Ball grid arrays are increasingly common in modern electrical equipment but are one of the most difficult types of components to attach to circuit boards and are prone to failure caused either by errors in the assembly process or during use as a result of environmental conditions that induce excessive stresses to solder joints.

ERA is seeing an increasing number of failures to the solder joints of ball grid arrays. Failure typically occurs under the following circumstances:

	Weak bonds when lead-free BGA balls are attached with tin/lead solder
	Excessive voids when tin/lead BGA balls are attached in lead-free soldering processes
	Cracks in bonds due to poor alignment of BGA with pads or by solder resist and pads
	Technical alternatives to the restricted substances and reliability issues
	Cracking induced by thermal cycling
	“Black pad” caused by the formation of brittle intermetallic layers
	Excessive voiding due to poor assembly process control

Some of these defects, such as voids, can be seen by X-ray examination but others including cracks are not invisible and can be seen only by careful micro-sectioning followed by scanning electron microscope examination.

Many BGAs produce heat in use and thermal cycling can cause failures due to thermal fatigue. Heat sinks are often used to reduce temperatures but may be inadequate if an incorrect design choice is made.

Case study -identifying the cause of failure In several recent failure analysis investigations ERA has been able to identify the causes of unexpected early failures and recommend corrective action. Figure 1 shows a cross section through a BGA joint at the interface with the active device. There is in fact an interfacial crack but this can be seen only if it imaged at high magnification (Figure 2).
	Figure 1 SEM image of a cross section through a BGA joint. The red box defines the area shown at higher magnification in Figure 2
This fine crack failure was due to fatigue caused by thermal cycling exacerbated by the formation of a layer of the brittle intermetallic Sn4Ni3. It was possible to calculate the theoretical number of cycles that should occur before failure. There was good agreement between the time before failure and the actual number of cycles that had occurred. Figure 2 Image of area defined by the red box in Figure 1. The interfacial crack
Further information To find out more call us on +44 (0)1372 367444 or email.