2, comparing the two options. outside and inside to inside pad dimensions, respectively. The dimension X and Y indicate the width and the length of the pad, respectively. Two additional clearances C LL and C PL are also defined to avoid solder bridging. While C LL defines the minimum distance between land to land for the corner joints on adjacent sides, C PL defines the minimum distance between the inner tip of the peripheral lands and the outer edge of the thermal pad. E SE E2 D D2 S Figure 3. MLF (full lead design) component dimensions needed for PCB land pattern design. D D2 SD L L b e Full Connecting Bar (FCB) Design 0.4mm Nominal Lead Length Full Connecting Bar (FCB) Design 0.6mm Nominal Lead Length ZDmax D2 C LL C PL Half Etch Connecting Bar (HECB) Design 0.4mm Nominal Lead Length Half Etch Connecting Bar (HECB) Design 0.4mm Nominal Lead Length ZEmax GEmin E2 AEmax X Y HECB FCB Superimposed FCB and HECB Design HECB FCB ADmax GDmin Figure 4. PCB Pad Pattern Dimensions for footprint design. Figure 2. Full vs. Half Etch Connecting Bar Design. Figure 3 shows the bottom and side views of full lead option, indicating the dimensions needed to design the pad pattern for PCB. Although, the leads are pulled back in the HECB design, the PCB pad pattern does not need to be changed for these options. Since most packages are square with dimension D equal to dimension E and the leads are along the E direction for dual packages, the side view dimensions (D, S, D2, & L) are used to determine the land length on the PCB. The PCB pad pattern dimensions needed for the footprint design are shown in Figure 4. In the figure, the dimensions ZD max and GD min (and ZE max and GE min ) are the outside to The development of pad pattern for MLF requires tolerance analysis with consideration of a) component tolerances, b) PCB fabrication tolerances, and c) the accuracy of the equipment used for placing the component. IPC has established a procedure for such an analysis [1], which has been used to determine the recommended pad patterns for various MLF packages. The complete details of the methodology and recommended pad patterns are documented in an application notes [2] for this package which can be downloaded from The details are not repeated here due to space limitations. SURFACE MOUNT PROCES S CONSIDERATIONS Because of the small lead surface area and the sole reliance on printed solder paste on the PCB surface, care must be

3 taken to form reliable solder joints for MLF packages. This is further complicated by the large thermal pad underneath the package and its proximity to the inner edges of the leads. Although the pad pattern design suggested above might help in eliminating some of the surface mounting problems, special considerations are needed in stencil design and paste printing for both perimeter and thermal pads. Since surface mount process varies from company to company, careful process development is recommended. The following provides some guidelines for stencil design based on Amkor s experience in surface mounting MLF packages. Stencil Design for Perimeter Pads The optimum and reliable solder joints on the perimeter pads should have about 50 to 75 microns (2 to 3 mils) standoff height. The first step in achieving good standoff is the solder paste stencil design for perimeter pads. The stencil aperture opening should be so designed that maximum paste release is achieved. This is typically accomplished by considering the following two ratios: Area Ratio = Area of Aperture Opening / Area of Aperture Wall, and Aspect Ratio = Aperture width / Stencil Thickness For rectangular aperture openings, as required for this package, these ratios are given as Area Ratio = LW/2T(L+W), and Aspect Ratio = W/T Where L and W are the aperture length and width, and T is stencil thickness. For optimum paste release the area and aspect ratios should be greater than 0.66 and 1.5, respectively. It is recommended that the stencil aperture should be 1:1 to PCB pad sizes as both area and aspect ratio targets are easily achieved by this aperture. The stencil should be laser cut and electro polished. The polishing helps in smoothing the stencil walls resulting in better paste release. It is also recommended that the stencil aperture tolerances should be tightly controlled, especially for 0.4 and 0.5mm pitch devices, as these tolerances can effectively reduce the aperture size. Since not enough space is available underneath the part once soldered to the board, it is recommended that No Clean, Type 3 paste be used for mounting MLF packages. Nitrogen purge is also recommended during reflow. Stencil Design for Thermal Pad In order to effectively remove the heat from the package and to enhance electrical performance the die paddle needs to be soldered to the PCB thermal pad, preferably with minimum voids. However, eliminating voids may not be possible because of presence of thermal vias and the large size of the thermal pad for larger size packages. Also, out gassing occurs during reflow soldering, which may cause defects (splatter, solder balling) if the solder paste coverage is too large. It is, therefore, recommended that smaller multiple openings in stencil should be used instead of one big opening for printing solder paste on the thermal pad region. This will typically result in 50 to 80% solder paste coverage. Shown in Figure 6 are some of the ways to achieve these levels of coverage. 1.5mm dia. 1.6 mm Pitch : 37% 1.35x1.35 mm 1.65 mm Pitch : 68% 1.0mm dia. 1.2 mm Pitch : 50% 1.35x1.35mm 1.5 mm Pitch : 81% Figure 5. Thermal Pad Stencil Designs for 7x7 and 10x10mm MLF Packages. Via types and solder voiding As the MLF package incorporates a large center pad, controlling solder voiding within this region can be difficult. Small voids with greater than 50% solder coverage under the exposed pad do not result in thermal or board level performance degradation in general. However, voids within solder joints under the exposed pad can have an adverse effect on high speed and RF applications as well as on thermal performance if the maximum size for a void is greater than via pitch in the thermal pad area. In order to control these voids, solder masking may be required for thermal vias to prevent solder wicking inside the via during reflow. This wicking displaces the solder away from the interface between the package die paddle and thermal pad on the PCB. There are different methods employed within the industry to avoid such wicking, such as via tenting (from top or bottom side) using dry film solder mask, via plugging with liquid photo-imagible (LPI) solder mask from the bottom side, or via encroaching. These options are depicted in Figure 6. In case of via tenting, the solder mask diameter should be 100 microns larger than via diameter. All of these options have pros and cons when mounting MLF package on the board. While via tenting from top side may result in smaller voids, the presence of solder mask on the top side of the board may hinder proper paste printing. On the other hand, both via tenting from bottom or via plugging from bottom may result in larger voids due to outgassing, covering more than two vias. Finally, encroached vias allow the solder to wick inside the vias and reduce the size of the voids. But this also results in lower standoff of the package, which is controlled by the solder underneath the exposed pad. Figure 7 shows representative x-ray pictures of MLF packages mounted on boards with different via treatments.

4 (a) (b) (c) (d) Figure 6. Solder mask options for thermal vias: (a) via tenting from top, (b) via tenting from bottom, (c) via plugging from bottom, and (d) via encroached from bottom. (a) (b) Figure 1 B11-U10 Figure 1 B11- U22 (c) (d) Figure 7. Voids in Thermal Pad Solder Joint for (a) vias tented from top, (b) vias tented from bottom, (c) via plugged from bottom, and (d) via encroached from bottom surface of the PCB. Encroached via, depending on the board thickness and amount of solder printed underneath the exposed pad, may also result in solder protruding from the other side of the board, as shown in Figure 8. Note that the vias are not completely filled with solder, suggesting that solder wets down the via walls until the ends are plugged. This protrusion is a function of PCB thickness, amount of paste coverage in the thermal pad region, and the surface finish of the PCB. Amkor s experience is that this protrusion can be avoided by using lower volume of solder paste and reflow peak temperature of less than 215 o C. If solder protrusion cannot be avoided, the MLF components may have to be assembled on the top side (or final pass) assembly, as the protruded solder will impede acceptable solder paste printing on the other side of the PCB. Solder joint standoff height The solder joint standoff is a direct function of amount of paste coverage on the thermal pad and the type of vias used for MLFs with exposed pad at the bottom. Board mounting studies by Amkor in partnership with customers have clearly shown that the package standoff increases by increasing the paste coverage and by using plugged vias in the thermal pad region. This is shown in Figure 9 below. The standoff height varies by the amount of solder that wets or flows into the PTH via. The encroached via provides an easy path for solder to flow into the PTH and decreases package standoff height while the plugged via impedes the flow of solder into via due to the plugged via closed barrel end. In addition, the number of vias and their finished hole sizes will also influence the standoff height for encroached via design. The standoff height is also affected by the type and reactivity of solder paste used during assembly, PCB thickness and surface finish, and reflow profile. To achieve 50 micron thick solder joints, which help in improving the board level reliability, it is recommended that that the solder paste coverage be at least 50% for plugged vias and 75% for encroached via types. Standoff Height (mils) PLUGGED 37% COVERAGE PLUGGED 67% ENCROACH 37% ENCROACH 67 % Paste PLUGGED 50% PLUGGED 81% ENCROACH 50% ENCROACH 81% 48I/O 48I/O 48I/O 48I/O 68I/O 68I/O 68I/O 68I/O Figure 9. Standoff height as a function of Via type and paste coverage. Figure 8. Solder protrusion from the bottom side of PCB for encroached vias. Reflow profile and peak temperature also have a strong influence on void formation. Amkor has conducted experiments with different reflow profiles (ramp -to-peak vs. ramp -hold-ramp), peak reflow temperature, and time above liquidus using Alpha Metal s UP78 solder paste. Generally, it is found that the voids in the thermal pad region for plugged vias reduce as the peak reflow temperature is increased from 210 o C to o C. For encroached vias, it is found that the solder extrusion from the bottom side of the board reduces as the reflow temperature is reduced. BOARD LEVEL RELIABILITY CONSIDERATIONS Since MLF is a leadless package, reliability of solder joints is one of the major concerns for this package. The board level reliability is affected by material and design parameters of the package itself as well as by the board design and thickness. Amkor has generated extensive data to evaluate the board level reliability both for temperature cycling and drop conditions. Multiple package sizes were tested to investigate the effect of package design and material parameters, board parameters, and temperature cycle condition. All tests were conducted with exposed pad of the packages soldered to the test board, unless specifically mentioned otherwise. The data presented below was generated using three different test conditions: 1. TC1: -40 to 125 o C, 1 cycle/hour, 15 minutes ramps and dwells 2. TC2: -55 to 125 o C, 2 cycles/hour, 2 minutes ramps, 13 minutes dwells

5 3. TC3: 0 to 100 o C, 2 cycles/hour, 10 minutes ramps, 5 minutes dwells The footprints of all test boards were designed according to the calculations described in the application notes [2]. All packages had daisy chain leadframe and dummy silicon die inside the package to simulate the mechanical and material construction of a real package. The test boards had alternating daisy chains so a complete net was formed once the package was attached to the board. The electrical continuity of these nets was monitored through out the test to detect opens as soon as they occur. The nets were scanned and electrical resistance was measured every 2 minutes. A resistance value beyond 500 ohms threshold was considered as open. Any such open was confirmed by 15 additional opens within 10% of the time of first open. The opens were manually confirmed also before declaring the net as failed. Effect of Package Material and Design Mold Compound Material: Selecting the proper mold compound was one of the main tasks in developing this package. The selection was primarily based on two criteria: 1. The selected mold compound must result in meeting the minimum package level reliability requirements, i.e., moisture sensitivity level, and 2. The selected mold compound must result in acceptable package to board attachment reliability. To meet these requirements actual tests were conducted for both package level and board level reliability. The package selected for this evaluation was 7 x 7 mm in size with 48 lead (12 leads on each side) at 0.5mm pitch and 3.8 x 3.8 mm dummy die inside. The packages were mounted on a 1.6mm thick, 4 layer FR-4 board with OSP surface finish. The board level reliability was evaluated under TC2 condition. Table 1. Mold Compound Material properties (supplier data) and BLR Result Summary Mold Compound alpha 1 alpha 2 (ppm/ o C) (ppm/ o C) Tg ( o C) Modulus (kg/mm 2 ) Cycles Completed # of Failures 1st Failure Mean Life EMC EMC EMC EMC EMC EMC EMC NA EMC Table 1 above provides the material properties as well as the summary of failure data. As expected, the data revealed that the board level reliability is directly dependent on the CTE of the mold compound. The compounds with lower CTE values performed worse than the ones with CTE value closer to the board CTE (around 17 ppm/ o C). Also, the compounds with lower CTE generally have higher modulus, resulting in a stiffer package. Based on the results above as well as the results from package level reliability (MSL evaluations), EMC7 was selected as the mold compound of choice for this package. Die Size: The die size inside the package can have a significant effect on the board level reliability of the package. To quantify this affect, a number of board level tests were conducted on 7 x 7mm-48 lead, 10x10mm-68, and 12x12mm-100 lead packages. The tests were run using two test conditions (TC1 and TC2) and two board thickness (0.8 and 1.6mm). Table 2 provides the summary of package, board, and temperature cycle condition used for these evaluations. Figure 10 shows a plot of first failure point Vs Die to Package size ratio. The data is normalized for the same test condition and board thickness using the acceleration factor multipliers for these variables. These acceleration factors are presented in a later section of this paper. The plot shown is for first failure based on sample size of 30, as not enough failures were observed to calculate the mean life for small die/package ratio. The plot also shows the best fit to this data as well as the fitted equation. It is clear that the board level reliability is dependent on this ratio, not the actual die and package size. The life can be very low if die/package size ratio is very high, but increases non-linearly for lower values of die to package ratio. Table 2: Test matrix for die size effect evaluation. Package Size 7mm -48 7mm-48 7mm-48 10mm-68 10mm-68 12mm-100 Package Size Die Size Board Thickness Test Condition TC2 TC2 TC1 TC1 TC1 TC1 Fatigue Life (1st Failure) y = x R 2 = Die to Package Size ratio Figure 10. Effect of Die to Package ratio on board level reliability Land Size: Since MLF is a leadless package, the reliability of solder joints is also a function of package land size. This effect is shown in Figure 11 by comparing 7mm-48 lead and 7mm-28 lead packages. Both packages used the same die size (5.1 x 5.1 mm), however, have different lead pitches; 0.5mm for 48 lead package and 0.8mm for 28 lead package. Because of higher pitch for 7mm-28 package, the land size of this package is 0.28 x 0.6 mm compared to 0.23 x 0.4 mm for 48 lead package. The Weibull plot, shown in Figure 11, clearly shows that this increase in land size resulted in

6 2X improvement in fatigue life. The larger land results in wider and longer solder joints, and thus longer path for the crack to go through before a complete separation is achieved. Cumulative % Failed β1=8.60, η1= , ρ=0.98 β2=9.61, η2= , ρ=0.93 Cycles to Failure Weibull 7mm-28 F=30 / S=0 7mm-48 F=23 / S=7 Figure 11. Land size effect on fatigue life. 7mm-28 lead package has bigger lands compared to 7mm-48 lead package. Full Vs Half Etched Leadframe: As explained in an earlier section (Figure 2), the package lands come in two options; a) a full lead option, and b) a half etched option with lead pull back. To investigate if this half etch lead pull back has any deleterious consequence on board level reliability, test are being conducted on 5mm-32 lead package with both options. The packages on test have 2.54 x 2.54 mm dummy die inside and are mounted using the same footprint on the board. TC1 test condition is being used for this evaluation with the board thickness of 1.6mm. As of this writing more than 3000 cycles have been completed without any failure for both of these options. This shows that lead pull back will not cause reduction in board level reliability as the land length is the same for both options. It should be noted that the land length for these packages is 0.4mm nominal, which is smallest that Amkor offers for these packages. Effect of Board Thickness Since the MLF package is targeted for multiple applications that require quite different board thickness, understanding the effect of board thickness on board level reliability was also investigated. Tests were conducted on 10mm-68 lead as well as 5mm-32 lead packages using 0.8 and 1.6mm thick boards. The die sizes were 7.62 mm square and 2.54mm square for 10mm and 5mm packages, respectively. The effect of board thickness was evaluated using TC1 test condition. Figure 12 shows the effect of board thickness for 10mm-68 lead package. The data shows that mounting the package on 1.6mm thick board results in 33% reduction in board level reliability. This is consistent with other packages as thicker board provides stiffer assembly, resulting in lower life. The tests on 5mm-32 lead packages are still on-going. At the time of this writing, the packages mounted on 1.6mm thick board started failing at around 6300 cycles, while the packages on 0.8mm thick board have not failed yet at the end of 8100 cycles. Based on this data, the life for thinner board is at least 30% higher. Since no failures have occurred on 0.8mm thick board, this life improvement is expected to be in the same range as that for 68 lead package. Cumulative % Failed mm Thick Board 0.8mm Thick Board β1=13.69, η1= , ρ=0.97 β2=9.42, η2=854.83, ρ=0.94 Cycles to Failure Weibull F=26 / S=4 F=29 / S=1 Figure 12. Weibull plot showing thicker board resulted in lower reliability. Effect of Exposed Pad Mounting The MLF package has a big exposed pad on the bottom of the package that is supposed to be soldered to the board for enhanced board level reliability, and electrical/thermal performance. However, because of routing considerations and actual application requirements, the exposed pad may not be soldered to the board. It should be realized that this will impact the board level reliability, which can be significant in some case. Cumulative % Failed Paddle Not Soldered β1=9.55, η1= , ρ=0.98 β2=10.84, η2= , ρ=0.88 Paddle Soldered Cycles to Failure Weibull F=28 / S=2 F=28 / S=2 Figure 13. Weibull plot showing the effect of not soldering exposed pad to the board.

7 Figure 13 shows the comparison of board level reliability for 10mm-68 lead packages mounted on the board with and without exposed pad soldered to the board. The evaluation was performed using TC1 test condition and 0.8mm thick boards. The Weibull plot shows a 20% reduction in board level reliability when exposed pad is not soldered to the board. Tests are also being conducted for exposed pad effect using 5mm-32 lead packages, TC1 condition, and 0.8mm thick board. Although the tests are not completed yet, the current data shows at least 60% improvement in life when the exposed pad is soldered for this particular package. Effect of Temperature Cycle Condition It is well known that depending upon the accelerated test condition used, the board level reliability of a package can be significantly different. The same is true of MLF package, as shown in the Table 3 below. The data shows that for a 5mm-32 lead package, the board level reliability can be 20% lower for TC2 condition compared to TC1 condition. Similarly, since no failures have occurred so far when using TC3 test condition, the reliability can be at least 75% higher for this condition compared to TC1 condition. This is because of faster ramp rates for TC2 and lower temperature extremes for TC3, respectively. The data is consistent with other packages, which show a similar trend. Table 3. Effect of temperature cycle condition Body Size Lead Count Pitch Board Thickness (mm) Test Type TC1 TC2 TC3 Cycles Completed # of Failures # Parts on Test st Failure (cycles) NA Mean Life (cycles) NA 5960 NA BOARD LEVEL RELIABILITY ENHANCEMENTS The data presented above clearly shows that the MLF package has superb board level reliability characteristics, except for the case where die to package size ratio is greater than 70%. Only for these extreme cases, the reliability can be below 1000 cycles in TC1 condition, but still better than 500 cycles requirement for most handheld and consumer electronic applications. However, some other applications, such as automotive and network hardware, will require enhancements in board level reliability for large die to package ratio designs. There are two possible ways to further improve the reliability: (a) increase the standoff height, and (b) solder fillet formation. These options are discussed below. Bumped MLF While increased standoff can be achieved by using thicker stencil, there are limits to this option due to aperture area and aspect ratio requirements for paste release. Also, since multiple types of components are mounted on the same board, using a thicker stencil for one or two components is not desirable. To resolve this issue, Amkor has developed Bumped MLF option. The Bump refers to thicker lead frame plating on the underside of the package, as shown in Figure 14. The resulting bump height can be as much as 4 mils (100 um), which in turn results in increased standoff (joint thickness) by 4 mils. Although FEA simulations showed that this can potentially increase the reliability by 2X, actual tests were conducted to further prove this concept. Figure 14. Plate-up Bumped MLF option. Plate-up Bump on Lead Plate-up Bump on Die Paddle Figure 15 shows the results of board level reliability test on bumped Vs non-bumped standard package (7mm, 48 lead, 3.81mm die). The packages were mounted on 0.8mm thick board using Sn4.0Ag0.5Cu paste and are being tested under TC1 condition. Although not enough failures have been observed so far, the data clearly shows 2X improvement in board level reliability for bumped version. Cumulative % Failed Non Bumped (Std) Bumped β1=10.08, η1= , ρ=0.97 β2=14.00, η2= , ρ=0.97 Cycles to Failure Weibull F=5 / S=25 F=7 / S=23 Figure 15. Weibull plot showing the effect of bumping the package lands. The bumped package also has the advantage in rework over the standard package. If the package is replaced with a new one during board rework process, solder paste needs to be printed either on the board site or to the underside of the new package at the board assembly site. However, with an already bumped package, this paste printing operation can be avoided during rework at the board assembly site. Data collected by customers have validated this aspect of bumped package.

8 Standoff Height and Solder Fillet Formation The MLF packages have either full or half the thickness of lead exposed on the sides of the package. However, since the plating operation is done prior to package singulation, the exposed sides of the leads are not plated and have bare copper surface. Since bare Cu oxidizes readily in an uncontrolled environment, the package falls into the category of Bottom Only Termination as per IPC/EIA J-STD-001C. This joint industry standard does not require fillet formation on the side of the package for the packages in this category. Due to factors not controlled by Amkor, the company does not guarantee the fillet formation on the side of the package during board assembly. In fact, the fillet formation is solely dependent on the surface mount process. Some of the variables that control this fillet formation are: solder paste used and its flux activity level, PCB land size, printed solder volume, and the package standoff height. The pad size calculated above along with 1:1 aperture will provide sufficient solder for fillet formation if the package standoff is not excessive. Since there is only limited solder available, higher standoff - controlled by paste coverage on the thermal pad may not leave enough solder for fillet formation. Conversely, if the standoff is too low, large convex shape fillets may form. This is shown in Figure 16. Since center pad coverage and via type were shown to have the greatest impact on standoff height the volume of solder necessary to create optimum fillet varies. Package standoff height and PCB pads size will establish the required volume. extends the length of solder joint and provides a longer path for the crack to go through the entire joint, thus improving the reliability. However, since fillet formation and standoff height are not independent of each other, the solder joint reliability may not increase as much due to fillet formation if the standoff height is too low (less than 2 mils). Also, the improvement in life goes down as the package becomes more compliant (lower die to package size ratio, longer lands, smaller packages). It should be realized that the added improvement in life is not needed for most applications. CONCLUSIONS The data presented in this paper shows that the attachment reliability of MLF package is excellent and can be further improved for extreme applications. This, however, requires careful attention in board design as well as in assembly process development. Proper footprint design and assembly processes will yield reliability far exceeding the requirements for most applications. Amkor has optimized the package design and material selection for this package with focus on both board level and package level reliability and offers options (such as bumped MLF) to enhance the reliability and re -workability of this package. ACKNOWLEDGEMENTS The authors would like to thank MLF engineering team (Jing Alaban, JD Kwon, Bob Bancod, and Terry Davis) for their continued support in board level reliability evaluations and preparing samples for various studies reported here. Acknowledgements are also due to YH Ka and YJ Kim for carrying out most of the tests and failure analysis. Last, but not least, thanks are also due to Cisco Systems, Solectron Corporation, Xetel Corporation, and Delphi Automotive for their collaboration in surface mount studies. (a) (b) REFERENCES 1. IPC-SM-782, Surface Mount Design and Land Pattern Standard. 2. (c) (d) Figure 16. Solder fillet shape as a function of paste coverage in the thermal pad, via type, standoff, and PCB land size. (a) 37% paste coverage, Plugged via, 1.4 mils standoff (b) 37% paste coverage, Encroached via, 0.6 mils standoff (c) 50% paste coverage, Plugged via, 2.9 mils standoff (d) 81% paste coverage, Encroached via, 2.1 mils standoff Finite element simulations and actual test data generated by customer have shown that the fillets - if formed - can improve the board level reliability by as much as 2X for a package with large die to package size ratio. The fillet