Monte McGuire said:
I've thought about this too, but one of the issues that comes up is not only the need to cool the contents somewhat rapidly, but also the need to provide a relatively fast positive temperature ramp.
[...]
So, I'd have to dig into the specs to work it out, but having a relatively fast, controlled rate ramp is really crucial to getting this to work, both going up in temperature and going down. Going down can be easy if you have liquid Nitrogen, but that's also a hardship for small labs.
So, I know people have made this work, but it seems on the ragged edge. Any promising thoughts?
FWIW, at my university lab we have a cobbled-together reflow oven (was a junkyard rescue, actually...).
It is rather small, some 15 x 20 x 20cm and I'd assume it's not really powerful, as it was not meant for baking stuff.
It does not have enough power to meet the ramp-up specs, and we're really unable to meet ramp-down by just opening the doors. The standard profile is way too fast for the oven: we begin liquid phase when the standard curve is already cooling down.
So, we basically let the PID follow the curve as much as possible, then we just switch it in manual mode and keep it at peak-ish temperature "for a bit" before turning it off and letting it cool with the door open. Very, very much imprecise and guesswork.
As much as I wouldn't advocate this setup for anything resembling production, we've been able to reliably solder 1mm and 0.8mm BGAs, 0.4mm QFN and so on, with minimal need for rework, except some occasional tombstoning on 0402 components. Actives (power devices, FPGAs, fast SiGe logic andRF stuff, and so on) seem to get thru the process with no damage/deterioration; same goes for passives, connectors and reflow compatible plastic parts.
Bottom line, in my experience, if you are doing small or single quantities and can afford to test each board and fix the occasional issue, you can maul the standard curve quite significantly and still get away with it.
Caveats:
- We still use leaded solder (research, not sale, so RoHS doesn't bother us)
- My definition of "reliably solder" is: all the pins we use (e.g. in FPGAs not all are used) make electrical contact, and we don't observe performance degradation or failures that we can relate to bad soldering or overcooked components.
- Can't comment on long-term (> 1-2 year) reliability as we got that thing just a couple years ago.