BGA/LGA chips

[Chris Hamilton]

I was curious if anyone had experience laying out BGA chips.  I know I don't and they appear to require multilayer boards just to be fully connected.  I have been quite unsure of exact chip selection vs. limiting the number of boards.  Motorola just announced some new K20 based 802.15.4 capable MCUs and I already am evaluating a low end K20 as our standard USB driver. - http://cache.freescale.com/files/rf_if/doc/data_sheet/MKW22D512V.pdf
The top end version combines the USB OTG, but it uses LGA (and is only available in tray, but that is a minor issue).  I know there is only about half a dozen people here, but I thought I would ask.

[Chris Hamilton]

OK, well I was looking at the LGA for this chip, and it appears that the 6 pins under it are all factory test, so I don't think it will be hard to lay this out with just a special ground plain that doesn't touch those pins.   Still, I am curious if anyone has worked with real BGA :).  There are quite a lot of tiny charger and fuel gauge chips that utilize it and BGA would open us up to more small MCU designs.  I know that BGA+heat+lead free solder usually ends up with the balls failing.

Also, I just noticed I had to turn on post by email.  Any other group features that are missing, let me know.

On Feb 18, 2014, at 6:42 PM, Chris Hamilton <ccha...@gmail.com> wrote:

I was curious if anyone had experience laying out BGA chips.  I know I don't and they appear to require multilayer boards just to be fully connected.  I have been quite unsure of exact chip selection vs. limiting the number of boards.  Motorola just announced some new K20 based 802.15.4 capable MCUs and I already am evaluating a low end K20 as our standard USB driver. - http://cache.freescale.com/files/rf_if/doc/data_sheet/MKW22D512V.pdf
The top end version combines the USB OTG, but it uses LGA (and is only available in tray, but that is a minor issue).  I know there is only about half a dozen people here, but I thought I would ask.

--
You received this message because you are subscribed to the Google Groups "Fyber Labs Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fyberlabs+...@googlegroups.com.
To post to this group, send email to fybe...@googlegroups.com.
Visit this group at http://groups.google.com/group/fyberlabs.
For more options, visit https://groups.google.com/groups/opt_out.

[Chris Hamilton]

Just some more info about this.  I spent some time researching ‘proper’ footprints for chips.  I am not sure there is a solid standard, but diptrace ‘oval’ (actually rounded rectangle) pads appear to be what the PCB Expert/IPC people recommend.  I have gone through most of the footprints I already did and tried to make them more correct.

I think it is debatable whether the length of the pad should be based on the maximum or nominal published in the datasheet.  It appears that many manufacturers recommend laying out a pad twice the length centered on the chip edge for similar QFP D shaped pads (which the LGA in this case really is).  So the rounded rectangle is a pad shaped like an elongated D back to back.  I have seen a couple designs that stop at exactly the edge though and so only have a single D side.  The width standard appears to be nominal.  For now I think we should go with rounded rectangle maximum length, nominal width.

I plan to use the reference nRF51822 DCDC reference design with 0402 with a stencil and see how that works.  I met some other hardware guys at a meetup last week and was recommended a cheap place to get laser cut polyimide stencils.  The hardware community here is definitely building.