When soldering the REF50xx chips, I put the boards on a pre-heater and slowly bring them up to 150c over the course of 30-45 minutes, then I apply a small heatsink to the chip itself and quickly drag solder each side at 265c with eutectic solder. Afterwards I ramp the pre-heater down to 85c over 15 minutes, then 50c over 15 minutes. Then I turn the heater off and leave the boards in place, so they're still receiving heat while the heater cools. I give it an hour or two to come back to ambient.
Overkill? Maybe. But I've got 8 of them going in a 25c chamber for the last 6 months gathering data, so I wanted to give them the best shot I could.
Next I want to put 10 REF5010's in series and make a 100 volt reference to classify.
Sent from my Smartphone
I played with some quad eutectic solder on a few boards a while back.....I didn't notice any significant improvement in the quality of the joints or heat vs time issues.....also played with some N2, using lead free....that most def yielded superior results....in terms of temps and times.....nice shiny joints....it's entirely possible I don't have the equipment or patience to track down the "ultimate result"....generally I found that a proper tip and nice multicore silver solder yielded very acceptable results....without the hassle etc....I know there are compendiums of information on these subjects, and i have read quite a bit, but never had a huge problem with the more traditional (hassle free) methods.....a simple heatsink on the IC or lead = totally acceptable results....and with a simple thermocouple fixed to the IC case or legs I was able to get at least ballpark temps going into the IC.....
I do believe there is a lot of validity to the socket vs shock concept though.....pick your poison.....I did notice VERY little negative impact from premium quality sockets though....however "bargain bin" sockets did change things for the worse in virtually every test I could conceive....
I played with all kinds of systems for suspending REF IC's above PCB's and even the opposite....dead bug thermally bonded to a common thermal layer.....in the end the results were mixed, and no appreciable differences were made in drift, over the sDEV of the components themselves....it's kind of like chasing 0.5ppm when the difference from IC to IC can be orders of magnitude higher.....
I did notice that aging and binning "premium chips" yielded substantially improved results though.....however that is shockingly expensive (on all resource fronts)....
The easiest solution to all of these "problems", for me, has simply been proper temp control in the system....best solution we came up with was a simple aluminum enclosure....with a TeC and TeC controller....bond the TeC to common thermal layers....ceramic coat the enclosure (external) and line the inside with some foil (we saw no impact between gold foil and the cheap stuff).....also simple heaters made from resistance wire, sandwiched between glass slides....yielded very nice results....less control and monitoring, but no issue about isolation the "cold" side of the TeC....which is a hell of a bear to implement....
routing around the REF IC on the PCB was another appreciable benefit (when the REF IC has an internal heater). Obviously the issues of supporting components comes up as well.....we found that "ovenizing" the entire package offered the best price to performance benefits.....the more esoteric stuff just seemed like time/money pissed away (lots of fun though)