POP Memory assembly and fab (production/proto runs)

Hello Gerald,
                  Seeing as you seem to have the most knowledge in the
area of fab/assy for the Beagleboard (with regards to POP technology),
I had a few questions for you - We're trying to evaluate the
feasibility of an OMAP3530 design with regards to fab/assy (cost,
yield, etc). I saw your presentation slides from ESC (ESC-120 Slides
by you and Jason Kridner) and the myriad of assembly issues you faced
for POP.

1) Is POP technology complicated enough that i'd be limited to
specific assy vendors? Like CircuitCo?
2) This process of discovery/trial and error/multiple iterations that
you mentioned - was that undertaken by CircuitCo themselves?
3) What would be your general advise with regards to using POP, and
pitfalls, etc. to watch out for?

Thank you kindly,

Jerry Johns

POP assembly is not at all complicated. Anyone can do it if they are willing to learn how to do it. But, I have seem several assembly houses that don’t have the guts to try it becuase they make it out as being complicated than it is. Heck, even the military loves POP!

CircuitCo and I worked together on that trial an error phase. That is not something that you would need to replicate, unless of course you just enjoy pain. Listen and do what we say, and it will work. We have replicated this with dozens of customers successfully

My advice is to forget POP and focus on doing the PCNB right. If you do that, you won’t have any POP issues. Remember in my presentation, we had no issues that we could trace back to POP only. POP is not difficult at all if the PCB is correct. It is no worse than soldeing a BGA package on a thin PCB.

I hope this helps!


First of all would like to say, that I totally agree with Gerald on this. POP isn’t the problem.

Production of the PCB for the 0.4mm pitch BGA’s is the main problem. Mainly due to:

  1. Too thin (or no) solder mask in between pads (Beagle uses SolderMaskDefinedPads AFAIR)

  2. Miss-registered solder mask (compared to the pad location)

  3. Fine etch technology (for 75um/100um track/spacing)

Last but not least the routing of all the pins from the OMAP, requires in-pad u-VIAs (for the CBB package).

@Gerald: With respect to in-pad u-VIAs on Beagle. Are they actually filled (to avoid solder voids due to raising air during reflow)?

From the ctc045.brd file I would tend to think they aren’t, since it seems that areas marked by note 6 are only around the connectors. Have you experienced any problems with this? And in given case, can you comment on it? Or are the somehow implicitly filled?

Best regards and thanks


We fill al the vias on the vias on the board. OFr the pads, you don’t want nay dips to where air can collect. Via in pad is not required on all pins. It really depends on how each pin is used.


Thanks for clear and quick answer – And I admit, that I was unclear with the “all the pins”-expression in my previous email.

Totally agree, that you don’t need via-in-pad for all pads, and that actual usage depends highly on what kind of signals you need routed from the OMAP. What I meant with the sentence was, that you need to use via-in-pad in some of the pads in order to be able to route (all) the many signals from the OMAP. I hope this expression was a bit better J

Time to go to sleep - Goodnight from Europe J




Thanks for the feedback guys, its been much appreciated!


My pleasure!


Hi Jerry;

As Gerald says, the POP isn't really any different then just another
BGA. The important part is just doing a good job with the layout and
fab. Filling and plating over the vias is a must. Good soldermask
registration is a must. Beagleboard does have the vias filled and
plated, as described in their documentation.

We've assembled POP here at Screaming Circuits without issue.
Duane Benson