Academia.eduAcademia.edu

Figure 3 – uploaded by Faxing Che

See full PDFdownloadDownload figure
Fig. 3 Nonlinear relationship between G;, and Gou

Figure 3 Nonlinear relationship between G;, and Gou

Related Figures (16)

Fig. | FCOB assembly schematics and chip number Test vehicle and setup: Flip chip is suitable interconnect technology for high I/O and performance microelectronics devices. Flip chip packaging can cater to the increasing demands on miniaturization, high density of I/O channels, high performance, high reliability. So in this study, flip chip on board (FCOB) assembly was selected for vibration test to analyze the dynamic response of the FCOB to sinusoidal vibration. The specimen and chip number is shown in Fig. 1. Six larger flip chip modules of 8.5mmx8.5mmx0.65mm (LxWxH) silicon die with 388 I/Os and six smaller flip chip modules of 3.5mmx3.5mmx0.65mm(LxWxH) silicon die with 48 I/Os were mounted on the PWB assembly which is made of FR-4 of 185mmx1i50mmx1.13mm(LxWxH) in size. The solder joints are eutectic 63Sn/37Pb solder with a diameter of 0.16mm, ball pitch of 0.35mm and standoff of height of 0.1mm. A total of 4 test boards were tested to different acceleration levels to assess the reliability of FCOB assembly subjecting to constant G-level and block G-level loading. Board 1 for varying amplitude test including 3G, 5G and 10G (G is gravity acceleration) blocks, Board 2, 3, 4 for constant 3G, 5G, 10G level vibration test, respectively.
Fig. | FCOB assembly schematics and chip number Test vehicle and setup: Flip chip is suitable interconnect technology for high I/O and performance microelectronics devices. Flip chip packaging can cater to the increasing demands on miniaturization, high density of I/O channels, high performance, high reliability. So in this study, flip chip on board (FCOB) assembly was selected for vibration test to analyze the dynamic response of the FCOB to sinusoidal vibration. The specimen and chip number is shown in Fig. 1. Six larger flip chip modules of 8.5mmx8.5mmx0.65mm (LxWxH) silicon die with 388 I/Os and six smaller flip chip modules of 3.5mmx3.5mmx0.65mm(LxWxH) silicon die with 48 I/Os were mounted on the PWB assembly which is made of FR-4 of 185mmx1i50mmx1.13mm(LxWxH) in size. The solder joints are eutectic 63Sn/37Pb solder with a diameter of 0.16mm, ball pitch of 0.35mm and standoff of height of 0.1mm. A total of 4 test boards were tested to different acceleration levels to assess the reliability of FCOB assembly subjecting to constant G-level and block G-level loading. Board 1 for varying amplitude test including 3G, 5G and 10G (G is gravity acceleration) blocks, Board 2, 3, 4 for constant 3G, 5G, 10G level vibration test, respectively.
Table I Weibull parameters, MTTF and FTTF
Table I Weibull parameters, MTTF and FTTF
accelerometers were used in the test to determine the dynamic response of the specimen subjected to vibration, one for fixture and the other for PWB, thus the vibration transmissibility of the PWB can be determined when evaluated at different acceleration levels. In order to minimize the effect of the accelerometer’s mass on the specimen, a miniature accelerometer (2 grams of mass) was adopted. The specimen was clamped along two long opposite sides using aluminum fixture, which was bolted to the shaker header. This gives a clamped-clamped boundary condition for the specimen. The resistance change of the connecting loops to the solder joints was used to determine the failure of the solder joints. There are three loops for each larger flip chips except for chip6 with outer, middle and inner loops. There is only one outer loop for each smaller and chip6. Total 22 loops were monitored simultaneously by the monitoring device called the Event Detector during vibration test. The threshold of the event detector was preset before the test. Any resistance change exceeding a preset threshold with minimum duration of 0.1us can be detected by the Event Detector. In general, 50% increase of resistance was used as a rule to determine the threshold setting. Fig. 2 shows the setup of the vibration test. Natural Frequency Scanning Test: The resonance frequency-scanning test was conducted first with sweep sine from 20Hz to 1000Hz. The first order natural frequency of 194Hz is dermined by scaning test. The transmissibility of the specimen can be estimated using the following formula
accelerometers were used in the test to determine the dynamic response of the specimen subjected to vibration, one for fixture and the other for PWB, thus the vibration transmissibility of the PWB can be determined when evaluated at different acceleration levels. In order to minimize the effect of the accelerometer’s mass on the specimen, a miniature accelerometer (2 grams of mass) was adopted. The specimen was clamped along two long opposite sides using aluminum fixture, which was bolted to the shaker header. This gives a clamped-clamped boundary condition for the specimen. The resistance change of the connecting loops to the solder joints was used to determine the failure of the solder joints. There are three loops for each larger flip chips except for chip6 with outer, middle and inner loops. There is only one outer loop for each smaller and chip6. Total 22 loops were monitored simultaneously by the monitoring device called the Event Detector during vibration test. The threshold of the event detector was preset before the test. Any resistance change exceeding a preset threshold with minimum duration of 0.1us can be detected by the Event Detector. In general, 50% increase of resistance was used as a rule to determine the threshold setting. Fig. 2 shows the setup of the vibration test. Natural Frequency Scanning Test: The resonance frequency-scanning test was conducted first with sweep sine from 20Hz to 1000Hz. The first order natural frequency of 194Hz is dermined by scaning test. The transmissibility of the specimen can be estimated using the following formula
From test results, a plot of G-level versus fatigue life can be obtained as shown in Fig. 5. In log-log scale, the liner relationship is clear. The fatigue model can be obtained based on Gin-N curve for FCOB assembly used in this study as expressed followings
From test results, a plot of G-level versus fatigue life can be obtained as shown in Fig. 5. In log-log scale, the liner relationship is clear. The fatigue model can be obtained based on Gin-N curve for FCOB assembly used in this study as expressed followings
Fig. 5 Plot of input G-level versus fatigue life Where A is input G, N is fatigue life. Therefore, Eq.(4) can be used to predict fatigue life of FCOB assemblies subjected to different G-level input vibration tests. The high speed camera with ability to record upto 4500 full frames persecond was used to capture the dynamic response of specimen. In this test, the displacement was measured at the edge of PWB. Some markers were placed on the PWB edge to allow the motion analysis software to recognise the pixel easily, as shown in Fig, 6. The mark on the fixture is selected as reference point, then the relative displacement on the PWB can be obtained for different G- level tests. Fig. 7 shows the plot of vibration displacement versus time for mark 2 on PWB at 10G test. according to refence [1], the maximum displacement can be determined ac
Fig. 5 Plot of input G-level versus fatigue life Where A is input G, N is fatigue life. Therefore, Eq.(4) can be used to predict fatigue life of FCOB assemblies subjected to different G-level input vibration tests. The high speed camera with ability to record upto 4500 full frames persecond was used to capture the dynamic response of specimen. In this test, the displacement was measured at the edge of PWB. Some markers were placed on the PWB edge to allow the motion analysis software to recognise the pixel easily, as shown in Fig, 6. The mark on the fixture is selected as reference point, then the relative displacement on the PWB can be obtained for different G- level tests. Fig. 7 shows the plot of vibration displacement versus time for mark 2 on PWB at 10G test. according to refence [1], the maximum displacement can be determined ac
Cumulative Damage Index Fatigue Analysis: For the electronic assembly subjected to different blocks of acceleration level (G-level) vibration test, the fatigue damage due to each acceleration level can be superimposed using the linear superposition method by Miner’s law [13]. Miner’s cumulative fatigue damage ratio is based on an analysis method that sums up the ratio of the actual number of fatigue cycles (7) accumulated in a specific element, in different environments, divided by the number of fatigue cycles (N) required to produce a fatigue failure in the same specific element in the same environment. When the ratios are added together, a sum of 1.0 or greater means that the fatigue life has been used up and should fail. The cumulative damage index (CDI) using Miner’s cumulative damage law, assuming a linear summation given by
Cumulative Damage Index Fatigue Analysis: For the electronic assembly subjected to different blocks of acceleration level (G-level) vibration test, the fatigue damage due to each acceleration level can be superimposed using the linear superposition method by Miner’s law [13]. Miner’s cumulative fatigue damage ratio is based on an analysis method that sums up the ratio of the actual number of fatigue cycles (7) accumulated in a specific element, in different environments, divided by the number of fatigue cycles (N) required to produce a fatigue failure in the same specific element in the same environment. When the ratios are added together, a sum of 1.0 or greater means that the fatigue life has been used up and should fail. The cumulative damage index (CDI) using Miner’s cumulative damage law, assuming a linear summation given by
Table II Natural frequency for bear PWB modal analysis underfill. In case 4, the IC chips were modeled as shell and the solder joints were modeled as effective two-node beam elements with equivalent stiffness of solder joint [6]. According to Steinberg [1], the fundamental resonance frequency for a PWB with the Clamped-Clamped boundary condition can be obtained from the following equation
Table II Natural frequency for bear PWB modal analysis underfill. In case 4, the IC chips were modeled as shell and the solder joints were modeled as effective two-node beam elements with equivalent stiffness of solder joint [6]. According to Steinberg [1], the fundamental resonance frequency for a PWB with the Clamped-Clamped boundary condition can be obtained from the following equation
Table IV Comparison of the natural frequencies results Quasi-static analysis: In this study, a quasi-static analysis method was developed to calculate the stress strain behavior of the solder joints, which can be used for fatigue life prediction using high cycle fatigue model. The dynamic loading due to vibration was replaced by effective static loading in this method. According to Newton’s second law, the pressure acting on the PWB or component can be obtained
Table IV Comparison of the natural frequencies results Quasi-static analysis: In this study, a quasi-static analysis method was developed to calculate the stress strain behavior of the solder joints, which can be used for fatigue life prediction using high cycle fatigue model. The dynamic loading due to vibration was replaced by effective static loading in this method. According to Newton’s second law, the pressure acting on the PWB or component can be obtained
Fig. 11 Schematic of the location of the measured point For constant G level test, the pressure loading can be obtained from Eq. (1) and Eq. (8) when transmissibility is known. It is observed that transmissibility is not a constant value along transverse locations. The transmissibility was measured only for the half of the PWB due to symmetry. Eight locations uniformly distributed shown in the Fig. 11 were selected to measure transmissibility of the PWB subjected to vibration. Fig. 12 shows the transmissibility result for different location at fundamental resonance frequency for 10G input test and a linear relationship between transmissibility and location is clear.
Fig. 11 Schematic of the location of the measured point For constant G level test, the pressure loading can be obtained from Eq. (1) and Eq. (8) when transmissibility is known. It is observed that transmissibility is not a constant value along transverse locations. The transmissibility was measured only for the half of the PWB due to symmetry. Eight locations uniformly distributed shown in the Fig. 11 were selected to measure transmissibility of the PWB subjected to vibration. Fig. 12 shows the transmissibility result for different location at fundamental resonance frequency for 10G input test and a linear relationship between transmissibility and location is clear.
Fig.12 Transmissibility at natural frequency for 10G test
Fig.12 Transmissibility at natural frequency for 10G test
b can be determined by curve fitting shown in Fig.14 using testing data by Yao et al. [14] and they are 177.15MPa and —0.2427, respectively. For multi-axes stress situation, the equivalent von: Mises stress is used to output the state of stress in the solder joint. In order to minimize the effect of stress concentration, the volume-weight method was used to calculate the equivalent von Mises stress for the die/solder interface layer
b can be determined by curve fitting shown in Fig.14 using testing data by Yao et al. [14] and they are 177.15MPa and —0.2427, respectively. For multi-axes stress situation, the equivalent von: Mises stress is used to output the state of stress in the solder joint. In order to minimize the effect of stress concentration, the volume-weight method was used to calculate the equivalent von Mises stress for the die/solder interface layer
Fig. 15 Transmissibility Vs frquency for 3G test Then fatigue life can be determined by Eq. (9) when stress amplitude was known. The frequency range from 175Hz to 215Hz was used for vibration test. Different transmissibility occur at different frequencies as shown in Fig. 15. Therefore, specimen was sujected to different effective pressures with different slop triangle distribution at utmost location during test. Miner’s law shown in Eq.(6) was used and fatigue life can be predicted combining with Eq. (9). In this study, CDI of 0.5 was used. The fatigue life result comparisons between quasi-static method and testing for chip3 are listed in Table V. It can be seen that the quasi- static can give reasonble fatigue life pridiction result compared to testing result, especially for lower G level vibration test. By imitating S-N curve, the G-N curve can be obtained based on fatigue life result for different G level test as shown in Fig. 16. Different CDI values were used to predict the flip chip solder joint fatigue life in order to investigate the CDI effect. The factor is defined as the ratio of fatigue cycles predicted by FEA to experimental fatigue cycles. Different fators corresponding for four CDI of 1, 0.7, 0.5, and 0.33 were calculated and shown in Fig. 17. It can be seen that fatigue life prediction has a good agreement with test data for 5G vibration test,
Fig. 15 Transmissibility Vs frquency for 3G test Then fatigue life can be determined by Eq. (9) when stress amplitude was known. The frequency range from 175Hz to 215Hz was used for vibration test. Different transmissibility occur at different frequencies as shown in Fig. 15. Therefore, specimen was sujected to different effective pressures with different slop triangle distribution at utmost location during test. Miner’s law shown in Eq.(6) was used and fatigue life can be predicted combining with Eq. (9). In this study, CDI of 0.5 was used. The fatigue life result comparisons between quasi-static method and testing for chip3 are listed in Table V. It can be seen that the quasi- static can give reasonble fatigue life pridiction result compared to testing result, especially for lower G level vibration test. By imitating S-N curve, the G-N curve can be obtained based on fatigue life result for different G level test as shown in Fig. 16. Different CDI values were used to predict the flip chip solder joint fatigue life in order to investigate the CDI effect. The factor is defined as the ratio of fatigue cycles predicted by FEA to experimental fatigue cycles. Different fators corresponding for four CDI of 1, 0.7, 0.5, and 0.33 were calculated and shown in Fig. 17. It can be seen that fatigue life prediction has a good agreement with test data for 5G vibration test,
Fig. 16 G-N curve for different methods
Fig. 16 G-N curve for different methods
Fig.17 CDI effect on fatigue life prediction
Fig.17 CDI effect on fatigue life prediction
Related topics:
EngineeringStructural EngineeringFinite Element Analysis (FEA)Damage AssessmentMean Time To FailureVibration and Fatigue AnalysisFatigue Lifefatigue life prediction
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved