I'd like to do a home project where I take temperatures inside of my freezer and inside liquids. Liquid-compatibility aside, I'm mostly curious about whether or not anyone can recommend some best practices for the board design. I haven't found any resources online that explain what will happen in applications that are intended for 0C temperatures. I assume the solder joints will contract due to the cold, but does this mean that a surface-mount design will inherently fail? Therefore, should I limit my temperature (and possibly microcontroller) selections to through-hole parts? Is there a better solder formulation to go with, i.e. lead-free or not? Should the pads or through-holes be larger than what's normally found on the IC data sheets in the footprint section?

Any advice or insight would be really appreciated. I'm more than willing to experiment and learn from my mistakes, but if I could get a better head start, that would be nice. :)

-
0°C is nothing. Lead Free is good, because lead and food don't mix. – Passerby Mar 2 '14 at 1:17
Below -70degC solder rich in Tin >20% is considered a bad idea. At 0degC I would vorry about condensation as mentioned below and conformal coat or pot the circuitry. farnell.com/datasheets/315929.pdf – KalleMP Oct 24 '15 at 17:16

Some useful information about different types of solder: NIST Metallurgy

The main important tables are 1.12 (coefficient of thermal expansion/elastic properties of leaded solders) and 1.14 (tensile/shear strength of leaded solders).

I believe the document has information for lead-free solders as well (that is what it is called after all), I didn't look too hard for these.

The key properties for 63/37 leaded solder:

Coeff of thermal expansion: $$\alpha = 24 \frac{10^{-6}}{K}$$ Elastic Modulus (I'm using the 20 degree figure, it will be slightly higher near 0 degrees, not exceeding 38.1 GPa at -70 degrees): $$E = 30.2 GPa$$ Tensile strength: $$\sigma_{max} = 56.19 MPa$$

The worst case scenario is if the solder is mounted onto something completely rigid. Suppose we were to take the 0 stress state as room temperature (25C).

The contraction due to thermal expansion is:

$$\epsilon = \alpha (25C - 0C) = 0.0006$$

And the appropriate tensile stress is: $$\sigma = E \epsilon = 18.12 MPa$$

This is well below the tensile strength of the solder.

However! Even better is that the PCB board itself will contract with the solder as it cools down. Depending on the actual layup direction, this closely matches the 63/37 solder CTE (~20e-6/C for the primary direction), so the actual stress will be lower.

tl;dr: you'll be fine. You might have to worry more about moisture/condensation, as well as having components which are rated for below 0C operation instead of worrying about solder joints cracking.

-
That's a really great, thorough explanation. Thank you very much for posting that. – Dave Mar 1 '14 at 23:49

The differential contraction is pretty small; normal SMT works down to at least -40C with no special techniques. Thermal cycling can be a bit more of a problem, if you're going from -40 to +100 and back every few minutes something will snap after a while. BGA parts are more vulnerable to this.

PCB conformal coating (applied after assembly and test) will keep water out of the board.

-
Conformal coating sounds like a good idea -- thanks! – Dave Mar 1 '14 at 23:49