USF Libraries
USF Digital Collections

An investigation of BGA electronic packaging using Moiré interferometry

MISSING IMAGE

Material Information

Title:
An investigation of BGA electronic packaging using Moiré interferometry
Physical Description:
Book
Language:
English
Creator:
Rivers, Norman
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Ball grid array technology   ( lcsh )
Microelectronic packaging -- Reliability   ( lcsh )
Interferometry   ( lcsh )
Moiré method   ( lcsh )
electronic package
bga
moire interferometry
Dissertations, Academic -- Mechanical Engineering -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: As technology progresses towards smaller electronic packages, thermo-mechanical considerations pose a challenge to package designers. One area of difficulty is the ability to predict the fatigue life of the solder connections. To do this one must be able to accurately model the thermo-mechanical performance of the electronic package. As the solder ball size decreases, it becomes difficult to determine the performance of the package with traditional methods such as the use of strain gages. This is due to the fact that strain gages become limited in size and resolution and lack the ability to measure discreet strain fields as the solder ball size decreases. A solution to the limitations exhibited in strain gages is the use of Moiré interferometry. Moiré interferometry utilizes optical interferometry to measure small, in-plane relative displacements and strains with high sensitivity. Moiré interferometry is a full field technique over the application area, whereas a strain gage gives an average strain for the area encompassed by the gage. This ability to measure full field strains is useful in the analysis of electronic package interconnections; especially when used to measure strains in the solder ball corners, where failure is known to originate. While the improved resolution of the data yielded by the method of Moiré interferometry results in the ability to develop more accurate models, that is not to say the process is simple and without difficulties of it's own. Moiré interferometry is inherently susceptible to error due to experimental and environmental effects; therefore, it is vital to generate a reliable experimental procedure that provides repeatable results. This was achieved in this study by emulating and modifying established procedures to meet our specific application. The developed procedure includes the preparation of the specimen, the replication and transfer of the grids, the use of the PEMI, interpretation of results, and validation of data by finite element analysis using ANSYS software. The data obtained maintained uniformity to the extent required by the scope of this study, and potential sources of error have been identified and should be the subject of further research.
Thesis:
Thesis (M.S.M.E.)--University of South Florida, 2003.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Norman Rivers.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 87 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 52443500
notis - AJL4021
usfldc doi - E14-SFE0000078
usfldc handle - e14.78
System ID:
SFS0024774:00001


This item is only available as the following downloads:


Full Text

PAGE 1

AN INVESTIGATION OF BGA ELECTRONIC PACKAGING USING MOIR INTERFEROMETRY by NORMAN RIVERS A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering Department of Mechanical Engineering College of Engineering University of South Florida Major Professor: Thomas Eason, Ph.D. Daniel Hess, Ph.D. Muhammad Rahman Ph. D. Date of Approval: March 21, 2003 Keywords: electronic package, bga, moir interferometry Copyright 2003 Norman Rivers

PAGE 2

TABLE OF CONTENTS LIST OF TABLES ..............................................................................................................ii LIST OF FIGURES ...........................................................................................................iii ABSTRACT .......................................................................................................................vi 1. INTRODUCTION .......................................................................................................1 1.1. Overview ..............................................................................................................1 1.2. Literature Review .................................................................................................4 2. MOIR INTERFEROMETRY ....................................................................................8 2.1. Introduction ..........................................................................................................8 2.2. Theory ..................................................................................................................9 2.3. Fundamental Equations ......................................................................................16 2.4. Post Processing ..................................................................................................20 2.5. Experimental Setup ............................................................................................22 2.6. Grid Replication and Transfer ...........................................................................25 2.7. Procedure for Data Acquisition .........................................................................33 3. EXPERIMENTAL RESULTS ...................................................................................42 4. MODELING ..............................................................................................................60 4.1. Modeling and Assumptions ...............................................................................60 4.2. Material Properties .............................................................................................61 4.3. Results of Modeling ...........................................................................................64 5. DISCUSSION AND CONCLUSION ........................................................................71 REFERENCES ..........................................................................................................76 i

PAGE 3

LIST OF TABLES Table 1: U and V normal strains for the chip, mold and PCB of the FlexBGA. ...............45 Table 2: FlexBGA measured strains. .................................................................................49 Table 3: U and V normal strains for the chip, mold and PCB of the PBGA. ....................51 Table 4: PBGA measured strains. ......................................................................................55 Table 5: U and V normal strains for the chip, mold and PCB of the FlexBGA ................56 Table 6: TapeBGA measured strains. ................................................................................59 Table 7: Standard component list of BGA configurations. ................................................62 Table 8: Material properties of BGA components .............................................................63 Table 9: Modulus of elasticity of 63Sn/37Pb solder ..........................................................63 Table 10: U normal strain comparison between calculated and experimental results for the FlexBGA. ............................................................................................69 Table 11: V normal strain comparison between calculated and experimental results for the FlexBGA. ............................................................................................69 Table 12: U normal strain comparison between calculated and experimental results for the PBGA. .................................................................................................69 Table 13: V normal strain comparison between calculated and experimental results for the PBGA. .................................................................................................70 Table 14: U normal strain comparison between calculated and experimental results for the TapeBGA. ...........................................................................................70 Table 15: V normal strain comparison between calculated and experimental results for the TapeBGA. ...........................................................................................70 ii

PAGE 4

LIST OF FIGURES Figure 1: Schematic of Portable Engineering Moir Interferometer. ...................10 Figure 2: Isolated coherent beams striking specimen grating. ..............................11 Figure 3: Undeformed V-field grating. .................................................................18 Figure 4: V-field reference grating with 10% elongation in specimen grating. ...18 Figure 5: V-field reference grating with 10% elongation and 5% rotation. .........19 Figure 6: U-field reference grating with 10% elongation and 5% rotation ..........19 Figure 7: Front side view of PEMI setup. .............................................................23 Figure 8: Back side view of PEMI setup. .............................................................23 Figure 9: Separation apparatus. .............................................................................26 Figure 10: Alignment apparatus. ...........................................................................29 Figure 11: Specimen placed on a grid positioned on the heat sink. ......................32 Figure 12: Front side view of the PEMI including labels. ....................................35 Figure 13: Back side view of the PEMI including labels. ....................................36 Figure 14: Example of Image J measurement and scale lines on image. .............40 Figure 15: Example of Image J gray scale plot. ....................................................40 Figure 16: Full figure view of the U field fringes for FlexBGA specimen one. ...42 Figure 17: Full figure view of the U field fringes for FlexBGA specimen two. ..42 Figure 18: FlexBGA u fringes on the right three solder balls of specimen one. ..43 Figure 19: FlexBGA u fringes on the right three solder balls of specimen two. ..43 iii

PAGE 5

Figure 20: FlexBGA before grid transfer with description of components and numbering convention of solder balls. ..........................................................44 Figure 21: FlexBGA U fringes on profile after grid transfer. ...............................44 Figure 22: FlexBGA V fringes on profile after grid transfer. ...............................44 Figure 23: FlexBGA U fringes on solder balls 1, 2, 3 and 4. ...............................45 Figure 24: FlexBGA V fringes on solder balls 1, 2, 3 and 4. ...............................46 Figure 25: FlexBGA U fringes on solder balls 5, 6 and 7. ...................................46 Figure 26: FlexBGA V fringes on solder balls 5, 6 and 7. ...................................47 Figure 27: FlexBGA U fringes on solder balls 8, 9 and 10. .................................47 Figure 28: FlexBGA V fringes on solder balls 8, 9 and 10. .................................48 Figure 29: FlexBGA U fringes on solder balls 11, 12 and 13. .............................48 Figure 30: FlexBGA V fringes on solder balls 11, 12 and 13. .............................49 Figure 31: PBGA before grid transfer with description of components and numbering convention of solder balls. ..........................................................50 Figure 32: PBGA U fringes on profile after grid transfer. ....................................50 Figure 33: PBGA V fringes on profile after grid transfer. ....................................51 Figure 34: PBGA U fringes on solder balls 1, 2 and 3. ........................................52 Figure 35: PBGA V fringes on solder balls 1, 2 and 3. ........................................52 Figure 36: PBGA U fringes on solder balls 4, 5 and 6. ........................................53 Figure 37: PBGA V fringes on solder balls 4, 5 and 6. ........................................53 Figure 38: PBGA U fringes on solder balls 6 and 7. ............................................54 Figure 39: PBGA V fringes on solder balls 6 and 7. ............................................54 Figure 40: TapeBGA before grid transfer with description of components and numbering convention of solder balls. ..........................................................55 Figure 41: TapeBGA U fringes on profile after grid transfer. ..............................56 iv

PAGE 6

Figure 42: TapeBGA V fringes on profile after grid transfer. ..............................56 Figure 43: TapeBGA U fringes on solder balls 1, 2, 3 and 4. ..............................57 Figure 44: TapeBGA V fringes on solder balls 1, 2, 3 and 4. ..............................57 Figure 45: TapeBGA U fringes on solder balls 4, 5 and 6. ..................................58 Figure 46: TapeBGA V fringes on solder balls 4, 5 and 6. ..................................58 Figure 47: Schematic of BGA configurations. .....................................................62 Figure 48: U field normal strains for FlexBGA as calculated with ANSYS and compared to experimental values of individual components. .......................65 Figure 49: V field normal strains for FlexBGA as calculated with ANSYS and compared to experimental values of individual components. .......................65 Figure 50: U field normal strains for PBGA as calculated with ANSYS and compared to experimental values of individual components. .......................66 Figure 51: V field normal strains for PBGA as calculated with ANSYS and compared to experimental values of individual components. .......................67 Figure 52: U field normal strains for TapeBGA as calculated with ANSYS and compared to experimental values of individual components. .......................68 Figure 53: V field normal strains for TapeBGA as calculated with ANSYS and compared to experimental values of individual components. .......................68 v

PAGE 7

AN INVESTIGATION OF BGA ELECTRONIC PACKAGING USING MOIR INTERFEROMETRY Norman Rivers ABSTRACT As technology progresses towards smaller electronic packages, thermo-mechanical considerations pose a challenge to package designers. One area of difficulty is the ability to predict the fatigue life of the solder connections. To do this one must be able to accurately model the thermo-mechanical performance of the electronic package. As the solder ball size decreases, it becomes difficult to determine the performance of the package with traditional methods such as the use of strain gages. This is due to the fact that strain gages become limited in size and resolution and lack the ability to measure discreet strain fields as the solder ball size decreases. A solution to the limitations exhibited in strain gages is the use of Moir interferometry. Moir interferometry utilizes optical interferometry to measure small, in-plane relative displacements and strains with high sensitivity. Moir interferometry is a full field technique over the application area, whereas a strain gage gives an average strain for the area encompassed by the gage. This ability to measure full field strains is useful in the analysis of electronic package vi

PAGE 8

interconnections; especially when used to measure strains in the solder ball corners, where failure is known to originate. While the improved resolution of the data yielded by the method of Moir interferometry results in the ability to develop more accurate models, that is not to say the process is simple and without difficulties of its own. Moir interferometry is inherently susceptible to error due to experimental and environmental effects; therefore, it is vital to generate a reliable experimental procedure that provides repeatable results. This was achieved in this study by emulating and modifying established procedures to meet our specific application. The developed procedure includes the preparation of the specimen, the replication and transfer of the grids, the use of the PEMI, interpretation of results, and validation of data by finite element analysis using ANSYS software. The data obtained maintained uniformity to the extent required by the scope of this study, and potential sources of error have been identified and should be the subject of further research. vii

PAGE 9

1. INTRODUCTION 1.1. Overview The general trend of electronic packages is progressing towards the decrease in the pitch size of solder balls. The pitch size is the center-to-center distance of the solder balls. As this distance is decreased, the size of the solder balls becomes smaller and the ball count of the connection increases, thereby giving a greater interconnection density. The increase in interconnection density also allows for a higher input/output density, since the solder balls act as electrical as well as mechanical connections. The higher input/output density facilitates the use of chips that exhibit a decreased die size and require a much smaller footprint during manufacture. This allows for a more efficient use of space on the printed circuit board (PCB); and therefore reduces manufacturing cost. The result is a smaller, more efficient, more cost effective electronic package. 1 A commonly used surface mount technology that facilitates decreased pitch size is the Ball Grid Array (BGA). The BGA consists of an evenly spaced solder ball pattern that is used to connect the chip to the printed circuit board. This pattern further facilitates ease of manufacture due to the fact that it is self-centering during the wave soldering process. While the decreased pitch size of the BGA results in the ability to 1

PAGE 10

produce smaller, more efficient electronic packages, it also introduces sources of failure inherent to smaller electronic packages. A major mode of failure in electronic packaging is cyclic fatigue in the solder balls that attach the chip to the printed circuit board. This failure is due to thermo-mechanical stresses caused by the mismatched coefficients of thermal expansion in the printed circuit boards and the chips. As the printed circuit board expands at a greater rate than the silicon chip, the solder balls that connect the two are placed in a state of shear stress. As the electronic package undergoes thermal cycling due to normal operating conditions, the solder balls exhibit fatigue effects due to these stresses; leading to crack propagation in the corners of the solder balls and ultimately failure of the solder connection. In surface mount applications, the solder connection mechanically and electrically connects the chip to the printed circuit board. The failure of the solder connection therefore implies a failure of the entire package. 2 A commonly used method of countering thermo-mechanical effects involves the use of heat sinks. Heat sinks typically attach directly to the die, and are used to dissipate the heat generated during normal operation of the electronic package. The smaller die size of packages utilizing a decreased pitch size, such as the BGA, makes the attachment of a heat sink difficult or impossible. In addition, the greater input/output density possessed by these packages results in the generation of more heat during normal operating conditions, thereby increasing the effect of thermo-mechanical considerations, and placing a greater emphasis on the ability to predict the fatigue life of the package. 2

PAGE 11

To predict the fatigue life of the solder connections, one must be able to accurately model the thermo-mechanical performance of the electronic package. 3-5 As the solder ball size decreases, it becomes difficult to determine the performance of the package with traditional methods such as the use of strain gages. This is due to the fact that strain gages become limited in size and resolution and lack the ability to measure discreet strain fields as the solder ball size decreases. 6 A solution to the limitations exhibited in strain gages is the use of Moir interferometry. Moir interferometry utilizes optical interferometry to measure small, in-plane relative displacements and strains with high sensitivity. Moir interferometry also has the unique ability to yield a discreet strain field over the area of application, whereas a strain gage gives an average strain for the area encompassed by the gage. 7 This ability to measure discreet strains is especially useful when measuring strains in the solder ball corners, where failure is known to originate. While the improved resolution of the experimental data yielded by the method of Moir interferometry results in the ability to develop more accurate fatigue models, that is not to say the process is simple and without difficulties of its own. Moir interferometry is inherently susceptible to error due to experimental and environmental effects. For instance, mistakes in experimental methodology, such as the introduction of rigid body rotation, can result in misleading data. Environmental effects, such as vibration or thermal convection, can result in no data at all. Therefore, it is vital to generate a reliable experimental procedure that yields repeatable results. The objective of this work was to formulate and document a procedure to measure the thermo-mechanical strains in BGA electronic packaging using Moir 3

PAGE 12

interferometry. This was achieved in this study by modifying established procedures to meet our specific application, namely the investigation of thermo-mechanical strains in BGA electronic packaging. This procedure includes the preparation of the specimen, the replication and transfer of the grids, the techniques of Moir interferometry, interpretation of results, and validation of data by finite element analysis using ANSYS software. 1.2. Literature Review Electronic packaging is typically described by a hierarchy consisting of four main levels of packaging. The terminology used in this study is consistent with that of Pecht. 8 The zero level is used to describe semi-conductor chips as well as discreet passive devices such as resistors and capacitors. When the chips are packaged in a protective chip carrier, this is referred to as first level packaging. The connection of the chip carrier to a mounting surface is described as the second level of packaging. The third level of packaging refers to the interconnection of circuit boards and power supplies to a physical interface. Often the circuit boards are bundled together in a protective structure called a cabinet. When several cabinets are joined together, it is referred to as fourth level packaging. There are two major categories in which connections are made in first level packaging: pinned components and surface mount components. Pinned components utilize wire leads that are inserted through the mounting surface. Advantages of this technology include the capability of multiple insertion and withdrawal cycles as well as good resistance to thermo mechanical stresses. Surface mount components are 4

PAGE 13

connected directly to the mounting surface and are becoming more popular due to the increases input/output to package area ratio when compared to pinned components. 9 The smaller size of surface mount components allows for more efficient board utilization, thereby decreasing overall size and increasing cost effectiveness. There are many types of carriers used in surface mounting, but they fall into two main categories: leadless and leaded carriers. Leadless carriers consist of metal pads, both on the carrier and board, which are joined by a soldered connection. Metal leads extend from the leaded carriers to solder pads on the board. Both types of carriers rely on the surface solder joints to make both the mechanical and electrical connections to printed circuit boards. The major difference between the carriers is that the leaded carriers are more resistant to the thermo mechanical stresses, while the leadless carriers are generally much smaller in size. 10 Leadless carriers are also categorized by the type of second level connection which is used to attach them to the circuit boards. Three major types of second level connections commonly used on leadless carriers are: direct chip attach (DCA), chip scale packaging (CSP), and ball grid array (BGA). DCA technology attaches the chips directly to the circuit board. The two major categories of DCA are wire bond and flip chip. The main difference between the two techniques is that wire bond DCA uses organic based adhesives to attach the chip, whereas the flip chip connection is accomplished using solder bumps that are formed on the chip before it is attached to the circuit board. DCA technology facilitates the reduction of die size, giving it increased cost and space efficiency; but the small 5

PAGE 14

footprint lends to increased failure rates due to thermo-mechanical stresses, and water absorption is a problem due to the lack of any chip encapsulation. The configuration of a CSP consists of an interposer sandwiched between the silicone die and the solder ball. The interposer is usually no more than 20 percent over the size of the chip itself. The slightly larger footprint due to the interposer allows for a slight increase in resistance to thermo-mechanical failure, but water absorption is still a problem due to the lack of any encapsulation. The carrier of a BGA consists of one or more chips laminated to an epoxy substrate using an encapsulant. It is attached to the circuit board via rows of solder balls, which allows for increased interconnection density, as well as the ability to self-center during attachment. There are several types of BGA commonly used, including thin cavity ball grid array (TBGA), plastic ball grid array (PBGA), thick laminate ball grid array (LBGA), ceramic ball grid array (CBGA), flex ball grid array (FlexBGA), and tape ball grid array (TapeBGA). TBGA is a cavity down array, meaning that the chip is sandwiched between the substrate and the circuit board. This configuration allows the attachment of a heat spreader directly to the substrate, but considerably reduces the interconnection density due to the fact that no solder balls can be placed directly under the chip. PBGA is a cavity up array, which means that the chip is on top of the substrate. This allows for a full solder field at the connection, thereby giving it the greatest interconnection density. LBGA and CBGA are similar to PBGA except that LBGA has a thicker substrate and CBGA has a ceramic substrate, which increases the strength as well as cost of both configurations due to material considerations. FlexBGA and TapeBGA 6

PAGE 15

are also similar to PBGA except they have a thinner, more flexible substrate, which is useful in smaller applications such as laptop computers and cellular phones. PBGA, FlexBGA, and TapeBGA configurations will be the focus of this study. 7

PAGE 16

2. MOIR INTERFEROMETRY 2.1. Introduction As technology progresses towards smaller electronic packages, thermo-mechanical considerations pose a challenge to package designers. One area of difficulty is the ability to predict the fatigue life of the solder connections. To do this one must be able to accurately measure the strain the solder balls experience during thermal cycling. As the solder ball size decreases, it becomes difficult to accurately measure these strains with traditional methods such as the use of strain gages. This is due to the fact that strain gages become limited in size and resolution and lack the ability to measure discreet strain fields as the solder ball size decreases 6 A solution to the limitations exhibited in strain gages is the use of Moir interferometry. Moir interferometry utilizes a combination of geometric Moir and optical interferometry to measure small, in-plane relative displacements and strains with high sensitivity. Note that displacement in this study refers to elongation of the specimen not translation. Moir interferometry also has the unique ability to yield a discreet strain field over the area of application, whereas a strain gage gives an average strain for the area encompassed by the gage. This ability to measure discreet strains is useful in the analysis of electronic package interconnections; especially when 8

PAGE 17

used to measure strains in the solder ball corners, where crack propagation is known to originate. 2.2. Theory Moir interferometry is a combination of the concepts and techniques of geometrical Moir and optical interferometry and is capable of measuring small, in-plane relative displacements with sensitivity. Moir fringes are produced by the interference of two grid patterns of the same frequency when one of the grids undergoes elongation. The resolution of the relative displacement that can be measured is directly proportional to the frequency of the grating used. In Moir interferometry, one grid is formed from a cross-line diffraction grating produced directly on the surface of the specimen. This grid is compared to a virtual reference grating formed from two interfering coherent beams of light reflected from an internal reference grating as shown by figure 1. A null fringe pattern is initially formed if the virtual grating is a multiple of the specimen grating. When the specimen undergoes elongation, thereby slightly changing the frequency of the grating attached to it, a fringe pattern is formed. The elongation of the specimen can then be calculated from simple governing equations involving the frequency of the reference grating and number of fringes recorded. These equations, along with an overview of the optical theory involved, are discussed below and follow closely to Post et al 11 9

PAGE 18

From sourceTo cameraU beam V beam Internal reference g ratin g S p ecimen U2 V2V1U1 Figure 1: Schematic of Portable Engineering Moir Interferometer. The interferometer used in this study is the Portable Engineering Moir Interferometer (PEMI), which is the variation of the four-beam interferometer illustrated in Figure 1. A single laser beam enters the PEMI and is redirected to strike a reference cross-line diffraction grating at normal incidence. This internal reference grating has a frequency equal to 1200 lines/mm. The 1 diffraction orders in both the U and V directions are reflected by mirrors U1, U2 and V1, V2 respectively to form a virtual reference grating on the specimen. The frequency of this virtual grating is twice that of the internal grating, or 2400 lines/mm. The specimen gratings used in this study have a frequency of 1200 lines/mm. Since the frequency of the virtual grating f is twice that of the specimen grating fs, an initial null pattern can be achieved with this setup. Shutters in front of the specimen allow the U and V (horizontal and vertical) beams to be isolated to form their respective reference gratings on the specimen 10

PAGE 19

separately. Light diffracted by the specimen grating is collected by the camera lens, which focuses the moir patterns onto the film plane of the camera as shown in Figure 2. The remainder of this section is devoted to an overview of the optical theory behind Moir interferometry and this figure will be referenced throughout the discussion. Beam 1 f = 2fs x -1 0 1 2 3 -3 -2 -1 0 1 Beam 2 Beam 1Beam 2 k S Camera m y z Figure 2: Isolated coherent beams striking specimen grating. The ray diagrams at the top and bottom of the figure illustrate the diffraction order m, and angle of diffraction m. These follow the commonly used sign convention where the normal reflection path is referred to as the 0 order, and diffraction order increases in the counter-clockwise direction. The angle of diffraction 11

PAGE 20

is measured relative to the horizontal with positive angles in the direction of increasing diffraction order. The center portion of the figure consists of a detailed description of beams 1 and 2, which result from the split beams diffraction from the internal grating in either the U or V direction. These beams approach the specimen grating at the symmetrical angles of incidence. The relationship between these angles and the reference grating frequency, f, is described by the relationships sin2 f sin2 f For beams 1 and 2 respectively and where is the wavelength of the beams. The beams then strike the specimen grating and are diffracted to form two mutually coherent beams, having wave fronts 1 and 2 which coexist in space and generate optical interference. The behavior of these beams as they leave the specimen is described by smfm sinsin smfm sinsin For beams 1 and 2 respectively, where m is the diffraction order, fs is the specimen grating frequency, m is the angle to the mth diffraction order. When f s = f /2, and the lines of the reference grating and the specimen are parallel, light diffracted from the specimen grating in the orders of m = 1 and m = -1 emerge at angles 1 = -1 = 0. This can be shown for beam 1 as follows: 12

PAGE 21

smfm )sin(sin With m = 1 and substitution from the above equation, sff 2sin1 sff2sin1 Substitution of the relationship f s = f /2 yields: 0sin1 ssff Similarly, for beam 2: smfm )sin(sin Let m = -1 and substitute the corresponding equation from above, sff 2sin1 sff2sin1 Substitute the relationship f s = f /2, 0sin1 ssff This indicates that wave fronts 1 and 2 emerge with a path perpendicular to the grating and having an angle of intersection, equal to 0. This dictates that no interference between the beams exists; therefore, the result is a null field. If the specimen undergoes uniform elongation in either the x or y direction such that the uniform normal strain is constant, the frequency of the specimen grating is decreased to: 13

PAGE 22

12/ffs Therefore, substitution into the above equations for fringe order m = 1 yields: sff2sin1 12/2sin1ff 12sin1f Similarly, substitution into the above equations for fringe order m = -1 yields: sff2sin1 12/2sin1ff 12sin1f For very small strains the following assumption can be made: 1 Therefore, the following relations hold true: 2sin1 f 2sin1 f 14

PAGE 23

It is known from basic optical theory of interference 12 that when any two coherent beams of light intersect at an angle 2; evenly spaced, parallel bands of constructive interference will form on a plane perpendicular to the bisector of the beams. Furthermore, the frequency with which these bands occur (also known as the fringe gradient F) is defined by the relationship: sin2F Where is the wavelength of the beams. Because and are symmetric, the angle of separation in this case is: 11122 Therefore, sinsin1 Substituting these relationships into the equations defining the fringe gradient yields: 12F Notice that if for the case of the null fringe is substituted into this relationship, the following holds true: 021sff 002F For the case where the specimen has undergone elongation, 15

PAGE 24

221 fffs ffF22 This shows that the fringe gradient is proportional to the frequency of the reference grating and the strain exhibited in the specimen. It is therefore possible to develop equations to calculate this strain if the fringe gradient and frequency of the reference grating are known. These equations are presented in the next section. 2.3. Fundamental Equations Another way of representing the fringe gradient F is dNi/dj. Here Ni refers to the fringe count resulting from the interference of beams traveling parallel to the x or y axis (in the horizontal plane U or vertical plane V), while j refers to a distance along the x or y axis. An example would be dNx/dy, which would read: the derivative of the fringe count along the vertical with respect to the distance associated with that count resulting from the interference of horizontal beams. Using this nomenclature the following relationships for linear normal strain can be formed 13 : dxdNfdxdUxx1 dydNfdydVyy1 From these relationships, the following equations can be formed for relative displacement: 16

PAGE 25

),(1),(yxNfyxUx ),(1),(yxNfyxVy Linear shear strain is calculated as follows: dxdNdydNfdxdVdydUyxxy1 The in-plane rigid body rotation of the specimen can also be calculated from: dxdNdydNfdxdVdydUyxxy1 It is important to note that the rigid body rotation of the specimen produces mutually dependant cross-derivatives, namely: dxdNdydNyx Four beam Moir interferometers, such as the PEMI used in this study, take advantage of this relationship. Since the x and y virtual reference gratings are fixed in space, the magnitude of the specimen rigid body rotation, relative to each of them, is identical. Although accidental rigid body rotation of the specimen alters the fringe patterns, the rotation has no effect on the calculated shear strains. Accordingly, none of the strains (x, y, xy) measured by the PEMI are affected adversely by rigid body rotations. 11 Geometric Moir can be used to demonstrate this. Figure 3 represents two sets of evenly spaced lines superimposed one on the other. The grid lines are perpendicular to the field under consideration, in this example the V field. Notice there are no fringes 17

PAGE 26

present, as is expected in the case of no elongation. If the spacing of one set of lines undergoes a 10% elongation, then Moir fringes can be observed. This is shown in Figure 4. Figure 3: Undeformed V-field grating. Figure 4: V-field reference grating with 10% elongation in specimen grating. This imitates elongation of the specimen in the vertical direction. Notice the presence of five fringes oriented perpendicular to the y-axis. This indicates normal strain and 18

PAGE 27

could be measured using the relationships previously developed. Figure 5 demonstrates the effects of adding rotation as well as elongation. Figure 5: V-field reference grating with 10% elongation and 5% rotation. Notice that there are still five fringes crossing the y-axis, indicating that the normal strain is unchanged. It would appear at first glance that shear strain is also present, but to measure shear strain, the U field is needed as well. Figure 6 is a representation of the same rotation and elongation as seen from the U field. Figure 6: U-field reference grating with 10% elongation and 5% rotation 19

PAGE 28

Notice that the slopes of the fringes are equal and opposite in the U and V fields as predicted in the previous equation. When these slopes are added together the value of the strain is zero. The amount of the rotation could also be determined by subtracting the two. 2.4. Post Processing A major challenge in using Moir interferometry is the post processing of the data once it is collected. Strain measurements in particular are difficult because they depend on an approximation of the fringe gradient dNi/dj. There exist several methods of estimating the fringe gradient including mechanical differentiation, the fringe vector method, and the finite increment method. 11 Mechanical differentiation, also known as the fringe shifting method, is a whole field method; meaning that it measures averaged strains across a section of fringes. It does this by superimposing 2 identical Moir patterns. If the patterns are aligned, then the original field is obtained. If one of the patterns is shifted a finite distance in the x or y, a super Moir pattern is created. The super Moir pattern is essentially a contour map of the derivative field that can be evaluated to yield gradient values. 14-17 The fringe vector method is a discreet method of approximation, meaning that the localized strains can be measured at discreet points. The fringe vector is a vector whose direction is normal to the tangent of the fringe. The fringe angle is the angle this vector makes with the x and y-axis. Relationships can then be developed between the gradient components for both the Nx and the Ny fields respectively: 20

PAGE 29

tandxdNdydNxx tandxdNdydNyy These relationships become useful when dx or dy is difficult to measure for either gradient field. Now instead of measuring two fringe gradients, one fringe gradient and the fringe angle can be measured to yield the same result. The method used in this study to approximate the fringe gradient is the finite increment method. When the fringe patterns are fairly uniform and spaced closely together, the fringe gradient can be approximated by: jNdjdNii Often the image of the fringe patterns is magnified either optically or with the use of computer software to facilitate the measurement of j. In this case a magnification factor M is introduced in the equation. A value for M is obtained by measuring a dimension in the magnified image whose actual value can be obtained. The magnified value divided by the actual value is the magnification factor. The fringe gradient therefore becomes: jNMdjdNii Where j is obtained from the magnified image. It should also be noted that sometimes when the spacing of fringe patterns are uniform, but not as closely spaced as desired; fractional fringe patterns are introduced by creating artificial lines that bisect the fringes while maintaining the contour of the fringe pattern. In this case N i 21

PAGE 30

will have a fractional value. Conversely, integer values of N i would result in averaged values over the area covered. In this study, the magnified distance was measured between individual fringe patterns to yield averaged strain values over small areas. Therefore Ni = 1 and a magnification factor was used. The resultant equations for normal strains as measured in this study are: xfMNxx1 yfMNyy1 Where x is the distance between fringes along the x-axis resulting from the horizontal beams, and y is distance between fringes along the y-axis as a result of the interference between the vertical beams. 2.5. Experimental Setup The experimental setup for this study was developed to better not only the quality, but also the repeatability of the data acquired. In order to accomplish this, several factors had to be considered. These included vibrations, control, mount rigidity, grid alignment, and grid replication. The PEMI setup was mounted on an optics table to help account for these factors. This optics table had a large mass top, which was mounted on legs containing air bladders to dampen vibration. The tabletop was a 1 breadboard of x 20 mounts, which enabled the PEMI and all pertinent equipment to be securely mounted to it. Figure 7 and Figure 8 are representations of 22

PAGE 31

the table as it was used in this study, and should be referred to throughout this discussion of the experimental setup. The remaining factors are addressed as the procedure is explained. Specimen m ount Y Z X Figure 7: Front side view of PEMI setup. Camera m ount Figure 8: Back side view of PEMI setup. A specimen mount was positioned behind the PEMI. The specimen mount consisted of a thin aluminum L shaped specimen holder attached to an optics mount, which was designed to have fine adjustment capabilities for translation and rotation in all three directions. This full range of adjustment is vital due to the sensitivity of the 23

PAGE 32

PEMI. This setup allowed the specimen and the transfer grating to be attached to the holder simultaneously. Once the transfer grating had been used to center the PEMI, only slight adjustment of the specimen mount was needed to bring the specimen into the focus plane. This process will be discussed in great detail in the experimental procedure portion. The camera mount was placed on the opposite side of the PEMI from the specimen mount. It also consisted of a modified optics mount, with rotational and translation capabilities in the three directions. The camera used was a FujiFilm FinePix S2 Pro Digital camera. The images were obtained remotely and downloaded directly to a computer, using Fujifilm shooting software. In this manner, fine adjustments could be made to the shoot settings (shutter speed, contrast, etc.) of the camera; and vibrations due to manual triggering were avoided. No lenses were used on the camera itself; rather, the mount was used in conjunction with a combination of lenses on the PEMI to focus the image directly on the CCD of the camera. Fine tune adjustment capabilities were of importance when obtaining full view images of the electronic packages. The combination of lenses necessary for this view mandated that the focal length from the PEMI to the camera be relatively short; and therefore, the achievement of a precise focus plane was necessary if quality images were to be obtained. Attaining a precise focus plane was not of as much importance when obtaining the magnified images of the individual solder balls. The combination of lenses used on the PEMI mandated that the focal distance be much greater, and therefore not as precise. The use of a tripod camera mount was found to be sufficient. 24

PAGE 33

Other components of the experimental setup included a separation apparatus, an alignment apparatus, and a variable range oven. These were all vital to the processes of grid replication and transference of the grids onto the specimen. These processes must be performed before the specimen can be used with the PEMI, and mistakes made during these processes would affect the quality of the data acquired. The next section discusses in detail these processes and the devices necessary to complete them. 2.6. Grid Replication and Transfer Grid replication and transfer involves several steps. Grid replication consists of the creation of a silicone negative from a master grating; creation of an epoxy submaster from the silicone negative; and finally, a reflective finish is formed on the epoxy submaster. Replication of grids is a process intended to make the experimental procedure more cost effective. Master gratings are expensive and are not reusable after transference has occurred, whereas one master grating can be replicated to make multiple negatives and submaster gratings at a vastly reduced cost. Transference is a four-part process involving alignment, adhesion, trimming and separation. It is important that these processes be performed correctly, or error will be introduced into the measurements. The refinement of these techniques was an important part of the formulation of a repeatable experimental method. The first step of replication is to produce silicone negatives from the master grating. This imprint will be formed in the silicone coating that will be placed on 3/7 25

PAGE 34

float glass that has been cut into 2 X 3 plates. A detailed description of the process is as follows: 1) Thoroughly clean and dry the glass plate to be used for the negative. 2) Apply SS4120 silicone primer to the plate using a lens cloth to assure a thin, even coating. Let dry in upright position for 30 minutes. 3) For each negative to be created, mix 10 grams of silicone and 1 gram of hardener in a 100 ml beaker. Working time is 24 hours. 4) Degas the mixture in a vacuum chamber, paying special attention to boil over during the first few minutes. The process takes approximately 1 hour. 5) Slowly pour silicone on the primed surface of the plate, leaving a trailing edge off one side, and ensuring there are no bubbles introduced into the silicone. 6) Slowly lower master grating onto silicone puddle, making sure no air bubbles are trapped between plates. 7) Align and tape the plates in position. Let cure for 3 5 days. 8) Trim excess silicone from the edge of the plates using a razor blade. 9) Place the plates in the separation apparatus with the master grating on top. The silicone negative is now ready to be separated using the separation apparatus shown in Figure 9. Figure 9: Separation apparatus. 26

PAGE 35

The separation apparatus is necessary to control the rate of separation between the master grating and the silicone negative. The process of separation involves securing the clamping bars on the edges of the bottom plate. Slowly turn the bottom thumbscrews until a separation crack propagates from the corner. Do not use excessive force to separate the plates, or tears might form in the silicone. The silicone negative is now ready to be used in epoxy submaster replication. An epoxy submaster is created by forming an imprint in the epoxy coating of a glass plate. This imprint is a negative of the silicone negative (i.e. an exact copy of the master grating). This copy will be made on a 2 X 3 plate. 3/7 float glass can be used for normal applications, but Ultra Low Expansion (ULE) glass must be used if a high temperature transfers is to be performed. A detailed description of the process is as follows: 1) Thoroughly clean and dry the glass plate to be used for the submaster. 2) For two submasters to be replicated, pour 15 ml of epoxy and 15 ml of hardener into separate100 ml beakers. 3) Degas separately in a vacuum chamber, paying special attention to boil over during the first few minutes. The process takes approximately 1 hour. 4) Mix epoxy and hardener thoroughly, then distribute evenly between the two beakers. Do not hold the beakers in hand while stirring, as body heat might accelerate the curing process. Working time is 1 hour. 5) Degas the mixture in a vacuum chamber, paying special attention to boil over during the first few minutes. The process takes approximately 20 minutes. 6) Slowly pour epoxy on surface of plate, leaving a trailing edge off one side, and ensuring there are no bubbles introduced into the epoxy. 7) Slowly lower the silicone negative onto the epoxy puddle, making sure no air bubbles are trapped between the plates. 27

PAGE 36

8) Align and tape the plates in position. Let cure for 1 day. 9) Trim excess epoxy from the edge of the plates using razor blade. Let cure for 2 days. 10) Place the plates in the separation apparatus, with the epoxy submaster on top. The epoxy submaster is now ready to be separated from the silicone negative using the same procedure as was described for the separation of the silicone negative from the master grating. The epoxy submaster is now ready to be given a fully reflective finish as described in the following process: 1) Coat the epoxy submaster with a thin layer of aluminum using vacuum deposition. 2) Mix distilled water with Photo-flo 200 solution, 200:1 by volume. 3) Apply the diluted Photo-flo solution to the aluminum surface of the plate using a lens cloth to assure a thin, even coating. Let dry in upright position for 30 minutes. 4) Apply a second layer of aluminum using vacuum deposition. Great care must be taken when depositing the two coats of aluminum onto the epoxy submaster. If the process is performed too fast, the epoxy surface will melt, destroying the grid pattern formed in it. The purpose of having two layers of aluminum is that once the photo-flo solution has been applied, an oxidation layer will form on the first coat of aluminum. When the second layer of aluminum is applied, it will not bond to the first layer. When the specimen is adhered to the second layer, it will separate from the first, thereby creating an exact reflective copy of the grating to be transferred onto the specimen. Once the grid has been coated with a reflective film, it must undergo an alignment procedure in order to assure that the grid lines are aligned with the 28

PAGE 37

specimen before the grid is transferred onto the specimen. The alignment procedure is performed with the assistance of the alignment apparatus shown in Figure 10. Figure 10: Alignment apparatus. The alignment procedure takes advantage of the fact that there are two perpendicular sets of lines etched onto the grids. The diffraction of a laser beam will be used to align one set to the perpendicular of the optics table; therefore, the other set will be parallel to the table. The procedure for alignment of the grid is as follows: 1) Mount the grid onto the apparatus using double-sided tape. 2) Direct a laser towards the reflective surface of the grid. 3) Adjust the vertical and horizontal alignment of the apparatus until the 0 order diffraction is directed back at the laser source. 4) The 1 order diffraction will be directed to either side of the laser source. Obtain vertical measurements of the 1 order beams at two positions equidistant from both the grid and the laser. 5) Rotational adjustment should be made to the grid until the two vertical measurements are equal. One set of grid lines is now oriented perpendicular to the optics table; therefore, the other set must be parallel to the table. 6) Using spacers of constant thickness, etch alignment lines onto the grid. 29

PAGE 38

The grids are now ready to be transferred onto the specimen. The alignment lines must be oriented parallel to the specimen during transfer. The transfer of the grating onto the specimen can be performed over a wide range of temperatures, and various adhesives exist that optimize adhesion at each temperature range. The following is a description of the method used to transfer the gratings replicated at room temperature using PC10-C unfilled adhesive. 1) Draw a little more than 3 ml of PC10-C adhesive into a syringe. Be careful not to introduce air bubbles into the adhesive. 2) Slowly dispense approximately 3 ml of adhesive down the side of a 6 ml test tube, ensuring that no bubbles are formed. 3) Draw approximately .5 ml of hardener into a syringe. Be careful not to introduce air bubbles into hardener. 4) Slowly dispense approximately .2 ml of adhesive down the side of a 6 ml test tube, ensuring no bubbles are formed. Exact ratio of adhesive to hardener is not as important as the absence of bubbles in the mixture. 5) Slowly lower a micro-spoon below surface of mixture. 6) Thoroughly mix the hardener and adhesive by rotating the micro-spoon, ensuring that no bubbles are introduced to the mixture. 7) If the above process was performed correctly, no air bubbles should be present in the mixture; therefore, there is no need to de-gas. Working time is 15 minutes. 8) Pour small puddle of adhesive onto the double-coated aluminum surface of the epoxy submaster. 9) Lower the specimen onto epoxy puddle, making sure no air bubbles are trapped between the specimen and the plate. 10) Place weights on the specimen to force out excess adhesive from between the specimen and the plate. 11) Use a cotton swab to clean the excess adhesive as it is forced out. Let cure for 3 hours. 12) Trim excess adhesive from the edge of the specimen using a razor blade. 30

PAGE 39

13) Separate the specimen from the epoxy submaster by hand. The two layers of aluminum should separate easily, leaving one layer of aluminum on the epoxy submaster, while the other layer is transferred to the specimen. The layer of aluminum that has been transferred to the specimen contains an exact replica of the grid on the master. The specimen is now ready to be used on the PEMI. Sometimes it is necessary to perform the grid transfer at temperatures other than ambient. One such case is when measuring the thermally induced strains in electronic packaging. If a room temperature transfer was performed, and an attempt was made to collect data at the elevated temperature; heat convecting from the specimen would distort the fringe patterns, making it difficult to obtain accurate readings with the PEMI. In addition, it would be difficult to keep the specimen at a constant elevated temperature. These issues can be avoided with the use of a high temperature transfer. If the grid is transferred on to the specimen at the elevated temperature and then allowed to cool to room temperature, the magnitude of the measured strain is the same as it would have been if the temperature had been raised from ambient to the elevated temperature. In addition, heat convection and temperature stability are no longer cause for concern. That is not to say that high temperature grid transfers do not possess inherent problems of their own. The specimen must be maintained at a fairly constant elevated temperature during alignment of the grid, curing of the adhesive, and post-cure trimming of the specimen. To aid in this, an oven containing a removable heat sink was designed, and an experimental procedure was developed. Figure 11 is a 31

PAGE 40

representation of the specimen placed on a grid positioned on the heat sink, as was used in the experimental procedure. Heat sink Transfer grating Specim e n Figure 11: Specimen placed on a grid positioned on the heat sink. The following is a description of the procedure for grid transference at 82 C: 1) Raise the oven temperature to 82 C. 2) Place the specimen and the ULE transfer grid in the oven for 30 minutes. 3) Thoroughly mix Tra-Bond BAF-230 high temperature adhesive, avoiding the introduction of air bubbles. 4) Remove the heat sink, specimen, and ULE transfer grid from oven. Note that the heat sink will keep the transfer grid and the specimen close to the elevated temperature during the next few steps. 5) Pour a small puddle of adhesive onto the double-coated aluminum surface of the epoxy submaster. 6) Lower the specimen onto the epoxy puddle, making sure no air bubbles are trapped between the specimen and the plate. 7) Use a cotton swab to clean the excess adhesive as it is forced out. 8) Align the specimen to alignment line. 9) Return the heat sink, specimen, and ULE transfer grid to the oven. 32

PAGE 41

10) Maintain a temperature of 82 C for 1 hour and 15 minutes. 11) Remove the heat sink, specimen, and ULE transfer grid from oven. 12) Trim excess adhesive from the edge of the specimen using a razor blade while specimen is still at elevated temperature. 13) Separate the specimen from the epoxy submaster by hand and allow it to cool to room temperature. 14) Trim excess adhesive from the edges of the chip, being careful not to separate the remaining grid from the face of the chip. The two layers of aluminum should separate easily, leaving one layer of aluminum on the epoxy submaster, while the other layer is transferred to the specimen. The layer of aluminum that has been transferred to the specimen contains an exact replica of the grid on the master. After the temperature of the specimen has cooled to room temperature the grid will have experienced the same magnitude strain as if it had been heated over the same temperature gradient. The specimen is now ready to be used on the PEMI. 2.7. Procedure for Data Acquisition Moir interferometry is inherently susceptible to error due to experimental and environmental effects. Therefore, it is vital to generate a reliable experimental procedure that yields repeatable results. This was achieved in this study by emulating and modifying established procedures11 to meet our specific application. The types of electronic packages used were a 10 x 10 mm TapeBGA having a 6.75 mm die, .8mm pitch, and 144 solder ball count; a 15 x 15 mm PBGA having a 6.35 mm die, 1mm pitch, and 196 solder ball count; and a 27 x 27 mm FlexBGA having a 6.35 mm die, 1 mm pitch, and 672 solder ball count. 33

PAGE 42

Each package was cut through the cross-section using a diamond blade saw. Care must be taken during this process, making sure not to clamp the package too tightly, cut to fast, or apply too much pressure; or else chip separation could occur. The cut should be made close to the midpoint and aligned with the solder row to be studied. The face of the cross-sectional cut was then polished using a rotary polisher. Polishing was performed using an iterative process that involved a different grain size in each step. First, 320 grit paper was used to create an even cross-sectional plane through the center of the solder balls. Two more polishing iterations were performed using 400 grit and then 600 grit. Care was taken to use a slow polishing speed and light pressure. The packages were then cleaned and allowed to dry in preparation of grid transfer. Alignment marks were placed on a ULE transfer grid. Transfer was then performed at 82 C from the ULE transfer grid for all 3 types of BGA packages, using the method described earlier. The electronic packages were attached to the bottom lip of the specimen holder, and the separation grid was attached to the back of the specimen holder after being rotated 90 degrees to the orientation of the packages. The rotation is necessary because the two sets of grid lines are not always exactly perpendicular. By rotating the grid 90 degrees, alignment of the transfer grid and the packages is ensured. Next, the PEMI was centered using the ULE transfer grating as a reference. The ULE grating experienced negligible thermal expansion when the grating was transferred to the specimens. At room temperature, the specimen gratings underwent 34

PAGE 43

negative thermal expansion while the transfer grating remained essentially the same, thereby making it the ideal reference to be used for centering. The centering process included alignment of the reference grating relative to the focal plane and adjustment of the U and V field mirrors in the PEMI. Once the centering process was completed, only slight adjustment of the electronic packages was necessary before data could be collected. Figure 12 and Figure 13 are representations of the PEMI 18 and will be referenced in the description of the process used to achieve a null field Figure 12: Front side view of the PEMI including labels. 35

PAGE 44

Figure 13: Back side view of the PEMI including labels. The axes used to describe the motion of the specimen are defined relative to the optics table and the PEMI as seen in Figure 7. The x-axis is defined perpendicular to the face of the PEMI. The y-axis is perpendicular to the optics table. The z-axis is normal to the xy-plane. The first step in centering the PEMI is the adjustment of the reference grating and the procedure is described as follows: 1) Remove all four caps from the PEMI. 2) Translate the reference grating along the x-axis until the four beams are focused on one shared area. This is the focal point of the PEMI. 3) Replace caps 1 and 2. 4) Rotate the reference grating about the y-axis. This should be performed while looking through the window located next to the beam attenuator. 5) A red dot should be visible on the circular disk inside the PEMI, and rotation of the grating about the z-axis will cause the dot to move vertically. Adjust the rotation until this dot is positioned beside the optical fiber at the center of the disk. 36

PAGE 45

6) Adjust the gratings angular orientation relative to the x-axis and the z-axis until the red dot is centered on the fiber tip. 7) Place one lens at the closest setting on the slider bar. Two blurry images should be observed on a sheet of paper when it is placed in front of the imaging lens. 8) Rotate the reference about the x-axis until the images merge. The reference grating is now adjusted properly in the focus plane of the PEMI. The PEMI can now be centered in the U and V fields using the following procedure: 1) Place two image lenses in series on the lens holder, and position them at approximately 20 cm along the slider. Translate the paper along the x-axis until it is in the focal plane of the lenses. This occurs when the images become small, well-defined spots. 2) Adjust the field changer to the V field, and adjust the V field mirrors until the spots merge. Make sure the red dot inside the PEMI stays on the fiber tip by making slight adjustments to the specimen holder. 3) Adjust the field changer to the U field. Two more small spots should be observed. Adjust the U field mirrors until the spots merge. 4) Remove one of the image lenses, and position the remaining lens at the closest position on the slider bar. Fringe patterns should now be observed on the image of the reference grating. Slight adjustments to the U field mirrors should yield a null fringe pattern (one or two fringes). 5) Adjust the field changer to the V field, and slightly adjust the V field mirrors until a null fringe pattern is obtained in the V field as well. The PEMI was now centered, meaning that it had been calibrated to the reference grating by obtaining a null field in the U and V directions. The PEMI was now ready to be used on the electronic packages, but the packages had to first undergo an alignment process. This process is similar to the process performed on the reference grating, except no adjustment is necessary on the U and V field mirrors of the PEMI, as it has already been centered. The process used was as follows: 1) Translate the electronic package along the x-axis until the four beams are focused on one shared area. This is the focal point of the PEMI. 37

PAGE 46

2) Adjust the field changer to the U field. 3) Rotate the reference grating about the y-axis. This should be performed while looking through the window located next to the beam attenuator. 4) A red dot should be visible on the circular disk inside the PEMI, and rotation of the package about the z-axis will cause the dot to move vertically. Adjust the rotation until this dot is positioned beside the optical fiber at the center of the disk. 5) Adjust the angular orientation of the package relative to the x-axis and the z-axis until the red dot is centered on the fiber tip. 6) Place two image lenses in series on the lens holder, and position them at approximately 20 cm along the slider. Translate the paper along the x-axis until it is in the focal plane of the lenses. This occurs when the images become small, well-defined spots. 7) Rotate the package about the x-axis until the spots merge. 8) Remove one of the image lenses, and position the remaining lens at the closest position on the slider bar. Fringe patterns should now be observed on the image of the package. 9) Fine adjust the rotation of the package about the x-axis until fringe symmetry is achieved about the vertical centerline of the electronic package. 10) Adjust the field changer to the V field. Symmetry should also be observed about the vertical centerline. Note that the vertical centerline of the packages could be used as a reference during the adjustment of the specimens. This was due to the fact that the vertical centerline of each package was also a line of symmetry, and therefore experienced no relative displacement. When properly adjusted, symmetry existed in the fringe patterns about the vertical centerline of the packages. If total symmetry was not achieved, it was due to the existence of a slight misalignment of the grid when it was transferred onto the electronic package. This would have affected the appearance of the fringe patterns, but not the results of the strain measurements obtained from the PEMI. 38

PAGE 47

The PEMI used in this study utilizes cross-sectional gridlines on the same internal reference grating. Since this internal grating was used as the reference for both the U and V fields, any fringe patterns due to misalignment that affected the U field would adversely affect the V field. This is due to the mutually dependent cross-derivatives of rigid body rotation mentioned previously. Therefore, when strain calculations were performed, any apparent strain due to rigid body rotation that was added to the U field would be subtracted from the V field 11 The resultant strain measurement would be the same as when no rigid body rotation was present. Before strain calculations could be made, the images of the fringes had to be recorded and analyzed. A full figure picture was taken of each package as well as a magnified view of the individual solder balls using a FujiFilm FinePix S2 Pro digital camera and FujiFilm photo shooting software. The images were then digitally analyzed using Image J software and the values for the strains could then be calculated using the finite difference method discussed previously. The use of Image J software is a tedious and time-consuming process, yet it is necessary due to the increase in accuracy of the data collected when compared to manually plotting and measuring the fringe differences. In addition, fringe resolution is greatly improved, which becomes a necessity when analyzing fringe patterns in composite materials such as the printed circuit board. Figure 14 and Figure 15 are to be referenced during the description of the procedure used to analyze the images using Image J software. 39

PAGE 48

Scale Line Measurement Line Figure 14: Example of Image J measurement and scale lines on image. 05010015020025000.10.20.30.40.5Distance mmGray Value Figure 15: Example of Image J gray scale plot. The following is a description of the process used to analyze the fringe using Image J software. 1) Draw a scale line on the image to be analyzed using the scale function in Image J. The line should span a feature that can be measured on the specimen the image represents. 2) Determine the value of the feature dimension represented by the scale line and enter its value in Image J. 40

PAGE 49

3) Using the analyze function in image J draw a measurement line on the image of the fringes to be analyzed. 4) Create a gray scale versus distance plot. The gray scale plot displayed light intensity versus length along the measurement line with 0 representing black and 255 representing white. This allowed the fringe difference measurements to be made between the peak intensities of the fringes, which yielded more accurate results than the method of manual plotting and measuring the differences. A local minimum on the plot created by the gray scale values accurately represented the maximum intensity of black on the fringe, and the difference between these intensity peaks gave an accurate estimation of the distance between the fringes. The finite difference method and equations discussed previously was then used to calculate strain data for the different components of the electronic packages. Pictures and tables of calculated data are presented in the experimental results section of this study. 41

PAGE 50

3. EXPERIMENTAL RESULTS Once a procedure for data collection was formulated and images were obtained, the repeatability of the procedure needed to be validated. This was accomplished by performing several iterations of the experimental procedure on all three BGA configurations under similar operating conditions. Two samples of each electronic package were analyzed, and the resultant data was checked for consistency between the separate specimens. The fringe patterns were found to be similar in pitch and shape for each specimen of all three BGA configurations. Figure 16 and Figure 17 are provided for comparison. These figures represent a full figure view of the U field fringes for FlexBGA specimens one and two respectively. Figure 16: Full figure view of the U field fringes for FlexBGA specimen one. Figure 17: Full figure view of the U field fringes for FlexBGA specimen two. Notice that both have similar fringe characteristics as well as spacing. The solder balls of the individual configurations were also compared, as demonstrated in 42

PAGE 51

Figure 18 and Figure 19 for the U field fringes of the right three solder balls of the Flex BGA. Figure 18: FlexBGA u fringes on the right three solder balls of specimen one. Figure 19: FlexBGA u fringes on the right three solder balls of specimen two. Notice that the same number of fringes is present on the solder balls and the fringes exhibit similar geometries. This process was repeated for the other two BGA configurations for all the solder balls with similar results. Once the repeatability of the procedure was validated, data could be collected and analyzed. Full figure images of the fringes, magnified images of the fringe 43

PAGE 52

patterns on the individual solder balls, and tabulated data of calculated strains of the three BGA configurations are presented in the following portion of this study. Figure 20 represents a full figure view of the FlexBGA before grid transfer was performed. The figure includes a description of the components as well as the numbering convention of the solder balls PCB Flex Substrate Mold Silicon Chip Solder Balls Figure 20: FlexBGA before grid transfer with description of components and numbering convention of solder balls. Solder Ball #: 1 2 3 4 5 6 7 8 9 10 11 12 13 Figure 21 and Figure 22 represent full figure views of the FlexBGA U and V fringe patterns after a high temperature grid transfer was performed at 82 C. Notice the symmetry in the fringe patterns about the vertical centerline. Figure 21: FlexBGA U fringes on profile after grid transfer. Figure 22: FlexBGA V fringes on profile after grid transfer. 44

PAGE 53

These images were used to measure normal strains in the U and V directions for the chip, mold, and PCB of the FlexBGA and are presented in Table 1. Table 1: U and V normal strains for the chip, mold and PCB of the FlexBGA. FlexBGA U V Chip 0.000408 Mold 0.000865 0.004481 PCB 0.001004 0.004831 Below resolution of the PEMI Figure 23 through Figure 30 are representations of the fringe patterns on the individual solder balls of the FlexBGA. Only the right half of the package is presented due to symmetry about the vertical centerline. The numbering convention of the solder balls is as described in Figure 20. 2 3 4 Figure 23: FlexBGA U fringes on solder balls 1, 2, 3 and 4. 45

PAGE 54

2 4 3 Figure 24: FlexBGA V fringes on solder balls 1, 2, 3 and 4. 7 5 6 Figure 25: FlexBGA U fringes on solder balls 5, 6 and 7. 46

PAGE 55

5 6 7 Figure 26: FlexBGA V fringes on solder balls 5, 6 and 7. 8 9 10 Figure 27: FlexBGA U fringes on solder balls 8, 9 and 10. 47

PAGE 56

8 9 10 Figure 28: FlexBGA V fringes on solder balls 8, 9 and 10. 11 12 13 Figure 29: FlexBGA U fringes on solder balls 11, 12 and 13. 48

PAGE 57

13 12 11 Figure 30: FlexBGA V fringes on solder balls 11, 12 and 13. Data was calculated for the normal strains at the midpoint of each solder ball using Image J software and the equations discussed in the Moir interferometry section of this paper. The data is presented in Table 2. Table 2: FlexBGA measured strains. solder ball # U Strain V Strain 1 0.0044450.009758 2 0.0047960.008327 3 0.0058870.012813 4 0.0069190.01091 5 0.0059620.008343 6 0.0065 0.009596 7 0.0105780.009832 8 0.0077450.011717 9 0.0080980.01004 10 0.0070340.010409 11 0.0075120.01597 12 0.0092410.015031 13 0.0072250.012316 49

PAGE 58

Figure 31 represents a full figure view of the PBGA before grid transference was performed. The figure includes a description of the components as well as the numbering convention of the solder balls. PCB Mold Silicon Chip Solder Balls Substrate Solder Ball #: 1 2 3 4 5 6 7 Figure 31: PBGA before grid transfer with description of components and numbering convention of solder balls. Figure 32 and Figure 33 represent full figure views of the PBGA U and V fringe patterns after a high temperature grid transfer was performed at 82 C. Figure 32: PBGA U fringes on profile after grid transfer. 50

PAGE 59

Figure 33: PBGA V fringes on profile after grid transfer. These images were used to measure normal strains in the U and V directions for the chip, mold, and PCB of the PBGA and are presented in Table 3. Table 3: U and V normal strains for the chip, mold and PCB of the PBGA. PBGA U V Chip 0.000198 0.00346 Mold 0.001443 0.003568 PCB 0.142086 0.003594 Figure 34 through Figure 39 are representations of the fringe patterns on the individual solder balls of the PBGA. Only the right half of the package is represented due to symmetry. The numbering convention of the solder balls is as described in Figure 31. 51

PAGE 60

3 2 1 Figure 34: PBGA U fringes on solder balls 1, 2 and 3. 3 2 1 Figure 35: PBGA V fringes on solder balls 1, 2 and 3. 52

PAGE 61

5 4 6 Figure 36: PBGA U fringes on solder balls 4, 5 and 6. 5 4 6 Figure 37: PBGA V fringes on solder balls 4, 5 and 6. 53

PAGE 62

6 7 Figure 38: PBGA U fringes on solder balls 6 and 7. 6 7 Figure 39: PBGA V fringes on solder balls 6 and 7. Data was calculated for the normal strains at the midpoint of each solder ball using Image J software and the equations discussed in the Moir interferometry section of this paper. The data is presented in Table 4. 54

PAGE 63

Table 4: PBGA measured strains. solder ball # U Strain V Strain 1 0.009665 0.014518 2 0.007106 0.013594 3 0.010227 0.016112 4 0.009563 0.011453 5 0.01005 0.015108 6 0.010509 0.013372 7 0.007248 0.016411 Figure 40 represents a full figure view of the TapeBGA before grid transference was performed. The figure includes a description of the components as well as the numbering convention of the solder balls. PCB Mold Silicon Chip Tape Substrate Solder Balls 6 5 4 3 2 1 Solder Ball # Figure 40: TapeBGA before grid transfer with description of components and numbering convention of solder balls. Figure 41 and Figure 42 represent full figure views of the TapeBGA U and V fringe patterns after a high temperature grid transfer was performed at 82 C. 55

PAGE 64

Figure 41: TapeBGA U fringes on profile after grid transfer. Figure 42: TapeBGA V fringes on profile after grid transfer. These images were used to measure normal strains in the U and V directions for the chip, mold, and PCB of the FlexBGA and are presented in Table 5. Table 5: U and V normal strains for the chip, mold and PCB of the FlexBGA. TapeBGA U V Chip 0.000378 Mold 0.002421 PCB 0.001113 * Below resolution of the PEMI Figure 43 through Figure 46 are representations of the fringe patterns on the individual solder balls of the TapeBGA. Only the left half of the package is represented due to symmetry. The numbering convention of the solder balls is as described in Figure 40. 56

PAGE 65

3 2 1 4 Figure 43: TapeBGA U fringes on solder balls 1, 2, 3 and 4. 3 2 1 4 Figure 44: TapeBGA V fringes on solder balls 1, 2, 3 and 4. 57

PAGE 66

4 5 6 Figure 45: TapeBGA U fringes on solder balls 4, 5 and 6. 4 5 6 Figure 46: TapeBGA V fringes on solder balls 4, 5 and 6. 58

PAGE 67

Data was calculated for the normal strains at the midpoint of each solder ball using Image J software and the equations discussed in the Moir interferometry section of this paper. The data is presented in Table 6. Table 6: TapeBGA measured strains. solder ball # U Strain V Strain 1 * 2 0.0074 0.009389 3 0.006026 0.00986 4 0.006395 0.010916 5 0.005069 0.011536 6 0.00639 0.007147 The experimental procedure developed in this study had now yielded repeatable results that could be used to measure strains in the solder balls of the three BGA configurations. This data now needed to be validated. This was performed by comparing the experimental results obtained in this study to the theoretical strains calculated by the finite element method using ANSYS software. The next section is devoted to a discussion of this process. 59

PAGE 68

4. MODELING 4.1. Modeling and Assumptions The data collected from Moir interferometry needed to be validated. This was accomplished by performing a non-linear finite element analysis on each of the chips using ANSYS software. During modeling, certain assumptions were found to be necessary. Past works were consulted to determine what assumptions were valid and the risks associated with them. 19 The following section discusses the process of developing the model, assumptions that were used, and their possible effect on the calculated results. The packages were modeled using factory specified dimensions as well as dimensions measured directly from the packages. Care was taken to use accurate dimensions, but some error may have been introduced due to slight discrepancies between the model and the specimen. The cross section of the electronic package was represented as a plane strain 2-D model. Existing studies have suggested that 2-D and 3-D models agree closely for Flip chip and BGA packages. 20-21 Only half the package was modeled due to symmetry about the centerline. In addition the model was constrained along the bottom in the vertical direction. This condition most emulated the boundary conditions of the experiment 60

PAGE 69

Certain material assumptions were made as well. The material properties used in the model assumed uniformity among the materials used in electronic packaging; when in fact, the materials of each individual package vary slightly. The printed circuit board is made of a composite material, but was modeled as a homogenous material. This was deemed acceptable because the complexity exhibited in the behavior of composite materials was beyond the scope of this study. These assumptions simplified the analysis, but may have introduced some error. In past studies, Anands model was used to define the creep behavior of the solder balls in ANSYS models 22 This was determined to be unnecessary, and the solder balls were modeled as a linear elastic material with temperature dependant properties. This decision was made based on the fact that the solder balls do not stay at the elevated temperature for a prolonged period of time; and therefore, the effects due to creep were assumed to be negligible. The simplifications that were made to the model were performed to meet the specifications of this study. The analysis was used only as a broad check of the experimentally determined data. A compromise had to be made between the accuracy of the model versus computation time and ease of modeling. 4.2. Material Properties The three BGA configurations modeled in this study were a 10 x 10 mm TapeBGA having a 6.75 mm die, .8mm pitch, and 144 solder ball count; a 15 x 15 mm PBGA having a 6.35 mm die, 1mm pitch, and 196 solder ball count; and a 27 x 27 mm FlexBGA having a 6.35 mm die, 1 mm pitch, and 672 solder ball count. Although 61

PAGE 70

package size, die size, pitch, and ball count varied among the three packages; each package was assembled from components common to all three. These components consisted of a printed circuit board, solder pads, solder balls, a substrate, a silicone die, and a mold. A schematic of the components is presented in Figure 47 and Table 7. 6 4 5 2 3 1 Figure 47: Schematic of BGA configurations. Table 7: Standard component list of BGA configurations. Figure # Material 1 Printed Circuit Board 2 Solder Pad 3 Solder Ball 4 Chip (Die) 5 Substrate 6 Mold 62

PAGE 71

The material properties of the components were obtained from manufacturing and compared with previous literature. 23-29 These properties are represented in Table 8 and Table 9. Table 8: Material properties of BGA components. # Material E(10 9 Pa) v (10 -6 / o C) Printed Circuit Board FR4 PCB 22 0.28 18.5 Solder Pad Copper 76 0.35 17 Solder Ball 63Sn/37Pb Table 9 0.4 21 Chip Silicon 131 0.3 2.8 Substrate A BT 26 0.39 15 Substrate B Polyimide 3.291 .41 67.5 Mold A SMT-B-1RC 1.31 0.35 15 Mold B SMT-B-1 1.41 0.35 14 Table 9: Modulus of elasticity of 63Sn/37Pb solder. Temperature ( o K) 280 320 360 E(10 9 ) Pa 33.367 27.299 21.231 The material properties of the printed circuit board, solder pads, solder balls, and silicon chip were consistent among the three BGA configurations. The components whose materials differed among the BGA configurations were the substrate and the encapsulant. The FlexBGA and Tape use substrate B and mold B. The PBGA uses substrate A and mold A. It should be noted that ANSYS differentiates between thermal and mechanical strains. The strain values presented in this document were obtained by summing these 63

PAGE 72

two components in order to allow the strains obtained experimentally from Moir to be compared to the strains obtained from the finite element analysis. 4.3. Results of Modeling A non-linear finite element analysis was performed on the 2-D FlexBGA model using the previously presented linear material properties. The model was meshed using 8-node, plane strain elements. A ramped temperature load was used to simulate the conditions 355 K 294 K (approximately 82 C 21C). Only half of the package was modeled due to symmetry, and the model was constrained in the x-axis along the centerline and the y-axis along the bottom of the model. The analysis was performed over several iterations while refining the mesh until convergence was achieved. The results of this process are presented in Figure 48 and Figure 49 both as contour plots and queried results of the individual package components. The values in parentheses are the experimental strains measured at the same location with the PEMI. 64

PAGE 73

Die -3.88E-4 ( 4.08E 4) Mold 1.05E-3 ( 8.65E 4) Solder -1.73E-3 ( 5.89E 4) PCB -1.45E-3 ( 1.08E 4) Figure 48: U field normal strains for FlexBGA as calculated with ANSYS and compared to experimental values of individual components. Die -1.86E-4 ( ) Mold -1.09E-3 ( 1.35E 3) Solder -6.40E-3 ( 1.22E 2) PCB -4.02E-3 ( 4.83E 3) Figure 49: V field normal strains for FlexBGA as calculated with ANSYS and compared to experimental values of individual components. 65

PAGE 74

A non-linear finite element analysis was performed on the 2-D PBGA model using the previously presented linear material properties. The model was meshed using 8-node, plane strain elements. A ramped temperature load was used to simulate the conditions 355 K 294 K (approximately 82 C 21C). Only half of the package was modeled due to symmetry, and the model was constrained in the x-axis along the centerline and the y-axis along the bottom of the model. The analysis was performed over several iterations while refining the mesh until convergence was achieved. The results of this process are presented in Figure 50 and Figure 51 both as contour plots and queried results of the individual package components. The values in parentheses are the experimental strains measured at the same location with the PEMI. Die -2.67E-4 ( 2.40E 3) Mold -1.37E-3 1.44E 3) Solder -1.96E-3 ( 1.02E 2) PCB -1.27E-3 ( 1.42E 3) Figure 50: U field normal strains for PBGA as calculated with ANSYS and compared to experimental values of individual components. 66

PAGE 75

Die -1.09E-3 ( ) Mold -1.39E-3 ( 1.52E 3) Solder -3.28E-3 ( 1.61E 2) PCB -3.29E-3 ( 3.59E 3) Figure 51: V field normal strains for PBGA as calculated with ANSYS and compared to experimental values of individual components. A non-linear finite element analysis was performed on the 2-D TapeBGA model using the previously presented linear material properties. The model was meshed using 8-node, plane strain elements. A ramped temperature load was used to simulate the conditions 355 K 294 K (approximately 82 C 21C). Only half of the package was modeled due to symmetry, and the model was constrained in the x-axis along the center line and the y-axis along the bottom of the model. The analysis was performed over several iterations while refining the mesh until convergence was achieved. The results of this process are presented in Figure 52 and Figure 53 both as contour plots and queried results of the individual package components. The values in parentheses are the experimental strains measured at the same location with the PEMI. 67

PAGE 76

Die -3.96E-4 ( 3.78E 4) Mold -2.10E-3 (-2.42E-3) Solder -1.56E-3 ( 6.40E 3) PCB -1.22E-3 ( 1.11E 3) Figure 52: U field normal strains for TapeBGA as calculated with ANSYS and compared to experimental values of individual components. Die -1.98E-4 ( ) Mold -8.96E-4 ( ) Solder -2.28E-3 ( 9.86E 3) PCB -9.88E-4 ( ) Figure 53: V field normal strains for TapeBGA as calculated with ANSYS and compared to experimental values of individual components. 68

PAGE 77

The results of the above analyses were then compared to the experimental data obtained using the PEMI. As given in Table 10 through Table 15. This comparison is discussed in detail in the following section. Table 10: U normal strain comparison between calculated and experimental results for the FlexBGA. FlexBGA U Moir U mech x therm x total x % diff x Chip -4.08E-04 -2.17E-04 -1.71E-04 -3.88E-04 4.90% Mold -8.65E-04 -1.97E-04 -8.54E-04 -1.05E-03 21.50% PCB -1.08E-03 -3.16E-04 -1.13E-03 -1.45E-03 33.89% Solder -5.89E-03 -4.54E-04 -1.28E-03 -1.73E-03 70.56% Table 11: V normal strain comparison between calculated and experimental results for the FlexBGA. FlexBGA V Moir V mech y therm y total y % diff y Chip -1.46E-05 -1.71E-04 -1.86E-04 na Mold -1.35E-03 -2.35E-04 -8.54E-04 -1.09E-03 19.33% PCB -4.83E-03 -2.89E-03 -1.13E-03 -4.02E-03 16.77% Solder -1.22E-02 -5.12E-03 -1.28E-03 -6.40E-03 47.54% below resolution Table 12: U normal strain comparison between calculated and experimental results for the PBGA. PBGA U Moir U mech x therm x total x % diff x Chip -2.40E-04 -9.56E-05 -1.71E-04 -2.67E-04 11.08% Mold -1.44E-03 -4.57E-04 -9.15E-04 -1.37E-03 4.72% PCB -1.42E-03 -1.35E-04 -1.13E-03 -1.27E-03 10.92% Solder -1.02E-02 -6.81E-04 -1.28E-03 -1.96E-03 80.77% 69

PAGE 78

Table 13: V normal strain comparison between calculated and experimental results for the PBGA. PBGA V Moir V mech y therm y total y % diff y Chip -9.20E-04 -1.71E-04 -1.09E-03 na Mold -1.52E-03 -4.77E-04 -9.15E-04 -1.39E-03 8.42% PCB -3.59E-03 -3.16E-03 -1.13E-03 -3.29E-03 8.36% Solder -1.61E-02 -3.00E-03 -1.28E-03 -3.29E-03 79.57% below resolution Table 14: U normal strain comparison between calculated and experimental results for the TapeBGA. TapeBGA U Moir U mech x therm x total x % diff x Chip -3.78E-04 -2.25E-04 -1.71E-04 -3.96E-04 4.76% Mold -2.42E-03 -1.25E-03 -8.54E-04 -2.10E-03 13.06% PCB -1.11E-03 -9.45E-05 -1.13E-03 -1.22E-03 10.32% Solder -6.40E-03 -2.79E-04 -1.28E-03 -1.56E-03 75.64% Table 15: V normal strain comparison between calculated and experimental results for the TapeBGA. TapeBGA V Moir V mech y therm y total y % diff y Chip -7.28E-06 -1.71E-04 -1.98E-04 na Mold -4.19E-05 -8.54E-04 -8.96E-04 na PCB -3.45E-04 -1.13E-03 -9.88E-04 na Solder -9.86E-03 -9.97E-04 -1.28E-03 -2.28E-03 76.91% below resolution 70

PAGE 79

5. DISCUSSION AND CONCLUSION The objective of this work to formulate and document a procedure to measure the thermo-mechanical strains in BGA electronic packaging using Moir interferometry. This procedure includes the preparation of the specimen, the replication and transfer of the grids, the techniques of Moir interferometry, interpretation of results, and validation of data by finite element analysis using ANSYS software. The process was performed on FlexBGA, PBGA, and TapeBGA packages over several iterations and tested for repeatability. The fringe patterns obtained from the individual iterations maintained uniformity in geometry and spacing for each of the three packages. General trends were also observed among the different BGA configurations that implied uniform qualitative behavior in the common components of the packages. The chip, for example, exhibited less strain than the PCB for all three packages, as was expected due to the difference in coefficient of thermal expansion and effective stiffness. These trends held true for each iteration of the procedure to which the packages were exposed. The procedure was therefore determined to be repeatable, and the data that was collected needed to be quantitatively validated. Quantitative validation was performed by comparing the experimental data to data obtained from performing a finite element analysis on each of the packages using ANSYS software. U and V field linear normal strains obtained from the specimen 71

PAGE 80

were compared to strains calculated using a plane strain, finite element analysis on a 2-D model of the cross-section of the BGA packages. Linear normal strains were compared due to the ease of acquisition of experimental data. If agreement was found for both the U and the V field between the experimental and computed data, then the shear strains must also be in agreement. This is due to the compatibility relationship: 22222dxddyddxdydyyxxxy Though the substrates of the packages were too thin to obtain an accurate experimental strain measurement, the packages exhibited fairly good quantitative uniformity for most of the package components when comparing experimental to computational data. The experimental strain values obtained for the encapsulants of each package exhibited an average difference from FEM of 20.5%, 6.5%, and 13% for the FlexBGA, PBGA, and Tape BGA respectively. The experimental strain values obtained for the PCB exhibited an average difference from FEM of 25.3%, 9.6%, and 10.3% for the FlexBGA, PBGA, and Tape BGA respectively. The experimental strain values obtained for the die of each package exhibited an average difference from FEM of 4.9%, 11.1%, and 4.8% for the FlexBGA, PBGA, and Tape BGA respectively. The experimental strains in the solder balls did not agree as closely and exhibited an average difference from FEM of 59.1%, 80.2%, and 76.2%, and for the respective packages. Any nonconformity of experimental and computational data for the components could be due to several factors and should be addressed in a further study. Experimental error could have occurred due to the relatively short period that the 72

PAGE 81

packages were maintained at the elevated temperature during the grid transfer. The assumption made during this study was that the solder balls would undergo negligible creep at the relatively low elevated temperature used in this study; therefore the effects of creep could be ignored. In reality some undetermined amount of creep could have occurred while the grating was adhering to the specimen, thereby introducing apparent residual strains into the grating. If the specimens would have been kept at the elevated temperature for a sufficient period of time to allow them to creep to a steady state condition before application of the grating, then the residual strains in the grating would not have been an issue. The necessary amount of time the specimen must be kept at the elevated temperature could be determined experimentally 30 or predicted by a finite element analysis using Anands model in ANSYS. 31 Anands model is a creep model in ANSYS and has been shown to effectively simulate the creep behavior of materials with temperature dependant properties such as solder. Other sources of possible error could have occurred due to the over simplification of the computer model. For instance, dimensions used when modeling the specimens came from nominal manufacturing specifications. These specifications have large tolerances associated with them. For example, the solder ball width of the TapeBGA is given as .48 mm ( .05 mm). This implies that the dimensions of the solder balls used for this study could have differed by up to 10% from the manufacturing dimension used in the computer model. This would not only affect the accuracy of the model, but experimental error could occur if the dimension in question 73

PAGE 82

were used to determine the scale factor for the image analysis portion of the experimental procedure. To circumvent this source of error all the dimensions could be measured microscopically. This could be performed by obtaining magnified images of the specimen before the grid is transferred. These images could then be analyzed using image J software in much the same way the fringe images obtained from the PEMI were analyzed to obtain data. The result would be more exact dimensions of the specific package to be modeled and a more accurate scale factor for image processing. Another possible source of error could have come from the material properties used to model the components of the packages, specifically the solder balls. The material properties of solder are temperature dependent, but were assumed to be linear over the range of temperatures experienced in this study. A more accurate representation of the material properties of the solder balls could be obtained by curve fitting experimentally determined data to better represent the behavior of the components over the intended temperature range. 32-34 In addition, the PCB was assumed to have homogeneous material properties, when in fact it is made of a composite material. This was deemed acceptable because the complexity involved in modeling the behavior of composite materials is not justified by a noticeable increase in accuracy of the model. Another issue arose from the fact that acquisition of experimental data from the PCB was made difficult due to non-uniformity of the fringe patterns. An algorithm could be developed 35 to compute strain values from the images and display them as contour plots. This would reduce the error due to image post processing. 74

PAGE 83

This was the first in an ongoing series of studies intended to investigate the thermo-mechanical behavior of Ball Grid Array electronic packages. A reliable, repeatable procedure to obtain strain measurements on the packages using Moir interferometer was developed and documented. The resultant experimental and computational data maintained uniformity to the extent required by the scope of this study. Potential sources of error have been identified and should be the subject of further research. 75

PAGE 84

REFERENCES 1 Harper, C., Electronics Packaging and Interconnections Handbook, 2000, The McGraw-Hill Companies, Inc. 2 Lau, J., Thermal Stress and Strain in Microelectronics Packaging, 1993, Van Nostrand Reinhold, New York, New York. 3 Su, B., Hareb, S., Lee, Y., Solder Joint Reliability for a 540-I/O Plastic Ball Grid Array Assembly, IEEE 1998 International Conference on Multichip Modules and High Density Packaging, pp. 422 428. 4 Sylvester, M., Banks, D., Kern, R., Pofahl, R., Thermomechanical Reliability Assessment of Large Organic Flip-Chip Ball Grid array Packages, IEEE 1998 Electronics Components and Technology Conference, pp. 851 860. 5 Darveaux, R., Banerji, K., Fatigue Analysis of Flip Chip Assemblies Using Thermal Stress Simulations and a Coffin-Manson Relation, IEEE 1991 Electronics Components and Technology Conference, pp. 797 805. 6 Ratanawilaj, T., Hunter, B., Rose, D., A comparison Between Moir Interferometry and Strain Gages for Effective CTE Measurement in Electronic Packaging, IEEE 2000 Inter Society Conference on Thermal Phenomenon, pp. 246 252. 7 Dally, J., Riley, W., Experimental Stress Analysis, 1991, McGraw-Hill, Inc., New York, New York. 8 Pecht, M., Handbook of Electronic Packaging Design, 1991, Marcel Dekker, Inc., New York, New York. 9 Morris, J., Electronics Packaging Forum, Vol. 1, Van Nostrand Reinhold, New York, New York. 10 Tummala, Rao R., Rymaszeweski, Eugene J., Microelectronic Packaging Handbook, 1989, Van Nostrand Reinhold, New York, New York. 76

PAGE 85

11Post, D., Han, B., Ifju, P., High Sensitivity Moir, 1994,Springer-Verlag New York, Inc., New York, New York. 12 Kobayashi, A., Handbook on Experimental Mechanics, 1993, Society for Experimental Mechanics, Bethel, CT. 13 Durelli, A., Parks, V., Moir Analysis of Strain, 1970, Prentice-Hall, Inc., Englewood Cliffs, N.J. 14 Han, B., Columbus, D., Wu, Z., Mechanical Fringe Shifting in Moir Interferometry, Experimental Techniques, v. 23 no. 1 (Jan./Feb. 1999), pp. 16 19. 15 Bastawros, A., Voloshin, A., Transient Thermal Measurements in Electronic Packages, IEEE Transactions on Components, Hybrids, and Manufacturing Technology, v. 13 no. 4 (Dec. 1990), pp. 961 966. 16 Miller, M., Mohammed, I., Dai, X., Jiang, N., Ho, P., Analysis of Flip-Chip Packages using High Resolution Moir Interferometry, IEEE 1999 Electronics Components and Technology Conference, pp. 979 986. 17 Liu, S., Wang, J., Zou, D., He, X., Quian, Z., Resolving Displacement Field of Solder Ball in Flip-Chip Package by Both Phase Shifting Moir Interferometry and FEM, IEEE 1998 Electronics Components and Technology Conference, pp. 1345 1353. 18 Photomechanics Inc., www.photomechanics.com. 19Ozmat, B., A Nonlinear Thermal Stress Analysis of Surface Mount Solder Joints, IEEE 2000 Inter Society Conference on Thermal Phenomenon, pp. 959 972. 20 Satoh,R., Arakawa, K., Harada, M., Katsuhiro, M., Finite Element Modeling of BGA Packages for Life Predictions, IEEE 2000 Inter Society Conference on Thermal Phenomenon, pp. 1059 1063. 21 Pang, J., Flip Chip on Board Solder Joint Reliability Analysis Using 2-D and 3-D FEA Models, IEEE Transactions on Advanced Packaging, v. 24 no. 4 (Nov. 2001), pp. 499 506. 22 Gustafsson,G., Finite Element Modeling of BGA Packages for Life Predictions, IEEE 2000 Inter Society Conference on Thermal Phenomenon, pp. 1059 1063. 77

PAGE 86

23Harper, C., Electrical Packaging and Interconnection Handbook, 1991, McGraw-Hill, Inc., New York, New York. 24Lau, J., Lee, S., Pan, S., Chang, C., Flip Chip Applications, 2002, Business News Publishing Co. 25 Practical Components, www.amkor.com 26 Cookson Electronics, www.cooksonsemi.com 27 Injectorall Electronics Corp., www.injectorall.com 28Online Material Database, www.matwb.com 29Mercado, L., Integrated Transient Thermal and Mechanical Analysis of Molded PBGA Packages During Thermal Shock, IEEE Transactions on Advanced Packaging, v. 24, No. 1, February 2000. 30 Frost, J., Creep and Tensile behavior of Lead-Rich Lead-Tin Solder Alloys, IEEE Transactions on Components, Hybrids, and Manufacturing Technology, v. 11 no. 4 (Dec. 1988), pp. 371-399. 31 Wang, J., Qian, Z., Zou, D., Liu, S., Creep Behavior of a Flip-Chip Package by Both FEM Modeling and Real Time Moir Interferometry, IEEE 1998 Electronics Components and Technology Conference, pp. 1438 1445. 32 Moore, T., Jarvis, J., Failure Analysis and Stress Simulations in Small Multichip BGAs, IEEE Transactions on Advanced Packaging, v. 24 no. 2 (May 2001), pp. 216 221. 33 Huang, X., Lee, S., Yan, C., Hui, S., Characterization and Analysis on the Solder Ball Shear Testing Condition, IEEE 19968 International Conference on Multichip Modules and High Density Packaging, pp. 611 635. 34 Ryohei, S., Arakawa, K., Harada, M., Matsui, K., Thermal Fatigue Life of Pb-Sn Alloy Interconnections, IEEE Transactions on Components, Hybrids, and Manufacturing Technology, v. 14 no. 1 (March 1991), pp. 224 232. 78

PAGE 87

35 Kadooka, K., Kunoo, K., Uda, N., Ono, K., Nagayusa, T., Strain Analysis for Moir Interferometry Using the Two-dimensional Continuous Waveform, Experimental Mechanics,v. 43, no.1 (March 2003), pp45 51. 79


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200457Ka 4500
controlfield tag 006 m d
007 cr bn
008 031007s2003 flu sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000078
035
(OCoLC)52443500
9
AJL4021
b SE
040
FHM
c FHM
049
FHME
090
TJ145
1 100
Rivers, Norman.
3 245
An investigation of BGA electronic packaging using Moir interferometry
h [electronic resource] /
by Norman Rivers.
260
[Tampa, Fla.] :
University of South Florida,
2003.
502
Thesis (M.S.M.E.)--University of South Florida, 2003.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 87 pages.
520
ABSTRACT: As technology progresses towards smaller electronic packages, thermo-mechanical considerations pose a challenge to package designers. One area of difficulty is the ability to predict the fatigue life of the solder connections. To do this one must be able to accurately model the thermo-mechanical performance of the electronic package. As the solder ball size decreases, it becomes difficult to determine the performance of the package with traditional methods such as the use of strain gages. This is due to the fact that strain gages become limited in size and resolution and lack the ability to measure discreet strain fields as the solder ball size decreases. A solution to the limitations exhibited in strain gages is the use of Moir interferometry. Moir interferometry utilizes optical interferometry to measure small, in-plane relative displacements and strains with high sensitivity. Moir interferometry is a full field technique over the application area, whereas a strain gage gives an average strain for the area encompassed by the gage. This ability to measure full field strains is useful in the analysis of electronic package interconnections; especially when used to measure strains in the solder ball corners, where failure is known to originate. While the improved resolution of the data yielded by the method of Moir interferometry results in the ability to develop more accurate models, that is not to say the process is simple and without difficulties of it's own. Moir interferometry is inherently susceptible to error due to experimental and environmental effects; therefore, it is vital to generate a reliable experimental procedure that provides repeatable results. This was achieved in this study by emulating and modifying established procedures to meet our specific application. The developed procedure includes the preparation of the specimen, the replication and transfer of the grids, the use of the PEMI, interpretation of results, and validation of data by finite element analysis using ANSYS software. The data obtained maintained uniformity to the extent required by the scope of this study, and potential sources of error have been identified and should be the subject of further research.
590
Adviser: Eason, Thomas
0 650
Ball grid array technology.
Microelectronic packaging
x Reliability.
Interferometry.
Moir method.
653
electronic package.
bga.
moire interferometry.
690
Dissertations, Academic
z USF
Mechanical Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
949
FTS
SFERS
ETD
TJ145 (ONLINE)
sv 6/16/03
4 856
u http://digital.lib.usf.edu/?e14.78