Components embedded on pads of BGA?

[sacherjj]

We were doing first run assembly of my ARM board and have a 441 ball 0.8mm pitch Processor (AllWinner A20).  We had to go from circle to square aperture on the stencil to get consistent and non-removing balls on the BGA.  So with a new setup, we take the board under xray to validate the BGAs and promptly freaked out a little.  Solder bridges everywhere.  Or where there...   We angled the xray (just 2d capable) and see the "solder bridges" climb just like the decoupling caps in the image.  So this is inside the chip.  Putting non-mounted extra chips in the xray confirms it.  I should note that this is a single component side board, i.e. nothing on the back side to ghost the xray.

Xray image


Highlighted "solder bridges" that are embedded components


Is this fairly common to embed components at the pad level in a package?  It was a surprise to the production engineer at our assembler.  I need to look at the pinouts and see if these are embedded decoupling caps or possibly series termination resistors to control edge speed.  The large ones look like 0402 sized packages.  But some are down 1/2 or 1/4 that size.

[free_electron]

it is common for such big SOC;s to have embedded passives in the cavity.
01005 parts.

[Dongulus]

I pried the cap off of a busted Xilinx Zynq SoC just yesterday at work for curiosity's sake and I was similarly surprised to see discrete components. The substrate is some fiberglass PCB material, FR-4 or possibly something more special  , to which the silicon die is bonded directly in the center. Surrounding the die a whole bunch of traces route signals away to a few discrete components and many, many vias that route the signals to the solder balls under the package.

[marshallh]

If it's overmolded wirebond like most monolithic bga packages, it probably has caps on the interposer (what the die is mounted and bonded out to)

For parts that are designed to run wtih a heatsink, they are mounted flip chip with the bottom of the die mating with the heatsink on the top.



My next question would be why you switched to square apertures for the ball lands? Are you using soldermask defined pads?

[sacherjj]

My next question would be why you switched to square apertures for the ball lands? Are you using soldermask defined pads?

We had recommended pads at first and tried a few pastes and ball sizes.  What was happening is the paste would periodically get removed with the stencil.  So we would have dry pads on the PCB.  The square shape works well.  It gets the correct amount of paste on the pad, but doesn't tug on the paste around the whole circumference of the circle.  It pulls from the corners of the square at first and works down to the middle of the side last.  This distributes the force sheering the paste, so the adhesion to the PCB pad is greater than the individual sheering force at a given time. 

Think of a cylindrical blade trying to cut through a trampoline.  All cutting force is distributed along the entire blade, as the trampoline bends around it.  Now use a square blade (like 4 sides of a cube).  The cutting force is going to be focused at each corner and make a quick puncture, then ride the edge towards the center.