So we have these multispectral sensor gizmos that go on large harvesters. They obviously have to work outdoors, so we're doing the whole IP67 thing--carefully designed O-ring seals, IP68 M12 connectors, carefully designed vents to keep liquid out while allowing air exchange (which keeps the seals from getting worked by the ~20% variation in air density over temperature and ambient pressure).
New to us, but entirely vanilla in the industry. (Simon did all that stuff.)
We put desiccant packs inside the box to prevent condensation on the window and on the board. They're clear plastic/Tyvek envelopes (intended for use in steam autoclaves) containing about 95% Linde 4A molecular sieve and 5% blue indicator silica gel. (I've mentioned them here before--5% is lots to be able to read the indicator, and the blue colour is cobalt chloride, not a super pleasant compound.)
For diagnostic purposes we put a Sensirion SHTC3 temperature & humidity sensor on the board, plus a dew point calculation. It reads out in
0.01% RH increments, and has a specified accuracy of 2% RH typical.Reading out the sensor continuously while closing up the box with the desiccant inside, the indicated humidity went from 45% or so right down to zero within 30 seconds or so. That probably means that it has a negative offset, but even getting the RH down below 2% is pretty good going.
Cheers
Phil Hobbs