> Or try : http://www.sipad.com/
I think sipad is incredibly cool and would be the perfect solution, but is way too expensive.
We're paying $0.80/each for four-layer ENIG boards, 50cm^2 with 0.125mm trace/space and 0.8mm-pitch LGA landings in lots of 1000 (and yes, the boards are working flawlessly at 800mbit/sec data rate / 400mhz toggle rate, please don't turn this thread into a PCB manufacturing quality religious war, I am tired of that happening).
Sipad would **quadruple** the price per board.
Personally I think their technology is great but their pricing is just stupid and unreasonable. All they do is run the board through an ordinary stencil printer, an ordinary oven, and then a special cold-press. I have a hard time believing that their special cold-press is that expensive to operate. Also our 0.8mm-pitch landings are slightly outside of their standard spec, so we'd be paying even more than the baseline pricing.
Also they don't seem to be too interested in new business. I requested a demo board from them three months ago and still haven't heard anything. That sort of thing is a Very Bad Sign from a vendor you are thinking of relying upon in a critical way.
The other problem is that we have LGAs on both sides of the board. This is not negotiable -- it keeps the critical traces short which matters more than anything else.
I have, believe it or not, begun experimenting with developing our own print-reflow-place-reflow process using a glass plate to planarize the solder deposits. The odds of it working are not great, but if it does work it will be a better fit than outsourced processing since we can do the following four steps in this order (each step is done for an entire lot of 1000 boards before the next step):
1. print+reflow the paste on the first side with planarization
2. place components on first side, re-reflow first side
3. print+reflow the paste on the second side with planarization
4. place components on first side, re-reflow second side
Step 1 and step 3 involve wet paste, so they have to be finished within one shift. That's not hard, they're the fastest and simplest steps. Steps 2+4 can be spread out over as many days as necessary.
Doing this with an outsourced provider would require shipping the boards twice and them being able to handle our boards with one side populated, which is both unlikely and risky in my opinion.
It might also be possible to do the steps in the order 1,3,2,4. In other words print+reflow the paste deposits on both sides before doing any placement. However this would require planarizing (or re-planarizing) the backside deposits during the reflow of the frontside components.
reply »