Ruggedized products and extreme design are always of interest, since they pose an extra challenge to the designer. While I have no personal need for them, I am interested in them from an engineering perspective. That's why announcements that say "rugged" or similar get my attention. (You can search for "extreme design" in the search box in the upper right corner to see the some other case-study stories published on the Planet Analog site.)
For example, I just saw the TPS50601 and TPS50301 point of load (PoL) converters from Texas Instruments, which offer extra-range temp operation to +125°:C and +210°:C, respectively. These could certainly be the building blocks of a larger design, which needs to work despite that high-end temperature challenge. But if you step back and look at the bigger design picture, rugged components alone may not be enough to meet specs, yet ironically they may actually be too much.
Why the contradiction? It's because there are really two dimensions to the challenge of a rugged design. First, defining specifically what sort of "ruggedness" is needed: temperature, vibration, pressure (high and low), supply voltage, or ESD (electro-static discharge), to cite a few. It's obviously much more difficult to design for several of these factors, compared to just one.
Second is what strategy you should use to actually achieve the system-level goals. The quick and easy answer is to start at the bottom with rugged components (such as ICs that can stand the heat or ESD) and then work up to improving the larger parts of the design, such as the PC board assembly. Using suitable building blocks certainly seems like a good way to have a solid foundation for the ruggedization of the larger design.
However, that may not be the best approach in the overall system-level context. Perhaps it would be better to use standard components coupled with aggressive cooling, or an ESD-wise PC board layout with an array of discrete protection devices, for the situations above. As in all things related to a "best" design, there are tradeoffs and constraints.
Even if you have components that can tolerate the extremes (such as heat or ESD) without failure, there's the associated issue of system performance to spec. For example, do you select analog parts with very low and specified drift over the temperature range? Or do you choose to instead just "deal with it" by calibrating the system at various temperature points, and then using these calibration factors to correct the acquired data values? Do you do this calibration at the factory, or do you use a precision, wide-range voltage reference and calibrate the unit in the field, while it is operating?
Sometimes, choosing the in-use calibrate/compensate approach with easier-to-get, lower-cost components gives better overall performance than trying to meet electrical specs with an inherently stable design—however inelegant or brute force that method may seem.
On the mechanical side, designers face the same choices: using rugged components and then building up, or using standard parts and then ruggedizing at the system level, if possible (but it often isn't for mechanical design; you can't calibrate a support bracket to have better performance).
Have you ever had to do a rugged design? In which aspects was it ruggedized, and to what extent? What was the biggest challenge you had, and the biggest surprise you encountered?