MEMS Packaging Reliability in Board Level Drop Tests Under Severe ...

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY

1

MEMS Packaging Reliability in Board-Level Drop Tests Under Severe Shock and Impact Loading Conditions—Part I: Experiment Jingshi Meng, Stuart T. Douglas, and Abhijit Dasgupta Abstract— The continuing increase of functionality, miniaturization, and affordability of handheld electronic devices has resulted in a decrease in the size and weight of the products. As a result, printed wiring assemblies (PWAs) have become thinner and more flexible, and clearances with surrounding structures have decreased. Therefore, new design rules are needed to minimize and survive possible secondary impacts between PWAs and surrounding structures because of the consequential amplification in acceleration and contact stress. This paper is the first of a two-part series and focuses on the drop test reliability of commercial off-the-shelf microelectromechanical systems (MEMS) components that are mounted on printed wiring boards (PWBs). Particularly in this paper, we are interested in gaining preliminary insights into the effects of secondary impacts (between internal structures) on failure sites in the MEMS assemblies. Drop tests are conducted under highly accelerated conditions of 20 000 g (“g” is the gravitational acceleration). Under such high accelerations, the stress levels generated are well beyond those expected in conventional qualification tests. Furthermore, secondary impacts of varying intensities were allowed by changing the clearance between the PWB and the fixture. As a result, the stress and accelerations are further amplified, to mimic unexpected secondary impacts in a product if/when design rules fail to avoid such conditions. The amplification of the test severity is quantified by comparing the characteristic life (η in a Weibull distribution) of all the tested MEMS components at each clearance. Multiple failure sites from drop testing are identified, from packaging-level failures to MEMS device failures. The participation of competing failure sites is also demonstrated via characteristic life representations of each failure site at various clearances. Index Terms— Durability, failure sites, microelectromechanical systems (MEMS), repetitive drop test, secondary impact, ultrahigh acceleration.

I. I NTRODUCTION

M

ICROELECTROMECHANICAL systems (MEMS) components are widely utilized in many different applications, such as sensors, portable consumer electronics, radio frequency switches, and power devices in automotive, aerospace, and military electronics. The use conditions for MEMS-based microsystems can be rather harsh in some Manuscript received March 22, 2016; revised September 12, 2016; accepted September 14, 2016. Recommended for publication by Associate Editor A. Chandra upon evaluation of reviewers’ comments. The authors are with the Center for Advanced Life Cycle Engineering Electronic Products and Systems, Mechanical Engineering Department, University of Maryland, College Park, MD 20742 USA (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCPMT.2016.2611646

applications (especially in sensing applications), such as harsh chemicals, extreme thermal and humidity environments, shocks, and drops [1], [2]. In particular, portable electronic devices are commonly exposed to impact loading due to accidental drops, therefore the reliability of MEMS assemblies under shock and impact is critically important. Especially, the dynamic response of the moving parts in MEMS [2], [3] makes identifying the root causes of failures in MEMS assemblies a very challenging task. At the moment of impact, kinetic energy of a freefalling packaged device converts into strain energy of both the external housing [6] and the internal structures of a device [7]. These dynamic deformations cause stress concentrations at the interconnects in the package [8]–[10]. Studies of product level drop tests have shown that when a mobile phone is dropped on a hard surface from a height of 1.5 m, accelerations at the printed wiring boards (PWB) level can reach 10 000 g [11]. Therefore, drop test methods for producing acceleration impulses of the order of 10 000 g have received recent attention. Moreover, the dense packaging in high-performance portable electronic products may generate internal collisions between the structures inside the product housing [6], [11], [12] after a shock experienced by the whole system. Such internal collisions, termed “secondary impact” in this discussion [13], could possibly happen between slender circuit cards and other relatively rigid internal surfaces, such as product housing interior, displays, battery case, etc. Typical board level drop tests specified in most industry standards (see [14]) for conventional IC packages [8], [12], [15]–[18] have no secondary impacts. In such tests, failures at package interconnects are mainly caused by large PWB bending deformation and inertial forces due to the mass of internal structures. By contrast, drop test with secondary impact can significantly increase the severity of the primary impact in two ways: 1) amplification of the dynamic loads discussed above, and 2) creation of a new source of stress pulse at the contact site and propagation of the contact stress wave through the structure [19]. Hart and Hermann [20], Harter et al. [21], and Kerwin [22] investigated the concept of achieving high-impact accelerations through the method of momentum transfer and velocity amplification through impacts between moving bodies. A velocity amplifier method is already commercially available as a shock test methodology to achieve high accelerations, based on the dynamics of velocity amplification through

2156-3950 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 2

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY

pairwise collisions between multiple masses in a chain [23], [24]. This commercial device is termed a dual mass shock amplifier (DMSA). Douglas et al. [13] demonstrated that secondary impacts in a DMSA can cause a significant amplification of accelerations at the impact site by conducting a simplified analysis of a two degreeof-freedom spring-mass system. Douglas et al. [13] and Meng et al. [25] have shown that one source of amplification in acceleration might be the multiple dynamic mode shapes of the PWB excited by the secondary impact. Drop test reliability of MEMS is a fast-growing field of research. Li and Shemansky, Jr., [3] analyzed the failure risks in a micromachined structure when subjected to a free-fall drop. The theoretically calculated impact acceleration can be up to 105 g from only a 1.2 m free fall on a hard surface. As a result, the micromachined structure in the experiment encountered a high failure rate. Sheehy et al. [26] reported that microscale cantilevers are generally durable in common drop tests, and failures appear to be significant only when the input acceleration is as high as 40 000 g. Kaplan et al. [27] developed a capacitance measurement technique for dynamic impact events. According to them, high acceleration up to 24 000 g attributes to the majority of circuit breakdown of surface-mount capacitors. The dominant failure mechanism in MEMS systems (e.g., silicon fracture, stiction, and contamination) varies depending on the MEMS design and loading condition [28]. Li et al. [29] performed via testing and multiscale simulation on a three-axis MEMS gyroscope and found that the package failures of the MEMS occurred at 8000 g, whereas stiction and fracture of the comb structure occurred at 4000 g. Lall et al. [30] analyzed the capacitance sensitivity of an off-the-shelf MEMS accelerometer when subjected to drop and shock via a multiphysics simulation. Srikar and Senturia [4] studied the transient response of MEMS devices due to a shock pulse applied to the silicon substrate. According to them, the package design in a MEMS assembly can alleviate the severity of the shock pulse, and they were therefore able to model the worst-case scenario by assuming the substrate to be a rigid body. On the other hand, Alsaleem et al. [31] pointed out that the presence of secondlevel packaging for MEMS devices may further amplify the dynamic response of the microstructure mounted on its top. Tilmans et al. [32] reviewed various approaches for MEMS packaging, from wafer-level (0-level) to chip-level (1-level) packages, and emphasized the importance of MEMS packaging in MEMS reliability. In a capacitance measurement technique study by Kaplan et al. [27], the flexural deformation of the PCB reduces the survivability of the capacitor mounted on PCBs. In view of the importance of the package design, this paper focuses on fully packaged MEMS assemblies rather than on bare MEMS devices. Secondary impacts on printed circuit board assembly (PCBA) and their effects on MEMS failures have been semiqualitatively addressed in the past [13], [25]. In particular, secondary impact drop test results for MEMS assemblies at 20 000 g were partially reported earlier [1], [2]. However, since tests were conducted on a limited number of conditions and replicates, only semiqualitative conclusions were provided

Fig. 1.

Drop tower with DMSA accessory.

because there were many survivors at the end of that test program (the maximum drop count was limited to 500). The data lacked sufficient granularity to differentiate between multiple dominant failure modes at each test condition. Also, the data lacked sufficient depth to study the variance in the likelihood of each failure site at each test condition. The present study extends the previous work by extending all tests to failure (requiring up to 2000 drops in some cases), increasing the number of test conditions to achieve a more comprehensive variation of test conditions, identifying the dominant competing failure modes in the tested MEMS structure, separating the failure data by dominant failure modes, and quantifying the statistical distributions of multiple competing failure modes. Furthermore, in Part II of this two-part series, semiempirical failure models are developed for the dominant failure modes. In this paper, we quantitatively demonstrate that drop testing of a printed wiring assembly (PWA) with MEMS microphones causes packaging-related failures more often than MEMS device failures. In addition, this paper highlights a transition of the dominant failure site as a function of severity of the impact between the PWB and the fixture underneath. II. T ESTING A PPROACH In this section, the test setup and test matrix for high-acceleration (of the order of 10 000 g) drop testing are introduced. A. Secondary Impact Test Setup and DMSA The setup for high-acceleration drop testing in this paper is an extension of the principle of multiple secondary impacts. One secondary impact occurs at the DMSA, (a commercial accessory of the drop tower, as shown in Fig. 1 [33], [34]). The other secondary impact is from the dynamic contact between the PWA and the fixture when the PWB deflection exceeds the designed finite clearance between them. Thus, including the primary impact, there are three consecutive impacts in this drop testing setup to enhance the acceleration magnifications in the board level drop test. As shown in Fig. 1, the tester consists of a drop table on four primary guide rods, a seismic base on four shock absorbers,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. MENG et al.: MEMS PACKAGING RELIABILITY IN BOARD LEVEL DROP TESTS PART I

Fig. 2.

3

20 000 g acceleration history measured on top of DMSA.

the DMSA accessory, and various pulse shaping materials. Pulse shaping materials help to determine the magnitude and duration of acceleration profiles. The DMSA accessory consists of a base rigidly mounted to the top of the drop table and a secondary drop table suspended on four secondary guide rods by four linear springs. The drop table falls along the guide rods, from a given height onto the seismic base. The primary impact between the drop table and the seismic table is capable of producing repeatable impact accelerations up to 5000 g at the DMSA base. Secondary impact is produced by the impact between the DMSA drop table and the DMSA base, to achieve almost 20× amplifications of the acceleration level at the DMSA table (producing accelerations up to 100 000 g). As discussed in the previous work by Douglas [35], the magnification of acceleration [left term in (1), A0 is the peak acceleration at the base and A1 is the magnified acceleration on the DMSA] is determined by a stiffness ratio and a mass ratio of the system. The stiffness ratio is defined based on the properties of two pulse-shaping materials: K 1 (placed between the contact pair of the primary impact) and K 2 (placed between the contact pair of the secondary impact). The mass ratio is taken from two impacting parts: DMSA base and drop table as one part (M1 ) and DMSA table and fixture as the other part (M2 ). A sample acceleration history measured on top of the DMSA is shown in Fig. 2. The drop tower with DMSA provides good drop-to-drop repeatability with an error range of ±3% for peak accelerations  K 2 M1 K2 A1 = + . (1) A0 K 1 M2 K1 As discussed above, a special fixture design, first introduced by Douglas et al. [13] is employed to generate additional impacts between the test PWB and the fixture for even further amplification of accelerations. This “tertiary impact,” between the bottom of the PWB and the fixture placed on top of the DMSA drop table, is shown in Fig. 3(d). To parametrically investigate the effect of clearance, a set of fixtures was designed with varying clearances between the PWB bottom and the fixture. The clearance ranged from zero clearance to infinite clearance, with various finite clearances in between [Figs. 3(a)–(c) and 4], ranging from 20% to 120% of the PWB thickness. The test specimen consists of MEMS microphone

Fig. 3. Finite clearance clamping method. (a) Zero clearance. (b) Infinite clearance. (c) Spacers for finite clearance. (d) Schematic of secondary impact between PWA and fixture base.

Fig. 4.

Drop orientations for MEMS microphone components.

Fig. 5. Test PWA with packages locations along the y-axis and generic specimen design.

components assembled on a PWB, as shown in Fig. 5. In the remainder of this paper, the term “secondary impact” only refers to the impact between the PWB and the fixture. Stress distribution in the package and interconnects also depends on the package orientation, due to the inertial force generated by the mass of the package and PWB strain at the package’s footprint. To explore the effect of component mounting orientation during drop tests, PWAs are tested with the package facing two opposite directions, either upward or downward. As demonstrated in Fig. 4, trenches are added on the top surface of the fixture, along the two orthogonal x and y centerlines containing the components

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 4

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY

TABLE I D ROP T EST M ATRIX (N UMBER OF MEMS C OMPONENTS )

Fig. 6.

Cross section of MEMS microphone.

being tested [Fig. 3(a) and (c)], to prevent a direct impact on the MEMS components during face-down drops. The test setup amplifies the stresses not only due to enhancement of the dynamic deformation modes of the PWB, but also by creating dynamic contact stress waves (due to secondary impact) that propagate through the test specimen. Therefore, stresses in the MEMS assemblies are amplified well beyond those experienced during typical life-cycle conditions or during traditional qualification test conditions, thus offering an opportunity for highly accelerated stress testing. B. Test Vehicle and Test Matrix The test board is designed as per JEDEC standard JESD22-B111 [14]. Fig. 5 shows the test board configuration, placement of the MEMS package, and the strain gage location for calibration of damping parameters. Unlike in JESD22-B111, in which PWAs are mounted on the fixture by four or six screws, in this paper the two short edges of the PWA are completely clamped, with a 71-mm span between them. This fixture design minimizes bending of the PWB along the x-axis, so that all the MEMS components on each PWA can experience a similar motion, thus providing a larger number of replicates from each drop, and enabling better statistical assessment of the failure data. Another benefit of this fixturing method is to simplify the stress state because of the unidirectional bending mode it generates. As shown later in this paper, this simpler stress state simplifies the simulation and analysis, thereby facilitating a more fundamental mechanistic understanding of the relative contributions from bending strain, contact force, and inertia of the component. Further studies are recommended before these results can be applied to actual applications where the PWB may have more complex boundary conditions leading to bidirectional flexural curvatures. The test specimen contains an assembly of functional commercial off-the-shelf MEMS microphone (Fig. 6) on the PWB. All the PWBs discussed in this paper have the same thickness of 1 mm. The microphone package consists of a polysilicon MEMS microphone device, and a globtop application-specific integrated circuit (ASIC) for signal processing, both mounted on an organic PWB substrate with die attach, and interconnected together with multiple wire bonds. The entire structure is covered with a brass lid that is soldered with SAC105 to the substrate. The multilayered

substrate with copper pads in the bottom surface is soldered to matching copper pads on the PWB, using SAC105. The MEMS microphone device itself, shown in Fig. 6, has a polysilicon substrate. On the top of the substrate, there is a floating diaphragm and a circular back plate, separated by an annular spacer. The floating diaphragm is connected to the substrate by a cantilever-type extension, called a diaphragm runner. A corresponding structure, extended from the back plate, covers the diaphragm runner and is called the back plate runner. The test plan, shown in Table I, is designed to compare the effect of placement orientation of the microphone packages and the effect of secondary impacts between the PWB and the test fixture base. The total number of tested MEMS components at each drop test condition (acceleration, clearance, and package facing direction) are listed in the drop test matrix presented in Table I. While this test matrix provides very important fundamental understanding of the damaging effects of a direct impact of PWB near the MEMS footprint, additional stress states can be explored using other drop orientations and by varying the location of the MEMS relative to the impact site on the PWB. These added variables will result in a significant increase of the test matrix and are considered beyond the scope of the current work. Such parametric investigations are therefore deferred to the future work. C. Testing and Failure Analysis Using the fixture design shown in Fig. 3, PWAs are tested at various clearances from the base fixture, with components facing either downward or upward. All the tests and failure analysis results in this paper are from an input acceleration profile of 20 000 g magnitude and 0.05 ms width. The impact pulse profile measured on the DMSA is always identical to Fig. 2, with high drop-to-drop repeatability. Each MEMS component is functionally checked after every 25 drops. The failure analysis method consists of X-ray inspection, followed by delidding and microscopic inspection of the package. X-ray is used to detect wire-bond failures and other internal damage modes while testing. However, the lid has to be removed to inspect the internal structures for damage. Delidding is accomplished either mechanically or through heating and desoldering. In [25], the dominant failure sites under these highly accelerated test conditions were identified to be either in the MEMS package (wire-bond breakage/fracture, die attach delamination, fracture of the soldered package lid-seal, delamination of components soldered on PWB), or in the MEMS microphone device itself.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. MENG et al.: MEMS PACKAGING RELIABILITY IN BOARD LEVEL DROP TESTS PART I

5

Fig. 8. Percentage of failure sites for all 20 000 g tests with component oriented downward.

Fig. 7.

Common drop test failure sites in MEMS package.

III. R ESULTS AND D ISCUSSION In this section, we discuss the drop durability and the dominant failure sites of MEMS microphones assembled on PWAs at 20 000-g drop tests, with and without secondary impact. A. Drop Test Failure Sites of MEMS Components According to the failure analysis results in [25], multiple failure sites were observed in MEMS microphone components after drop tests. Dominant failure sites can be grouped into four types, as summarized in Fig. 7: 1) solder fracture; 2) die attach delamination (MEMS detachment from microphone package substrate); 3) wire bond fracture; and 4) MEMS device failure. Typical MEMS device failure sites include diaphragm, back plate, diaphragm runner, and back plate runner. Excessive movement of the diaphragm can fracture the diaphragm and the diaphragm runner. Additionally, being very thin and brittle, collision between the diaphragm and the back plate provides another source of failure in the MEMS device. The observed failure sites in the tested MEMS devices are in good agreement with a recent drop test study for MEMS microphone device by Li et al. [36]. Other than MEMS device failures, all the other three types of failure sites in Fig. 7 are related to MEMS package. Wire bond fracture [Fig. 7(c)] is most commonly seen in the ball bond of the gold wires connecting the MEMS device and ASIC chips and in the wedge bond of the wires between ASIC chip and package substrate. Solder fracture [Fig. 7(a)] mainly refers to solder fractures at two locations of solder joints: One occurs between the brass lid and the substrate, and the other (second level) is located between the microphone module and the PWB. Fig. 7(a) shows the first case, where the brass lid is already detached because of solder fracture. Fig. 7(b) shows the top of the package substrate, with the MEMS device (IC) completely detached. The die footprint is discernible because of the residue of the die attach material. In this case, die attach delamination occurs between the MEMS device and the substrate. Delamination of MEMS from package substrate sometimes can also induce secondary failures at other sites (i.e., fracture of the wires connecting the MEMS and ASIC chips). Such potential interactions

between the detected failure sites are not further analyzed. When multiple failure sites are found in one test sample, all the failure sites are recorded. Fig. 7(b) also shows an audio port located at the center of the MEMS device footprint. Since the MEMS microphone is used in a nonhermetic environment, air pressure through the audio port is believed to be another source of damage in MEMS structure during the secondary impact. Further investigation regarding the effect of air pressure was reported by Li et al. [36]. PWB-related failures (Cu trace crack and crack in PWB) were less relevant to the MEMS component failure modes for the purposes of this paper, so they are grouped as “others” in Fig. 10. As Tilmans et al. [32] pointed out, failures in MEMS packages are as important as failures in the MEMS device in hampering the expected performance and reliability of MEMS. A quantitative comparison between MEMS device failures and MEMS package-related failures is presented in Fig. 8, based on the results from downward-oriented MEMS components. The result shows that only less than a quarter of the functional failures in MEMS microphone were caused by MEMS device failures [Fig. 7(d)], the rest are due to damage in the MEMS package. The percentage of each failure site varies with the clearance and component orientation (facing upward or downward). Either orientation may lead to failure site distributions that are different from the distribution in Fig. 8. In order to investigate the variation in failure site distribution as a function of the clearance and component orientation, the percentage of each failure site is further analyzed at each clearance level, as shown in Figs. 9 and 10. Fig. 9 shows the detailed percentage of failure site distribution at each clearance, when components are oriented downward. In general, each failure site is found to have a nonmonotonic dependence on the clearance amount. For example, the percentage of damage in the MEMS device is the highest at 40% and the lowest at 100% of PWB thickness, whereas a totally opposite trend is observed for solder fractures. Noticeably, in all the test conditions, there is always at least one MEMS package failure site more likely to occur than failures of the MEMS device. Fig. 10 presents the distribution of failure sites at all clearances, with components facing upward. Different from Fig. 9, wire bond fracture in Fig. 10 stands out as the only dominant failure site, independent of clearances.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 6

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY

TABLE II D OMINANT FAILURE S ITE OF MEMS M ICROPHONE C OMPONENT AT E ACH D ROP T EST C ONDITION

Fig. 9. Failure site distribution for different clearances, with components oriented down.

improving the packaging design to reduce the overall failure risks of the MEMS component. If the risk of secondary impact is found late in the product design phase, the module level vulnerability information shown in Figs. 9 and 10 could also help to prioritize process optimization/ruggedization efforts for risk containment. In conventional board level drop tests without secondary impacts, acceleration–strain correlations are generally monotonic. By contrast, drop test with secondary impact is more complex. The finite clearance not only generates an additional amplification in acceleration, but also invokes participation of multiple PWA flexural modes [13]. Such multimode interaction makes understanding the variations of acceleration and strain at each failure site very challenging. Moreover, failure sites with different natural frequencies may show different sensitivities to all the dynamic inputs above. In order to explain the observations in Table II, Part II of this paper focuses on the dependence of damage on the local dynamic mechanical response at each failure site. B. Characteristic Life of Each Failure Site

Fig. 10. Failure site distribution for different clearances, with components oriented up.

Based on the analyses in Figs. 9 and 10, the phenomenon of failure site transition is summarized in Table II. When MEMS components are oriented upward, wire bond failure occurs most frequently; when MEMS components are facing downward, the dominant failure site is observed to transition from wire bond fracture at clearances 0%–40% of PWB thickness h, to die attach delamination at 40%–80% h and to solder fracture at 80%–120% h. The dominant failure site is found to depend on the test condition, with several crossovers, as seen in Figs. 9 and 10. This information is valuable for

Secondary impact drop test results for MEMS assemblies at 20 000 g were partially reported earlier [25], [37]. However, since tests were conducted on a limited number of conditions and replicates, only qualitative conclusions were provided. In addition, the maximum drop count was limited to 500, when no failures were detected in some cases. Survival data were discarded in [25] and the failure distributions were based only on the data from samples that failed within 500 drops. Although this averaging method was capable of qualitatively showing the trend of drop test durability of MEMS components as a function of clearance, the actual estimated lifetimes were unavoidably underestimated. For the same reason, two test conditions (components facing up with 20% of clearance, and components facing down with no secondary impacts) provided no failure data because all survived 500 drops. The test results in this paper improved upon all the issues listed above as follows. 1) New tests were conducted up to 2000 drops when needed. 2) The numbers of test conditions and replicates at each condition are both increased significantly. 3) The expected drop counts to failure at each test condition were estimated based on both the failure data and censored data. 4) Enabled by additional replicates, lifetimes are estimated separately for each identified failure site. These improvements provide quantitative insights into the overall reliability of MEMS microphone components in severe

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. MENG et al.: MEMS PACKAGING RELIABILITY IN BOARD LEVEL DROP TESTS PART I

Fig. 12.

7

Drop durability results for MEMS assemblies.

Fig. 11. Unreliability versus time (cycles to failure) plot of tests at 20% clearance, with package facing upward. Legend shows two-parameter Weibull parameters (details discussed in the text).

drop tests with secondary impacts. Moreover, the failure site in MEMS microphone components is observed to vary as a function of the clearance magnitude. Experimentally measured durability data (Weibull characteristic life) are plotted for each test condition (clearance and component orientation). The Weibull distribution is found to do a better job of describing the failure statistics than the exponential distribution, normal distribution, and lognormal distribution. Thus the Weibull distribution is adopted for characteristic life calculations in this paper. The probability density function of the two-parameter Weibull distribution is defined as [38]        β−1 t β t β exp − (2) f (t) = η η η where η is characteristic life and β is shape parameter. The shape parameter and characteristic life for each test condition are both determined by fitting the drops-to-failure data to the two-parameter Weibull model. A sample plot for 20% h clearance test, with packages facing upward, is plotted using Reliasoft Weibull++, as shown in Fig. 11. The blue line is the Weibull probability plot, whereas the two red curves are the upper and lower bonds of the two-side 90% confidence interval. Because some censored data were used for calculation, maximum likelihood is selected as the parameter estimation method. In this example, based on the failure data obtained from 18 samples (including five censored data, shown in Fig. 11), the Weibull shape parameter β is equal to 1.98, and the Weibull characteristic life η is 999. With a 90% confidence interval, the upper bound for η reaches 1261 and the lower bound is at 791. Mean-time-to failure is equal to 885. Fig. 12 summarizes the Weibull analysis results at all the test conditions. For demonstration purpose, two-sided 90% confidence bounds are added to the two drop orientations at 20% h clearance. It is clear that the finite clearance drop tests are significantly more severe than in the infinite clearance cases. Furthermore, drop durability of PWAs tested with components facing upward is higher than with those components facing downward. According to Fig. 12, the durability of

Fig. 13. Drop durability results by failure sites, with packages facing downward.

MEMS microphone decreases as the finite clearance between the PWA and the fixture increases. This is true for components oriented both downward and upward. The Weibull characteristic life is estimated for failures at each observed failure site. Failure mode is identified for each detected failure by conducting immediate failure analysis; the characteristic life of each failure site can be plotted based on all the available data for the given failure site. Fig. 13 lists the durability at each failure site and clearance with components facing downward. Multiple failure sites are observed at some test conditions (as shown in Fig. 9). Therefore, the characteristic life values shown in Fig. 13 are expected to have a larger scatter than in Fig. 12. As a result, when too few failure data are available for life estimation, data point will be missing (for example, MEMS device failures with large clearances). Ideally, the failure site with the lowest characteristic life at each clearance is supposed to be the most vulnerable (dominant) failure site. Following this pattern, small discrepancies (i.e., at 0% and 60% of PWB thickness) can be found between the results in Fig. 13 and Table II. Such discrepancies reflect the qualitative and probabilistic nature of the effect of failure site transition, which can be further improved with additional test results. Overall, the results shown in Fig. 13 are compatible with the information in Table II, when taken in the context of the sources of variabilities discussed earlier. IV. C ONCLUSION In this highly accelerated drop test study, secondary impact between test board and fixture is found to significantly amplify

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 8

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY

the impact, even with a fully supported, tightly clamped PWA, as indicated by the drop durability results of the MEMS microphone. Therefore, secondary impacts on PCBAs should be avoided in design (for example, increase the stiffness of the PCB or increase the clearance). The durability of MEMS microphone decreases as the clearance between the PWA and the fixture increases. If finite clearance could not be avoided, the smaller the clearance, the better to minimize the impact from secondary impacts (within the range 0–1.2 h). At any given clearance, the drop durability of PWAs tested with components facing upward is higher than that of PWAs with components facing downward. This indicates that the inertial force due to component weight plays a significant role in failures. From this perspective, avoiding secondary impacts is more important on the populated side of the PWB than on the unpopulated side. The majority of functional failures in the tested MEMS microphone components are caused by failure sites in the MEMS package, including wire bond fracture, solder fracture, and die attach delamination. Less than a quarter of all failures are due to MEMS device failures. Furthermore, the failure site is observed to change, with changes in clearances for secondary impacts. Participation of multiple bending modes could contribute to the rapid change in the dominant failure site. Using suitable statistical distribution functions, the characteristic life is estimated for each failure site. The calculated characteristic life data for each failure site provide additional qualitative verification of the failure site transition. In sum, the value of this research is that it provides: 1) quantitative information about the extent to which drop damage is exacerbated by secondary impact on the PWB; 2) insights into which failure modes dominate at different test conditions; and 3) design considerations in order to contain the risks from secondary impacts. ACKNOWLEDGMENT The authors would like to thank Microsoft Mobile Oy (Finland) and the Lansmont Corporation for their assistance in the drop testing and research. They would also like to thank the members of the Center for Advanced Life Cycle Engineering Electronic Products and System Consortium (CALCE-EPSC) at the University of Maryland for sponsoring this paper. R EFERENCES [1] D. M. Tanner et al., “MEMS reliability in shock environments,” in Proc. 38th Annu. IEEE Int. Rel. Phys. Symp., Apr. 2000, pp. 129–138. [2] M. Tariq Jan, N. Hisham Bin Hamid, M. H. Md Khir, K. Ashraf, and M. Shoaib, “Reliability and fatigue analysis in cantilever-based MEMS devices operating in harsh environments,” J. Quality Rel. Eng., vol. 2014, p. e987847, Jan. 2014. [3] G. Li and F. Shemansky, Jr., “Drop test and analysis on micro-machined structures,” Sens. Actuators A, Phys., vol. 85, nos. 1–3, pp. 280–286, Aug. 2000. [4] V. T. Srikar and S. D. Senturia, “The reliability of microelectromechanical systems (MEMS) in shock environments,” J. Microelectromech. Syst., vol. 11, no. 3, pp. 206–214, Jun. 2002. [5] T. Hauck, G. Li, A. McNeill, H. Knoll, M. Ebert, and J. Bagdahn, “Drop simulation and stress analysis of MEMS devices,” in Proc. 7th Int. Conf. Thermal Mech. Multiphysics Simul. Experiments Micro-Electron. Micro-Syst. EuroSime, Apr. 2006, pp. 1–5.

[6] S. Goyal, S. Upasani, and D. M. Patel, “Improving impact tolerance of portable electronic products: Case study of cellular phones,” Experim. Mech., vol. 39, no. 1, pp. 43–52, Mar. 1999. [7] G. Kim and J. Choe, “Study on measurement free fall posture and height of a mobile device using MEMS sensors,” in Proc. IEEE Int. Conf. Adv. Intell. Mechatron. (AIM), Jul. 2015, pp. 1398–1401. [8] C. T. Lim and Y. J. Low, “Investigating the drop impact of portable electronic products,” in Proc. 52nd Electron. Compon. Technol. Conf., May 2002, pp. 1270–1274. [9] E. H. Wong, S. K. W. Seah, W. D. van Driel, J. F. J. M. Caers, N. Owens, and Y.-S. Lai, “Advances in the drop-impact reliability of solder joints for mobile applications,” Microelectron. Rel., vol. 49, no. 2, pp. 139–149, Feb. 2009. [10] J. Karppinen, J. Li, J. Pakarinen, T. T. Mattila, and M. Paulasto-Kröckel, “Shock impact reliability characterization of a handheld product in accelerated tests and use environment,” Microelectron. Rel., vol. 52, no. 1, pp. 190–198, Jan. 2012. [11] T. T. Mattila, L. Vajavaara, J. Hokka, E. Hussa, M. Mäkelä, and V. Halkola, “An approach to board-level drop reliability evaluation with improved correlation with use conditions,” in Proc. 63rd IEEE Electron. Compon. Technol. Conf. (ECTC), May 2013, pp. 1259–1268. [12] C. T. Lim, C. W. Ang, L. B. Tan, S. K. W. Seah, and E. H. Wong, “Drop impact survey of portable electronic products,” in Proc. 53rd Electron. Compon. Technol. Conf., May 2003, pp. 113–120. [13] S. T. Douglas, M. Al-Bassyiouni, and A. Dasgupta, “Experiment and simulation of board level drop tests with intentional board slap at high impact accelerations,” IEEE Trans. Compon., Packag., Manuf. Technol., vol. 4, no. 4, pp. 569–580, Apr. 2014. [14] Board Level Drop Test Method of Components for Handheld Electronic Products, document JESD22-B111, JEDEC Solid State Technology Association, Arlington, VA, USA, Jul. 2003. [15] Y. C. Ong, V. P. W. Shim, T. C. Chai, and C. T. Lim, “Comparison of mechanical response of PCBs subjected to product-level and boardlevel drop impact tests,” in Proc. 5th Electron. Packag. Technol. Conf. (EPTC), Dec. 2003, pp. 223–227. [16] S. K. W. Seah, E. H. Wong, Y. W. Mai, R. Rajoo, and C. T. Lim, “Failure mechanisms of interconnections in drop impact,” in Proc. 56th Electron. Compon. Technol. Conf., Jun. 2006, pp. 1–9. [17] J. Varghese and A. Dasgupta, “Test methodology for durability estimation of surface mount interconnects under drop testing conditions,” Microelectron. Rel., vol. 47, no. 1, pp. 93–103, Jan. 2007. [18] J. Varghese and A. Dasgupta, “An experimental approach to characterize rate-dependent failure envelopes and failure site transitions in surface mount assemblies,” Microelectron. Rel., vol. 47, no. 7, pp. 1095–1102, Jul. 2007. [19] J. Meng and A. Dasgupta, “Influence of secondary impact on printed wiring assemblies—Part I: High frequency ‘breathing mode’ deformations in the printed wiring board,” J. Electron. Packag., vol. 138, no. 1, pp. 1–12, Jan. 2016. [20] J. B. Hart and R. B. Herrmann, “Energy transfer in one-dimensional collisions of many objects,” Amer. J. Phys., vol. 36, no. 1, pp. 46–48, Jan. 1968. [21] W. G. Harter, “Velocity amplification in collision experiments involving superballs,” Amer. J. Phys., vol. 39, no. 6, pp. 656–663, Jun. 1971. [22] J. D. Kerwin, “Velocity, momentum, and energy transmissions in chain collisions,” Amer. J. Phys., vol. 40, no. 8, pp. 1152–1158, Aug. 1972. [23] B. Rodgers, S. Goyal, G. Kelly, and M. Sheehy, “The dynamics of multiple pair-wise collisions in a chain for designing optimal shock amplifiers,” Shock Vibrat., vol. 16, no. 1, pp. 99–116, 2009. [24] G. Kelly, J. Punch, S. Goyal, and M. Sheehy, “Shock pulse shaping in a small-form factor velocity amplifier,” Shock Vibrat., vol. 17, no. 6, pp. 787–802, 2010. [25] J. Meng et al., “Drop qualification of MEMS components in handheld electronics at extremely high accelerations,” in Proc. 13th IEEE Intersociety Conf. Thermal Thermomechanical Phenomena Electron. Syst. (ITherm), Jun. 2012, pp. 1020–1027. [26] M. Sheehy, M. Reid, J. Punch, S. Goyal, and G. Kelly, “The failure mechanisms of micro-scale cantilevers in shock and vibration stimuli,” in Proc. Symp. Design, Test, Integr. Packag. MEMS/MOEMS, 2008, pp. 2–7. [27] S. M. Kaplan, R. B. Greendyke, R. D. Lowe, and J. R. Foley, “Measurement of the electromechanical response of capacitors in dynamic loading conditions,” in MEMS and Nanotechnology, vol. 5, B. C. Prorok and L. Starman, Eds. Costa Mesa, CA, USA: Springer, 2016, pp. 97–108. [28] W. M. Van Spengen, “MEMS reliability from a failure mechanisms perspective,” Microelectron. Rel., vol. 43, no. 7, pp. 1049–1060, Jul. 2003.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. MENG et al.: MEMS PACKAGING RELIABILITY IN BOARD LEVEL DROP TESTS PART I

[29] J. Li, M. Broas, J. Makkonen, T. T. Mattila, J. Hokka, and M. Paulasto-Krockel, “Shock impact reliability and failure analysis of a three-axis MEMS gyroscope,” J. Microelectromech. Syst., vol. 23, no. 2, pp. 347–355, Apr. 2014. [30] P. Lall, N. Kothari, and J. Glover, “Mechanical shock reliability analysis and multiphysics modeling of MEMS accelerometers in harsh environments,” in Proc. ASME InterPACK, Jul. 2015, pp. 1–9. [31] F. Alsaleem, M. I. Younis, and R. Miles, “An investigation into the effect of the PCB motion on the dynamic response of MEMS devices under mechanical shock loads,” J. Electron. Packag., vol. 130, no. 3, p. 031002, Jul. 2008. [32] H. A. C. Tilmans et al., “MEMS packaging and reliability: An undividable couple,” Microelectron. Rel., vol. 52, nos. 9–10, pp. 2228–2234, Sep. 2012. [33] S. Douglas, M. Al-Bassyiouni, and A. Dasgupta, “Simulation of drop testing at extremely high accelerations,” in Proc. 11th Int. Conf. Thermal Mech. Multi-Phys. Simul. Experiments Microelectron. Microsystems (EuroSimE), 2010, pp. 1–7. [34] P. Lall et al., “Modeling and reliability characterization of area-array electronics subjected to high-g mechanical shock up to 50,000g,” in Proc. 62nd IEEE Electron. Compon. Technol. Conf. (ECTC), Jun. 2012, pp. 1194–1204. [35] S. T. Douglas, “High accelerations produced through secondary impact and its effect on reliability of printed wiring assemblies,” M.S. thesis, Univ. Maryland, College Park, MD, USA, 2010. [36] J. Li, M. Broas, J. Raami, T. T. Mattila, and M. Paulasto-Kröckel, “Reliability assessment of a MEMS microphone under mixed flowing gas environment and shock impact loading,” Microelectron. Rel., vol. 54, nos. 6–7, pp. 1228–1234, Jun. 2014. [37] S. Douglas, J. Meng, J. Akman, I. Yildiz, M. Al-Bassyiouni, and A. Dasgupta, “The effect of secondary impacts on PWB-level drop tests at high impact accelerations,” in Proc. 12th Int. Conf. Thermal Mech. Multi-Phys. Simul. Experiments Microelectron. Microsystems (EuroSimE), Apr. 2011, pp. 1-6–1-6. [38] J. W. McPherson, Reliability Physics and Engineering: Time-To-Failure Modeling. New York, NY, USA: Springer, 2010.

Jingshi Meng received the B.S. degree in materials science from the Dalian University of Technology, Dalian, China, and the M.S. and Ph.D. degrees in mechanical engineering from the University of Maryland, College Park, MD, USA. He was a Graduate Research Assistant with the Center for Advanced Life Cycle Engineering, University of Maryland, where he was involved in accelerated drop, shock and vibration testing, and physics of failure modeling of the interconnects in electronic packaging. He is currently a Reliability Engineer with Apple, Cupertino, CA, USA. Dr. Meng has been a Reviewer for the Journal of Microelectronics Reliability and the Journal of Electronic Packaging.

9

Stuart T. Douglas received the B.S. degree in aerospace engineering and the M.S. degree in mechanical engineering from the Center for Advanced Life Cycle Engineering, University of Maryland, College Park, MD, USA. He was involved in several generations of Kindle E-readers and tablets and led analysis into the reliability of glass displays at Lab126, Sunnyvale, CA, USA. He is currently a Senior Reliability Engineer with Google. His academic research has been featured in several journals and conferences. He has supported projects such as self-driving car, wing, glass, makani, and verily life sciences. His current research interests include the test and analysis of MEMS and electronic packages, brittle materials and thin films, and large mechanical systems.

Abhijit Dasgupta is currently the Jeong H. Kim Professor of Mechanical Engineering with the University of Maryland at College Park, College Park, MD, USA, where he is a Founding Investigator with the Center for Advanced Life Cycle Engineering, and conducts research on the mechanics of engineered, heterogeneous, active materials, with a special emphasis on the micromechanics of constitutive and damage behavior. He applies his expertise to several multifunctional material systems, including electronic packaging material systems and smart composite material systems. He applies these principles for developing effective virtual qualification tools, optimizing manufacturing process windows, real-time health monitoring, and devising quantitative accelerated testing strategies used in qualification and the quality assurance of complex electronic, electromechanical, and structural systems. He has published over 200 journal articles and conference papers on these topics and presented over 35 short workshops nationally and internationally. Prof. Dasgupta is an ASME Fellow, and was the Chair of the ASME Electrical and Photonics Packaging Division. He has served on the Editorial Boards of three different international journals, and organized several national and international conferences. He was a recipient of six awards for his contributions to materials engineering research and education.