Gentlemen,
I too have been atacked by that "soldering to gold" and will add my 2 cents.
I also had parts that would pop off the board when the fillet looked perfect. After lots of pain, experimentation, and sleepless nights, I solved it through a better understanding of the soldering process of gold boards. The fix was in the profile.
The boards were small and thin and my first mistake was using a fixture to reflow them in mass. To get a reasonable thermal profile I needed more time in the oven. I was also afraid of getting the boards too hot, and kept the peak temp at 210. With this I created two problems. One was the preheat time. It was too long. After the flux burned off there was time enough to oxidize the nickle anywhere oxygen could get to it. I proved this by applying some flux to the bare board and running it through the oven, getting dark pads at the other end. Without the flux the pads were not nearly as dark when just the bare board was sent through the oven. The other problem was the liquidous time and temp. With gold, the solder joint forms to the nickle, not the gold. The gold is removed by the tin in the solder. The tin breaks the tin-lead bond and forms a bond with the gold,leaving the lead behind. (As a sidenote, when too much gold is on a board, the joint will be very weak as the tin-gold alloy is brittle and the lead seems to concentrate at the pad surface, which is nickle. This lead nickle interface is extreemly weak.) What you are really soldering to is the nickle, not copper and not gold. To solder to nickle the temperature needs to be hotter. The time must also allow the tin-gold alloy to mix in with the rest of the solder, otherwise you have all of the tin-gold at the interface of the pad, along with the lead the tin left behind. You want this interface to be solder and nickle, and form a proper intermetallic.
In my situation, I removed the boards from the fixture to minimize the mass, I cut to the minimum time for the preheat period based on the paste manufacture recommendations, I increased the peak temp to 225, and I kept the solder liquidous for about 50 seconds. The time and temp were decided on by destructive testing of many boards in an experimental basis to determine the greatest resistance to solder joint breakage. This profile for me resulted solder joints breaking at the component end, or the component itself breaking.
I put about 6 months of effort into this problem, and it was not fun. But I learned a thing or two, and I hope I have passed some of that on. Good luck!
reply »