Minapad BGA package design and solder joints …

1 Minapad 2014, May 21 22th, Grenoble; France BGA package design and solder joints board level reliability Luc PETIT and Daniel YAP STMicroelectronics , Abstract BGA (Ball Grid Array) package is becoming more & more popular, it was developed first for mobile applications but now it is also penetrating other products segments, from consumer to Automotive. BGA platform is very flexible & many evolutions have been considered in term of integration & miniaturization. Improving BGA package in term of size & performance is putting a lot of challenges not only on PLR ( package Level reliability ) but also on BLR ( board Level reliability ) The different variables that can affect positively the BGA board Level reliability performances are analyzed, including package configuration, structure, material & design , solder ball layout, solder ball alloy metallurgy & substrate pad materials.

2 Practical results from different BGA packages are presented, including modeling of solder joints behavior during thermal cycling. Accurate modeling & specific tricks implementation (redundant solder balls in specific areas) are allowing an optimization of the BLR robustness for a given package in the field since the product design phase. Identification of critical solder joints can allow to optimize package balls matrix but also the functions attributed to solder balls creating dummy or redundant solder joints . Key words: BGA , solder joints , design , Thermal cycling , Drop test Introduction Using the components surface for locating the connections , rather than at the package periphery only, is a must for recent electronic components : BGA , LGA , CSP , The hidden connections , once reported on board by soldering are not only a challenge for inspection but also for the routing of signals at the minimum number of metallized layers in the associated boards (meaning at minimized cost).

3 But once mounted and soldered on board , in the application, both the board and the packages are submitted to various thermal and mechanical stresses which can impair their functionalities and which effects are conventionally tested through several standardized separated tests such as : Thermal cycling test, Drop test, Shock test, Bending test and Vibration test. The various solder joints of a BGA are submitted to different level of stress depending on their position versus the package geometry and versus the elements contained in the may have different thermo-mechanical behaviors( CTE , Young s Modulus, Tg ..) , see Figure 1 for typical failure modes under drop and under thermal cycles.

4 2 key parameters impacting the relative performance of a solder joint are : -the distance to the center of the package considered as the neutral point - the distance to any feature in the package having different stiffness . thermal expansion coefficient or mechanical properties ( die in most cases) . Typical break during drop test : cracks at interfaces Typical break during thermal cycling cracks inside bulk solder Figure 1 : solder joint reliability : typical failure modes after board level testing of solder joints In classic cases, corner joints are more stressed than central ones when noticeable difference in thermal expansion is occurring between the board and the package and when bending of board is occurring during applied stress: drop test, vibration test in direction perpendicular to board plane and bending test.

5 But the presence of silicon with very low thermal coefficient of expansion is leading also to high stress on connections close to the die corner for example . The move to Lead-free soldering alloys , with their higher melting temperatures, has lead to increased risk and number of failures for BGAs packages joints . Warning have been made quite early [1] about brittleness occurrence of solder joints and have led to huge amount of research in the industry as well as in universities and several hundreds of publications on Lead-free BGA in typically in 3 directions : - Test methods : shear tests , shock tests , thermal cycling .. - Materials Improvement : solder composition and minor (doping) elements - package and board design improvement [2] [3].

6 Materials Improvements : solder balls and pads metallurgy Material improvement has included : solder hardness and strain transmission leading to rather standardized low Silver alloys , see figure 2 [4], for mobile application considering that drop test is more critical than thermal cycling. But if low Silver alloys are behaving better than classic SAC405 SAC305 initial Lead-free alloys they usually exhibit reduced performances during thermal cycling . Some customers are expecting both good drop test performances as well as good thermal cycling performances leading to the development of new alloys. Figure 2 : 2007 review of solder balls used for handheld Electronic [ 4] Dopping elements in balls solder alloys have been the object of extensive research , sometimes surprising : Aluminium , Antimony , Bismuth , Cerium , Cobalt , Indium , Germanium, Lead , Manganese Nickel, Phosphorus, Titanium and aiming to improve the robustness of the intermetallics formed during the soldering between BGA and board pads and the solder balls and to reduce evolution of the solder microstructure and the growth of intermetallics during ageing or under stress.

7 See figure 3 about visualization of the impact of Nickel in Intermetallics formed on Copper pads : modified Cu6Sn5 intermetallics dopped with Nickel have different diffusion properties : after long ageing the thin dark layer made of Cu3Sn intermetallics , just at Copper interface, is not or less visible even at very low Ni dopping concentrations. This dopping has been widely implemented, as leading to more robust joining of solder to Copper pads. At the same time Nickel+Gold coating became less popular both on board and on BGA substrates for simplification and cost reasons but also in order to improve solder joint interface robustness versus shocks as SAC alloys on Nickel coating is always more sensitive to shock than SAC alloys bonding on Copper.

8 Minapad 2014, May 21 22th, Grenoble; France Figure 3 : Nickel doping effect is well visualized after ageing , such as 3 weeks at 125 C. Experimentals A study of LFBGA 15x15 package has been done focusing on substrates pads metallization and solder balls Nickel dopping effect Another study has been done on PBGA 25x25 comparing stress on different solder balls and impact of balls matrix distribution. Refer to table 1 for packages details. See figure 4 packages balls matrix and die size representation/ Table 1: packages data package LFBGA PBGA package size 15 x 15 25 x 25 package body thickness mm mm Balls count 396 512 and 512+36 Balls pitch mm mm Balls diameter mm mm Balls alloy SAC305 / SACN305 SACN305 BGA pads metallurgy Ni-Au electrolytic / Cu + OSP Cu+ OSP Die size x mm x Die position centered centered Note : SAC305 is standing for Sn Ag Cu and SACN for Sn-Ag- Cu Figure 4.

9 LFBGA 15x15 396 and PBGA 25x25 512 and its modified version Daisy chained packages are manufactured using modified substrates to create in the bottom copper layer, series of 2 by 2 balls links to create a resistive chain once the package is soldered on board . The other parts of the substrates are kept as similar as possible to the substrate for functional devices with only modification to allow the chain built-up. Even if in real BGA products several groups of balls are at the same potential, the daisy chained are design with all ball independently having potential to break the chain. Similar dies than in real products are used but without connections to the substrate.

10 Other manufacturing conditions are kept as close as possible to real BGA products. A variation of PBGA 25x25 was created by adding dummy solder balls along each diagonal of the package , these solder balls are not part of the daisy chain and are expected to protect the other balls from the original package design . Test boards : For practical reasons all test boards are standardized to the format ( X , Y , clamping ) defined in JEDEC standard for Handheld Electronic drop testing [5] even when using packages bigger than 15x15., see figure 5. The LFBGA 15x15 board has patterns allowing to measure 2 chains : one for the 3 x 4 corner balls(chain A) and one for all the other balls of the package (chain B) : see figure 6 board Level reliability tests : LFBGA 15x15 packages once mounted on board were submitted to fatigue drop test at 300G decceleration with 2milliseconds deceleration peak width , see results in table 2.