Ribbed Package Geometry for Reducing Thermal
Warpage and Wire Sweep During PBGA
Encapsulation
Sen-Yeu Yang, Shin-Chang Jiang, and Wen-Shu Lu
Abstract—The effects of changing package shape into ribbed
geometry on thermal warpage and wire sweep of PBGA are in-vestigated in this paper. Three rib geometries (border, diagonal, and cross) with a variation of rib widths and thicknesses are com-pared with the original plane geometry. Finite element analyses of thermal warpage during the reflow process of PBGA molding with and without ribbed geometry are carried out. Numerical modeling shows that the border rib has the least thermal warpage at the re-flow condition. Flow visualization was performed to study the ef-fect of rib geometry on wire sweep, and demonstrates that wire sweep in the ribbed packages is significantly less than that in the original nonribbed package.
Index Terms—Finite element analysis, flow visualization, plastic
ball grid array, ribbed package geometry, thermal warpage, wire sweep.
I. INTRODUCTION
B
ALL grid array (BGA) packages have emerged as the package of choice. An explosive growth in research and development efforts has been witnessed in recent years [1]. Plastic BGA (PBGA) employs resins as encapsulant for eco-nomical and productive reasons. Although PBGA has several advantages over fine-pitch quad flat pack (QFP) such as smaller package area and higher inputs/outputs (I/Os), potential relia-bility problems such as coplanarity have to be solved. PBGA is basically a composite structure consisting of different materials with different elastic moduli and coefficients of thermal expan-sions (CTEs). The thermal expansion mismatch between the materials generates thermomechanical stress at temperatures above and below the molding temperature and especially at the reflow temperature. The excessive thermomechanical stresses generated can result in warpage problems as shown in Fig. 1. This coplanarity-related concern has been increasing rapidly as a consequence of trends for achieving a high I/O density with a package that is as small and thin as possible. Liang [2], Ootaet al. [3], Ho et al. [4], and Mertol [5], investigated the effect
of geometry such as thickness and dimension, and material
Manuscript received August 8, 1999; revised October 19, 2000. This paper was recommended for Publication by Associate Editor R. Chanchani upon evaluation of the reviewers’ comments. This work was supported by National Taiwan University.
S.-Y. Yang is with Department of Mechanical Engineering, National Taiwan University, Taipei 106, Taiwan, R.O.C.
S.-C. Jiang is with Branchy Vacuum Technology Company, Ltd., Lung-Tang, Taiwan, R.O.C.
W.-S. Lu is with Acer Communication and Multimedia, Inc., Taipei, Taiwan, R.O.C.
Publisher Item Identifier S 1521-3331(00)10597-5.
Fig. 1. Warpage of plastic ball grid array package may cause coplanarity problem in solder balls during reflow.
properties such as CTE, Tg, and elastic moduli on coplanarity and their sensitivity. It was found that substrate warpage is the main cause of poor coplanarity. Increasing substrate thickness, higher substrate Tg and Young’s Moduli, and thinner encapsulant can reduce warpage. No research has attempted to investigate the effect of redistributing encapsulant into a ribbed geometry on warpage of PBGA. Ribbing is a common practice in enhancing structural rigidity [6]. This paper is devoted to investigating the possibility of changing package geometry by adding a rib structure to reduce thermal warpage during the reflow condition. Numerical stress and structure analysis based on Finite-Element-Method is employed to predict the thermal warpage of a package with the original plane structure and various ribbing structures.
While developing an innovative ribbed package geometry for reducing thermal warpage, wire sweep is another challenging problem. Wire sweep is a common defect in plastic integrated high density packages. Most integrated circuits are encap-sulated by transfer molding of epoxy molding compounds (EMCs). During the filling stage, the high viscosity epoxy exerts a drag force on the wires. If the wire sweep is too big, neighboring wires may touch, resulting in a short-circuit. With the trend of high wire density, thinner cavity depth, and higher filling rates, the wire sweep problem is becoming critical. To minimize wire sweep, a general practice involves adjusting lead frame design, mold design, device layout, molding compound and process parameters based on experience and trial-and-error. These practices are expensive and time-consuming. Many researchers have developed analytical and numerical models to predict the wire sweep [7]–[10]. Flow visualization experiment is another effective technology to comprehend the flow-related phenomena. Many researchers have successfully applied the visualization technique to unveil process mysteries [11]–[14]. In this study, the encapsulating mold is covered with a trans-parent top plate and a CCD camera is used to observe the cavity filling process during transfer molding. The wire sweep during filling cavities for original plane package and ribbed packages are to be measured and compared.
(a) (b)
(c) (d)
Fig. 3. Sketches showing geometries of packages (a) original plane package, (b) package with border ribs, (c) package with cross ribs, and (d) package with single diagonal rib.
II. FINITEELEMENTMODEL
The thermal warpages during reflow process in plane and ribbed PBGA packages are studied using a commercial finite el-ement software ANSYS. The original plane package under in-vestigation is a 196-pin perimeter cavity-up PBGA. There are four arrays of peripheral solder balls with 1.27 mm pitch. The solder material is eutectic Pb–Sn solder (63 Sn/37 Pb). A 8 mm 8 mm 0.5 mm IC chip is bonded on to a 21 mm 21 mm 0.6 mm bis-maleimide-triazine (BT) substrate with 0.07 mm thick thermally conductive adhesive. The package is square with body size of 18 mm 18 mm 1.3 mm. Due to the symmetry of the package only one quarter of the package (as shown in Fig. 2) is modeled using 3-D 10-node tetrahedral element, solid87 in ANSYS. Mechanical properties of all materials are assumed to be isotropic linear elastic, except BT substrate. The model con-sists of the chip, die attach, epoxy molding compound (EMC), solder ball, and substrate.
There are three types of ribbed packages investigated in this research: 1) “border”: 18 mm 18 mm 1.3 mm package sur-rounded with a rib as shown in Fig. 3(b); 2) “cross”: 18 mm 18 mm 1.3 mm with two cross ribs as shown in Fig. 3(c); 3) “diagonal”: 18 mm 18 mm 1.3 mm with a diagonal rib as shown in Fig. 3(d). The combination of widths and thicknesses of the ribs are shown in Table I. The summarized dimensions
TABLE II
MATERIALPROPERTIES ANDIMPORTANTDIMENSIONSUSED INTHETHERMAL
WARPAGESIMULATION
TABLE III
COMBINATION OFWIDTH ANDTHICKNESS FORBORDERRIBS(ON18 mm2 18 mm2 1.2 mm PACKAGE)
and material properties used in this simulation are shown in Ta-bles II and III.
The thermomechanical analyses are used to predict the nodal deflection distribution. Since one quarter of the package is mod-eled, the symmetry boundary conditions are used along the cut planes where the adiabatic condition is prescribed. To model the reflow process, the packages are heated to 160 C. The thermal loading is assumed to be 135 C above the stress free refer-ence temperature ( 25 C). The deflections of the solder balls at 160 C are then compared to the nonstressed condi-tion. Warpage is defined to be the maximum deflection differ-ence among all bottom nodes of solder balls. The warpage in
Fig. 4. Schematic of facilities for flow visualization experiment.
Fig. 5. Sketch showing the detailed construction of mold: 1) transparent top plate, 2) cavity plate, 3) substrate plate, 4) silicone seal, and 5) base plate with gate. The flow entrance, gate, upstream, and downstream wires are also marked.
the ribbed packages are compared with warpage in the original plane package.
III. EXPERIMENTALSETUP FORFLOWVISUALIZATION
To help understand the effects of the ribbed package geom-etry on the filling process and wire sweep during the transfer molding of the epoxy molding compound (EMC), wire displace-ments are observed during the transfer molding process for the original (no rib) and border-rib cavities. The flow-visualization set-up is shown in Fig. 4. A scaled-up mold with transparent window was constructed for this study as shown in Fig. 5. The mold was composed of four plates. The top plate is made of transparent poly(methyl-methacrylate) (PMMA) for flow visu-alization. The top plate, cavity plate and substrate plate form the package cavity (dimensions shown in Fig. 6). A square PMMA chip (30 mm 30 mm 1 mm) with four brass wires is attached on the substrate plate. The brass wire used is EDM wire HTF (HITACHI, Japan) The base plate provides the resin entrance and runner. The epoxy resin used is D.E.R.331 (Dow Chemical). Since its viscosity is sensitive to temperature, a temperature monitoring system is implemented between the pot and cavity. The transfer molding experiments were conducted with a mate-rial testing machine (MTS-810, MTS). The epoxy is pushed at constant flow rate from pot to cavity. The flow rate can be reg-ulated with specified stroke control through the machine con-troller. A digital CCD camera (SONY XC-77) with image pro-cessing card (Meteor 330 S/N AE 68577) is used to capture the cavity filling process.
Fig. 6. Sketch showing (a) geometry of the cavity and (b) typical wirebond geometry.
IV. RESULTS ANDDISCUSSION
A. Distribution of Thermal Deflections of Original Package and Ribbed Packages
Fig. 7 shows the distribution of thermal deflections in the original package. Several iso-deflection lines are marked with values. There is a calculated deflection value associated with each node. The largest substrate deflection occurs at the outer corner. Similar distributions are found in a package with border rib, a package with diagonal rib and a package with cross rib.
The coplanarity will be determined by the variation in deflec-tions of solder balls. The warpage is defined as the maximum difference among deflections of the central bottom nodes in the 48 solder balls.
Fig. 8 presents the warpage for the ribbed packages versus warpage for the original package. The term “warpage relative to original” is defined as the warpage (maximum deflection dif-ference among bottom nodes) in ribbed package divided by the warpage in original plane package. The border, diagonal, and cross ribs are presented in multiple columns to show results of different rib dimensions (e.g. w1-t 0.5 is the case with rib width 1 mm and rib height 0.5 mm) for each rib geometry. A value of 100% indicates the same warpage as that in original plane package under reflow condition. A value of warpage rel-ative to original of less than 100% implies reduce in warpage as compared to the original plane package. There is an im-provement in coplanarity. A positive value indicates the overall concave-convex configuration is similar to that predicted in the original plane package, while a negative value implies a con-trary pattern.
According to the results of numerical analysis, warpage rel-ative to original of all ribbed packages are all less than 100%. This implies that warpages in ribbed packages with specified dimensions are all smaller than that in original plane package.
Fig. 7. Distribution of deflections under reflow condition in plain (nonribbed) package at (a) EMC and substrate and (b) solder balls and substrate. Warpage is defined as the maximum difference among deflections of the central bottom nodes of the solder balls. Gradient lines:A = 04:77 m; B = 01:97 m;
C = 0:841 m; D = 3:65 m; E = 6:46 m; F = 9:26 m; G = 12:1 m; H = 14:9 m.
Fig. 8. Warpages in packages with various rib configurations and dimensions normalized by the warpage in the original plane package. Values less than 100% imply that warpage in packages with ribs is less than that in the original plain package.
Fig. 9. Warpage relative to plain package in packages with border ribs of various dimensions. The reduction in thermal warpage is significant even with thickness of EMC reduced.
Fig. 10. Effects of (a) coefficient of thermal expansion and (b) elastic modulus of ribs in thermal warpage of packages with border ribs relative to original plane package. Increasing rigidity of rib reduces warpage, but CTE of the rib should be close to that of EMC.
Furthermore, it is found that adding border ribs is more effec-tive in reducing thermal warpage than adding cross and diagonal ribs.
B. Effects of EMC Thickness, Dimensions, CTE, and Rib Elastic Modulus on Warpage
Packages with border ribs are the best among the three ribbed packages, not only because of reduced warpage overall, but also due to the convenience for subsequent marking and printing. The following studies of changing package thickness and ma-terial properties will be limited to the package with border ribs. Fig. 9 shows the attempt to reduce the original EMC thick-ness in packages with border ribs. The warpage relative to
orig-Fig. 11. Photographs show the filling process and wire sweep in cavity for original plane package. Wire sweep is observed after the epoxy resin flows across the bondwires.
inal is defined as the warpage in 1.2 mm thick package with border rib of specified dimension divided by the warpage in 1.3 mm thick plane package without rib. As can be seen, with the EMC thickness reduced from 1.3 mm to 1.2 mm, it is still pos-sible to achieve much less warpage in the border rib packages for all rib dimensions. This implies that a improved coplanarity can be achieved even with less EMC material by redistributing the EMC from plain geometry to more bending-resistant ribbed geometries.
It is possible that a different material could be used for the ribs. For example, one could select copper as a rib material for enhancing heat transfer. However, the difference in CTE and elastic modulus may affect the relative warpage. Fig. 10 demonstrates the sensitivity of warpage to rib CTE [Fig. 10(a)] and modulus of elasticity [Fig. 10(b)]. For CTE, deviation from the CTE of original EMC worsens the warpage. However, in-creasing the elastic modulus of the rib reduces the warpage.
C. Comparison of Wire Sweeps in Border Ribbed and Original Package Cavities
Fig. 11 displays the photographs reproduced from the CCD frames showing the cavity filling process during transfer
Fig. 12. Photographs show the filling process and wire sweep in cavity for package with border ribs. Reduced wire sweep is observed because the excessive flow along the thick channel for molding border rib reduces the flow across the bondwires.
molding of EMC in the original cavity. The profile of the resin fronts is parabolic, and it’s profile changes with time. After the resin passes the chip, the resin front is retarded, showing the chip exerting significant resistance on flow. Wire sweep can be observed as soon as resin flows across the bondwires.
Fig. 12 shows the flow patterns during filling the cavity for the border rib package. Due to the racetrack effect (i.e., resin flows faster in the thicker sections), the profile of the resin front is straighter than that in a original cavity. In addition, less wire sweep is observed in cavity for package with border rib.
Fig. 13 shows the maximum wire sweeps measured in cavities for the original package and ribbed packages at different flow rates. The wire sweep s is evaluated by the ratio , where is the maximum lateral wire deflection projected onto the plane of the lead frame and is the span of the bondwire. Wire sweep increases with flow rate. The flow rate is expressed in term of Reynolds number Re. Re is defined as / , where is the average flow velocity during filling, is the cavity height, and is the viscosity of epoxy resin during filling. For the same Reynolds number, the wire sweeps in the cavity for ribbed pack-ages are smaller than wire sweeps seen in the cavity for original plane package. A relatively large volume of resin flows through
Fig. 13. Wire sweep in cavity for packages with ribs is significantly smaller than that in cavity for original plane package.
the cavity via the rib passages, lowering the velocity over the chip, and thus exerting less drag force on wire. As a result, the wire sweeps in the ribbed cavities are significantly less than for the original cavity.
V. CONCLUSIONS
The effect of changing package shape by redistributing plain epoxy molding compound (EMC) into ribbed geometries on thermal warpage of PBGA has been investigated in this study. Finite element analysis of thermal warpage during encapsula-tion of PBGA packages with ribbed geometries is performed. Three types of ribbing (border, diagonal and cross) with a num-bers of combinations of rib widths and thicknesses are com-pared. In addition, the effect of such ribbed geometries on wire sweep is also measured by flow visualization experiments. The following conclusions can be drawn from this study.
1) The thermal warpage during reflow condition in ribbed packages (especially the package with the border rib) can be significantly reduced if the thickness and width of the rib are appropriately selected. Optimal rib dimensions de-pend on rib arrangement and EMC thickness. The reduc-tion in thermal warpage can still be significant even with the thickness of the EMC being reduced.
2) Changing the thermal expansion coefficient of the border rib would not help reduce thermal warpage. However, in-creasing the Young’s Modulus of the border rib would reduce thermal warpage.
3) The wire sweeps in cavities with ribs are significantly smaller than that in nonribbed cavities. For ribbed pack-ages, the additional flow area in the mold cavity due to a rib (especially the border rib) serves as a flow guide, re-ducing the flow rate over the chip, and thus drag exerted on the wires.
1995.
[2] D. Liang, “Warpage study of Glob Top Cavity-UP EPBGA Packages,” in Proc. IEEE Electron. Comp. Technol. Conf., 1996, pp. 694–701. [3] K. Oota and K. Shigeno, “Development of molding compounds for
BGA,” in Proc. IEEE Electron. Comp. Technol. Conf., 1995, pp. 78–85. [4] T. H. Ho, J. Y. Lee, R. S. Lee, and A. W. Lin, “Linear finite element stress simulation of solder joints on 225 I/O plastic BGA package under thermal cycling,” in Proc. IEEE Electron. Comp. Technol. Conf., 1995, pp. 930–936.
[5] A. Metrol, “Application of the Taguchi method on the robust design of molded 225 plastic ball grid array packages,” IEEE Trans. Comp.
Packag. Manufact. Technol. B, vol. 18, pp. 734–743, Nov. 1995.
[6] M. Weck, Handbook of Machine Tools (in translated form the original German by H. Bibring). New York: Wiley, 1984, pp. 15–20. [7] L. T. Nguyen, “Wire bond behavior during molding operations of
elec-tronic packages,” Polym. Eng. Sci., vol. 28, pp. 926–943, 1988. [8] E. Suhir and L. T. Manzione, “Predicted stresses in wire bonds of plastic
packages during transfer molding,” J. Electron. Packag., ASME Trans., vol. 113, pp. 16–20, 1991.
[9] S. Han and K. K. Wang, “A study on wire sweep in encapsulation of semiconductor chips using simulated experiments,” J. Electron. Packag.,
ASME Trans., vol. 117, pp. 994–1004, 1995.
[10] J. H. Wu, A. A. O. Tay, K. S. Yeo, and T. B. Lim, “A three-dimensional modeling of wire sweep incorporating resin cure,” IEEE Trans. Comp.,
Packag., Manufact. Technol. B, vol. 21, pp. 65–72, Feb. 1998.
[11] F. J. Lim and L. T. Nguyen, “Wire sweep during molding of integrated circuits,” in Proc. IEEE Electron., Comp. Technol. Conf., 1990, pp. 777–785.
[12] H. Yokoi, “Progress in visualization technologies for analyzing injection molding phenomena,” in Proc. 16th Polym. Proc. Soc. Conf., 2000, pp. 136–137.
[13] S. Y. Yang, S. J. Liou, and W. N. Liou, “Flow visualization of the gas-assisted injection molding process,” Adv. Polym. Technol., vol. 16, pp. 175–183, 1997.
[14] S. Han and K. K. Wang, “Analysis of the flow of encapsulant during un-derfill encapsulation of flip-chips,” IEEE Trans. Comp., Packag.,
Man-ufact. Technol. B, vol. 20, pp. 724–433, Nov. 1997.
Sen-Yeu Yang received the B.S. degree from National Taiwan University (NTU), Taipei, Taiwan, R.O.C., in 1976, the M.S. degree from Syracuse University, Syracuse, NY, in 1982, and the Ph.D. de-gree in mechanical engineering from the University of Minnesota, Minneapolis, in 1990.
He had several years of industrial experience and is currently a Professor in the Department of Mechan-ical Engineering and is in charge of Grace Laboratory for Polymer Processing, NTU, Taipei. His research interests span over a wide area of resin processing technology, including transfer compression molding, molding for EMI preven-tion, and flow-visualization of cavity filling process. He is the author of more than 30 international journal papers and several patents.
Dr. Yang is a member of SPE, PPS, and CSME, and served as the co-organizer of flow-visualization section in the PPS-13, the 13th meeting of International Polymer Processing Society.
Shin-Chang Jiang received the B.S. degree in mechanical engineering from the Taipei Institute of Technology, Taipei, Taiwan, R.O.C., in 1997 and the M.S. degree in mechanical engineering from National Taiwan University, Taipei, in 1999.
He is currently a R&D engineer at Branchy Vacuum Technology Co. Ltd., Lung-Tang, Taiwan.
Wen-Shu Lu received the B.S. and M.S. degrees in mechanical engineering from National Taiwan Uni-versity, Taipei, Taiwan, R.O.C., in 1995 and 1998, re-spectively.
He is currently a Software Engineer in Acer Com-munication and Multimedia Inc., Taipei.
