We're surrounding by instrumentation, ranging from basics such as voltmeters to sophisticated and complex units such as mass spectrometers. It is truly amazing what these instruments can do, and how they have developed with their various internal and external error sources designed out, compensated for, or calibrated away.
But sometimes, it's tough to develop an instrument for what seems to be an easy measurement. I visited a sophisticated weather station and they had all the modern instruments, as well as basics such as rain gauge, which simply, accurately, and reliably collects the rain in a tube and tells you how much rain has fallen. But the director said the instrument they really wanted was one which told them the type of rain drops, not the total amount of rain. They wanted to know if the rain was a fine mist, or larger drops, or something in-between. They did have a unit for this which they had rigged up, using a rotating drum with holes in its periphery, but it wasn't really very good and it required a lot of manual attention.
After I left, I gave the problem some thought. How could you measure and quantify the drop size? The obvious way would be to use a video camera as the front-end sensor, and do some image capture and analysis using various algorithms. This might work, but it would be relatively costly and power hungry, a drawback in some situations but not a problem in others. But how well would the system work? I don't know much about image-analysis algorithms, so I can't say; I assume that with enough development effort and memory and processing power, it could be done fairly well. Also, what practical issues would crop up, such as the lens of the camera getting covered inadvertent spray, thus interfering with the camera's field?
Perhaps there is some laser-based system that uses one or more wavelengths of light, and measures the transmittance through a known “rain zone”, or the reflectance from the drops? Or is there a flat sensor that can report back, based on how much of its surface gets covered, and in what sort of pattern, by the drops (it would have to be dried to reset, of course). What type of sensor could you use: capacitive? Resistive? Some type of X-Y matrix?
I don't have the answer, because there is no single answer. There are really two parallel problems here. First is the need to design a system which works reliably and consistently, using whatever techniques you can, as a proof of the underlying concept. Second is to find out which successful design can also meet the broader requirements of the application in terms of power, size, weight, and other factors, and these will differ depending in the specifics of the installation (what power is available, is the installation remote or accessible, how it will be delivered and installed, to cite just a few).
This is why instrumentation to assess real-world parameters is always such a challenge, and at the same time, so exciting. It's also why field trials and testing are so important, since the prototype bench is no substitute for the harsh reality of “out there.”♦