Hello all,

I did a board with a 1mm pitch BGA 676-pins. The problem is that the fanout connections to vias are completely unshielded via soldermask, since they're so small.

I have a solder stencil, which I had planned on using to apply a small amount of paste to each pad. Then, I plan on using a hot plate to flow the board.

Must I use paste on the BGA pads? I know that the BGA balls are composed of solder, can I not just use no-clean flux and count on the solder from the balls to make the connection? My fear is that I will apply too much paste via the stencil, then short out the connections between pads and vias between them.

Anyone have experience with this?

Steve

 

Hi Steve.
Dealing with it every day, or better said, suffer and struggle with it daily, with professional equipment it is still risky. Manually can be very difficult; be prepared for failure.
Troubleshooting will be a nightmare if something ends less than perfect.

If the chip has to be soldered manually, works better on not-bumpy tinned pads wiped with braid , avoid unleaded solder; the board has to be preheated, the chip too, controlled hot air flow, proper flux amount should allow the chip to float on it barely touching balls-to-pads.
Alignment is of course important, not referenced to silkscreening but to pads themselves.
The heat melts the flux and the chip descends on its own weight to bonding when the right temperature is reached. A very steady table is mandatory, following a known recipe time/temperatures for a given BGA chip is almost unavoidable, as all the balls should melt simultaneously.

The lack of soldermask is serious trouble, sometimes the balls are wicked into the vias leaving little solder to ensure proper joint.
I would not use the solder paste, just smooth flat tinned pads.

Good luck, you need it.
Miguel

 

Miguel,

Thanks a lot for your input, eventhough it was a bit less optimistic than what I was hoping for. Nonetheless, a reality check is very well what I might of needed.

In the past, I had soldered a 256-pin BGA with 1mm pitch pads with success. I believe I used similar tactics, except I think the PCB manufacturer was able to do a better job with the soldermask.

The board was professionally-made, so all pads are pre-tinned and ready for soldering. So, can I just use a bit of flux painted on, then the BGA set in accurately?

Steve

 

One of the newer issues with BGA balls is that they are becoming RoHS compliant, this means lead free solder balls which just make a complicated system almost impossible due to reflow issues with non-leaded solder.

There are now some businesses that will actually reball a BGA using leaded solder.

I mention this because if you are DIY, you may want to make sure you have lead balls (as opposed to brass ones, I guess).

BTW -- you may want to look around and query a few assembly houses if they can do it for you for free, or at cost or something, (starving student, etc.).

 

I don't think much of your chances of success unless you have some pretty sophisticated equipment that we are unaware of. In your position I would hire someone with the experience and the equipment to do it right.

 

It seems like I am SOL in this situation. I will attempt soldering with some flux and a lot of care, hopefully it will be okay. As I mentioned before, I have soldered a 256-ball BGA in the past in a toaster oven with a lot of success. I didn't think it would be a big deal to move up to a 676-bga. Perhaps I just got lucky in the past?

I will let everyone know whether or not I was successful in my attempt.

Thanks for your replies!

Steve

 

I'm still pulling for you in any case.

 

Well, I totally bit the dust on the BGA soldering
One of them soldered very well, but the other had shorts on the bottom. So, I decided to redesign the PCB with the components surrounding the FPGAs and bought a virtex-2 pro development board with breakout headers to interface with it. I'm getting it next week and hand soldering it under a magnifying glass. The reflow of the PCB depicted was poor, since I was silly and got a RoHS-compliant PCB and I overbaked it a little. A lot of the SMT moved around when they were connected to the same net, as you can see with the resistors at one end joining.

Here's a pic of my failed PCB I'm glad most of the ICs are samples... There are some bridges that I could've taken out, but it was a bit useless to try when the BGA failed.

 

Scubasteve_911,

I gotta give you a hand for taking on this challenge.

hgmjr

 

Thanks a lot I really hope I can get this thing working by the end of May !

I forgot to mention what really destroyed the board.. It was thermal deformation! I had a really good quality hot plate and I tried to bring the temperature up slowly. It didn't matter how I did it, the board deformed significantly. I learned this the hard way and the BGAs barely soldered. I took it off and used a 1/4" cast iron plate, drilled and tapped holes into it, then heated the plate. There was no deformation, but I imagine there was a lot of stress on the board. The one BGA flowed nicely and I checked known connections for shorting, it was perfect. The second one got too much heat and solder flowed down through the vias and made a mess.

I just wanted to tell anyone that came across this thread, it isn't really doable without the right equipment, as advised. Maybe if you were doing a really small board, like <15 square inches, it may work a bit better. I do have proof that I soldered a 256-BGA with 1mm pitch on a <5 square inch PCB.

Steve

Steve

 

Well, it was a pretty board, so it's really a too-bader that it didn't come out.

This is a suggestion made in ignorance - I do some smt work, but have never even considered BGA. I wonder if you can get results by limiting the heated area to just that under the BGA device. In other words, see if you can get it soldered in separately from the other devices. You might have to hand-solder the decoupling caps and such, but you might dodge the deformation and component slipping issue.

 

thanks for the compliment

I tried to do what you suggest, but it just made it worst. The entire PCB has to be brought up to temperature at once, otherwise, you get major deformation. This may be more significant with 6layer pcbs than 2-layer.

Steve

 

Wonder what the failure rate is in industry?

 

Finally, I broke the PCB into two parts, took out the spartans, and put connectors on it. The result is the following. I machined a custom heatsink for it too, which interfaces directly with the powerpad on the h-bridges.

Steve

 

Very pretty!

 

Thank you Sir Now hopefully it works as good as it looks!

Steve

 

Nice work Steve.

As a curiosity, where'd you get access to machine the heatsink?

Dave

 

Thanks Dave ,

I have a little machineshop room in my apartment. I have a Taig CNC-ready mill that I rigged up with a homebrew servo controller, so I use that for small jobs. The rest of the bigger stuff I have to do, I do it at school. I also have a small lathe harbour freight 8X12" that I use for making shafts and etc.

We have been having some issues with the machine, so the heatsink doesn't look exactly as planned. We use Mastercam to generate G-code, then Mach3 CNC to use the parallel port to provide step/dir signals to the controller.

Steve

 

A machineshop in your apartment -very nice Sounds like you have a snazzy little set-up, you'll have to post up some pictures some time to make us jealous!

Dave

 

hehe thanks, I probably can snap a picture of my girlfriend slapping me in the face for the little aluminum chips scattered throughout the apartment.

It's actually a bit lacklustre right now, not soo impressive with tools all over the place

Steve