There are a couple of pieces of data required to answer this question. We have the size of the component, the number of balls and their size. What TSY has not supplied is the use environment for the device.
In theory you do not need any underfill, however if you want to sleep at night you might want to consider using underfill. Some of our customers are very sophisticated, and can model a circuit board, and the potential vibration / shock it will see.
From this they can determine which parts on the board require underfill. Then again we have some major customers who have determined that the risk is to great not to add an underfill process in their assembly.
There are also trade offs you can make. You can have a full capillary underfill, corner dots, edge bonds or no-flow underfills. Full capillary will give the most reliable product, then everything else is a trade off of cost for reliablity.
If edge bond allows you to meet the design requirements for reliablity then use this process. There are other trade offs you can make with regards to underfill.
I have seen data that suggest a really good board clean to remove all traces of flux prior to underfilling a part, can give you double the number the thermal cycles compared to no cleaning.
If you are looking for a couple of tests to determine the reliablity of your products. Try the Jedec Drop test, where a, board is dropped a maximum of 30 times or until 80% of the electrical joints go to a high resistance (JESD22-B111).
Another test that many manufactures ofcell phones use, is a Temperature Cycling Test. Jedec test JESD22-A104 will give the conditions of the test for different reliablity levels. However many cell phone manufactures use -40degc to 125degC, for 1000 to 2000 cycles.
For further reading:
"Qualification approaches and thermal cycle test results for CSP/BGA/FCBGA"
Reza Ghaffarian, Jet Propulsion Laboratory, California Institute of Technology, MS 125-152 4800, Oak Grove Drive, Pasadena, CA 911009, USA
It all depends if you want to sleep at night.
