I am working on extending the features of a piece of consumer electronics(WiFi router), and the CPU is a ~350 pin BGA. I need access to several signals that are not broken out to the mainboard. My theory was to fab a pcb with many(!) vias that passed straight through to the bottom of the board, that would sandwich between the chip and MB to break out the signals I needed. Has anyone ever tried or heard of something like this? Thanks!
I think you are biting off more than you can chew. you will need BGA equipment just to remove and then reball and then re-install the BGA, not to mention the bake time before removing the part. In theory it could work, but a lot depends on the ball pitch, potential loading you are adding, etc, etc.
All in all, I think you run more of a risk of damaging the part/router than you will getting it to work once you have modd’ed it.
Croc4
There is a very good chance that you are right, however I do have access to BGA equipment, and I am just crazy enough to try things sometimes:) I accept that I have exceeded the norms of rationality, and economy(Gonna be an expensive little board), but I still may try… if so, I’ll pass on any techniques to all…
You can get conductive elastomer sockets for BGAs, so you wouldn’t have to solder up this stack once you pry things apart. They’re a little touchy about clamping force, but they work quite well.
Which would be cheaper: Making this special board to go between a bga part and an extisting board? Or having a proto of the existing board made with the signals you need routed to be accessible?
I am investigating a duplicate of the original board… testing my reverse-engineering skills:) If anybody has a full datasheet of a Broadcom BCM5354, drop me a PM:)
Possible food for thought…
Schmartboard has BGA boards in 1.27mm and 1.0mm pitch that take up to 400 pins:
http://www.schmartboard.com/index.asp?page=products_bga
-Bill
Thanks… thats probably exactly what I will do. What can I use for jumpers though when chip is running at 240 Mhz… cut pieces of cat5 guts? those are my standby for small bits o wire.
UGH… prototyping this is gonna be fun… its a a 400 ball BGA, and on something less than 1 mm pitch… Gonna need to design something like the SchmartBoard to work with it…
I hate to rain on your parade, but hand wiring the SDRAMs is just not going to fly. You would really have to reverse engineer the board and build your own version, even then while its not impossible without datasheets and support from Broadcomm, you’re probably in for a long, expensive and frustrating process.
Thats kinda what I figured… This is gonna be fun… and since the WIFI radio is on the BGA, impedance matching the trace there… I seem to have bit off a whole lot… But this is gonna be fun…
Having worked on a similar projects at one of the big T&M companies, there is one tip you should be aware of. We had BGA interposer/probe for a high-density high-speed package. We used Lead Solder on one side of the interposer and Non-Lead Solder on the other. This way we could bake the chip to the interposer and then the interposer to the original board/DUT. Otherwise, trying to make all 3 bake together at the same time is almost impossible. (The DUT is going to dissipate heat way too relative to your interposer and chip.)
As you mention, impedance matching is going to be difficult. (In case you are asking, how we did it… we were able to embed a resistive material on an inter-layer of the interposer. This effectively made a resister which isolated the extended traces from the top/bottom VIAs. You won’t have access to that process.)
Now thats the kind of info and experience I was hoping somebody had. I guess my biggest concern is the pin that is the WiFi radio. I wont be tapping that line, so the question is, how much does one pcb thickness and one extra solder ball affect impedance matching and potentially SWR of the antenna. Probably will have to re-tune the antenna, but that’s not so awful. The signals that I am attempting to tap are GPIOs and USB on the chip, at least to start with, so they should be fairly tolerant comparatively speaking. Ultimately I will need another address signal to be able to connect higher capacity RAM, but I dont believe that it is practical to attempt to tap it in this fashion. Probably will have to wait to re-design the entire PCB to do that. Thanks for all the help that everyone has given!
Your five second WAG on the joys of interconnect: the only interface type that’s problematic is SDRAM. Everything else you describe is unlikely to object to going through a funky connector stack. SDRAM is very touchy, and even straight-through your stack might screw it up. Before you get too excited about adding “another address signal” you should verify that the Broadcom part is driving it (and, for that matter, that it isn’t already wired to the SDRAM on the board).
The line(would be routed to A12) on the SDRAM exists on the Broadcom chip, however is not routed on the PCB. I’m not even gonna try to mess with that until I make a whole new PCB for the entire router. I am going to attempt the “stack” to access the GPIOs and USB connections, but I’m not going to attempt anything other than passing the existing RAM signals straight through.
After skimming this thread, a post on Hackaday about probing BGA package signals came to mind:
It’s an old topic but were there any results?