Predictive modeling of board level shock-impact reliability of the HVQFN-family
Introduction
In the past decade, quad flat no-lead (QFN), or micro-lead frame (MLF), packages have been developed for electronics applications. As they do not have leads they occupy a relatively small area, and their low solder height results in a low self-inductance. The large heat sink gives them a very good thermal performance. The version with a very low building height is usually called HVQFN: heat sink very thin quad flat-pack no leads. But nevertheless, until recently, the board level reliability of QFN- and HVQFN-components has received little attention. Particularly the low stand-off of these packages may be a concern.
Board level reliability tests consume both a lot of time and money. Still, they are very relevant to the development of electronic packages. A methodology that must allow for predictive simulation techniques for board level reliability is under development. According to this methodology, one can select a few packages from an entire family for testing, fit the simulation to the experimental data, and use the result to estimate the lifetime of the other members by interpolation. Earlier investigations were done by using this approach to thermal cycle stress loading of BGA-packages [1] and HVQFN-packages [2]. Something similar was also done for ultra chip scale packages [3].
The past decade has seen an increasing number of publications addressing the effect of impact loading on electronic assemblies. To restrict ourselves to the more recent ones, Wong et al. [4] presented an analytical and numerical model, with qualitative material properties. Various approaches for the numerical model were compared. The use of a high-speed camera to determine acceleration and strain was demonstrated by Tee et al. [5]. The focus was on finite element analyses. This has resulted in a predictive lifetime model. At the University of Limerick, Heaslip et al. studied the distribution of failure sites of BGA-assemblies. Analytical and numerical modeling of the peel stress was compared to material properties [6]. Chai et al. considered the assembly as a plate bending under impact [7]. Basic inputs for the modeling were the acceleration and the resulting strain. As an alternative to impact testing, high-speed bending was used for a detailed study of the failure mode and crack propagation [8]. Marjamäki et al. and Mattila et al. reported on using vibration loading with high amplitude to simulate the impact test [9], [10].
In addition to the papers referenced above, work has been published specifically on the mechanical impact performance of QFN-packages. Chai et al. reported results of a shock test according to the JEDEC test standard [11]. This was done on QFN-packages mounted with SnPb- or SAC-solder on two-layer boards [7]. The difference between these boards and the recommended (JEDEC) boards, with respect to the flexural modulus and the density, was checked. The JEDEC boards’ values were 24.8 GPa and 2039 kg/m3, whereas the two-layer boards’ values were 21.0 GPa and 1899 kg/m3. Chong et al. studied amongst others VQFN48-packages that were assembled with eutectic SnPb-solder and SAC405-solder [12], [13]. The boards had a dimension of 200 × 150 mm2. Prior to the drop test they received a thermal cycling stress. The clamping was along the long edge of the boards – this is not according to the JEDEC standard. Birzer et al. report on testing QFN48-packages on JEDEC-boards [14]. One has made these with Cu-track widths of 75 μm and 150 μm where the daisy chain is made on the top layer, and a version where the daisy chain is made through the vias-in-pad. The many variations in these experiments make it clear that a comparison of the results with our own data will be difficult.
In this paper experiments and simulations on the drop-impact test on HVQFN-packages will be combined for the first time. We will discuss a set of experimental data of various HVQFN-packages subjected to such loading. A numerical model was built to calculate the solder strain, and used to interpret the results. In addition, the board strain is estimated from an analytical model, and compared to the numerical model. The outcome of the experimental and the modeling results are compared to literature data. This investigation forms part of a coherent set of experiments and modeling studies on board level reliability.
Section snippets
Experiments and simulations
HVQFN-packages with various numbers of leads were assembled on printed circuit boards with a layout that was according to the JEDEC test standard [11], except that in this case they were equipped with four Cu-layers (each 35 μm in thickness) and that they did not have vias. As mentioned in the introduction, this does not appear to severely affect the relevant board properties [7]. The board finish was Cu-OSP. Soldering was done in a reflow oven using lead free soldering paste of SAC305
Results
The measured acceleration (see Fig. 3) shows the main shock pulse and some additional features. These additional features are due to higher order eigen frequencies, in particular the fifth at around 700 Hz. They have a second order impact on the durability of the interconnection.
The failure distributions are given in Fig. 7. In Table 2, the results of the statistical analysis are compiled.
One board with medium-size packages was put through 95 drops; at that point, only two products survived.
Discussion
As we intend to make a semi-predictive model, one of the first things to do is to check for the failure mode and mechanism. From the failure analysis it is clear that two modes dominate: solder joint cracks and intermetallic-layer cracks. Other failure modes such as breaking of the Cu-tracks were not found. In some cases cracks in the solder and the intermetallic layers appear in one and the same interconnection (see Fig. 9a). This is in line with many observations in literature, where
Conclusions
In this study the board level drop-impact lifetime of HVQFN-packages was determined experimentally. A semi-empirical model that is intended to help estimate the solder joint reliability was developed by means of finite element calculations. The results can be summarized with the following conclusions.
- –
A fatigue-based model is applied to describe a mixture of that with overstress-induced failure behavior. The authors admit that this is unusual, but in practice the model agrees with the
Acknowledgements
The authors thank J. Caers, R. Engelen, M. Jansen, I. Lemmens and W. Vogels (Philips) and F. Kessels (NXP) for their valuable advice and support.
References (22)
- et al.
On the difference between thermal cycling and thermal shock testing for board level reliability of soldered interconnections
Microelectron Reliab
(2007) - et al.
Solder-joint reliability of HVQFN-packages subjected to thermal cycling
Microelectron Reliab
(2009) - et al.
Solders fatigue prediction using interfacial boundary volume criterion
J Electron Packag
(2003) - et al.
Impact life prediction modeling of TFBGA packages under board level drop test
Microelectron Reliab
(2004) - et al.
Drop impact reliability testing for lead-free and lead-based soldered IC packages
Microelectron Reliab
(2006) - et al.
Time-independent elastic–plastic behaviour of solder materials
Microelectron Reliab
(2004) - et al.
Drop impact test – mechanics & physics of failure
Proc EPTC
(2002) - et al.
A stress-life methodology for ball grid array lead-free and tin–lead solder interconnects under impact conditions
Proc EuroSime
(2005) - et al.
Board level drop test reliability of IC packages
Proc ECTC
(2005) - et al.
Failure mechanisms of interconnections in drop impact
Proc ECTC
(2005)
A comparative study of the failure mechanisms encountered in drop and large amplitude vibration tests
Proc ECTC
Cited by (2)
The reliability of component boards studied with different shock impact repetition frequencies
2012, Microelectronics ReliabilityCitation Excerpt :Again, the same failure modes were observed, regardless of the impact repetition frequency used. Similar types of failures have also been observed earlier [19]. Since the same failure modes were observed and the impact repetition frequency is the only difference causing the different average number of impacts-to-failure in Fig. 7, the root cause of the different lifetimes must be related to events that take place between the successive impacts.
Simulation Study of the High-G Drop Impact Testing of Microelectronic Packaging Products
2021, InterSociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems, ITHERM