I’ve done QFN (stm32) and LGA12 (accelerometer and other sensors) but this seems a bit more daunting…
I’m thinking of tinning the pads with the iron and then reflowing with moderate flux amt. Dunno if I smear paste instead whether it’s going to create bridges. LGA seems worse than BGA…
Or do I ball the LGA pads and then pretend it’s a BGA?
It’s an OSH park PCB which didn’t come out to my liking (my fault):
The U-Blox stuff says to use 0.37mm solder mask opening, but that ends up exposing a lot of trace which skews the pads which doesn’t seem good. Maybe it doesn’t matter… Dunno whether using smaller openings would improve things or create other issues…
You might want to read up on the use of solder mask defined pads ("SMD"pads) where the desired exposed copper size of the pad is defined by the mask, not the etched copper. The etched copper is then made larger than the opening of the mask so that there are not any problems due to alignment issues between the copper and the solder mask.
Looks like the two main downsides of SMD pads are reduced spacing between pads for routing and lower adhesion of the solder to the pad. Sounds like that’s what I should have chosen. Next time…
Still wondering how best to solder. I think I’ll try minimal solder on the board using wick and then mounds on the chip and sand flat on paper.
like mentioned by others mask to find pads may have been a good choice for you.
ultimately you’re going to have problems because you’re doing high density interconnect work with conventional tolerances.
The image of your OSH Park board just don’t look good enough to get a quality soldering.
We do a good amount of hand-placed BGA work for prototypes.
I would not try to do this without at least a cheap stencil.
I would consider buying a u blocks module that is already been placed on a breakout PCB.
I am worried you’re going to go through a lot of boards and a lot of modules trying to get this to work. And without x-ray it is very difficult to see whether or not you’ve got bad solder joints.
If this is for one-off prototyping, you can probably get away with tinning the pads, clean, then apply fresh flux on top, and then using a hot-air tool to install the IC like doing a board rework. I lightly bounce the IC a bit to ensure good solder contact. This one may be a bit challenging since the amount of exposed copper vary significantly so the amount of solder on the tinned pads may be uneven.
I’m using a module for the software development, but I’m designing a 1.5 gram device and the modules I know of weigh a multiple of that just by themselves. Unfortunately they also don’t allow me to measure the current consumption accurately (3.3V operation vs. 1.8V).
So, I decided to “just try” and learn. Thanks for all the input!
I concur with Joseph. One thing that has helped me a lot is a board pre-heater for the underside. I let the board preheat for a minute or two, then I solder the part with a hot air gun. I find it much easier getting the chip I want on there without blowing all of my passives off the board with the air gun since I can use very low volume air flow.
I would add: buy a stencil from OSH Stencils or at the very least, get the appropriate repair stencil. Repair stencils are made for the cellphone repair business. Since it’s your design - get a real stencil. I think the OSH Stencils Polyamide stencils will be fine for 3 units, I’ve used them for 10 without problems.
I ended up tinning the board pads flat (solder wick), then adding solder mounds to the chip pads, then sanding flat on a sheet of paper, and then reflow with hot plate and hot air. Looks like it worked well enough for me to get GPS fixes .
I need to get some solder balls, that would have made things easier, I think…
The worst parts on this board are actually the ESD protection diodes on the antenna pads. I got “ultra-low parasitic capacitance” ones in an 0201 package. The size isn’t that bad, the real problem is that the devices have a mirror black finish. That makes them basically indistinguishable from shiny solder or flux. I had one that didn’t go where it was supposed to and finding an angle to see the darn thing was a nightmare. Next time I’ll settle for “very-low parasitic capacitance”
.