I'm trying to set up an analogue to digital converter to read electrical noise. The noise is kinda Gaussian with a very very very rough amplitude of the order of 2Vp-p. I'm using a 10 bit range ADC that spans 2.5V, and have the opportunity to alter the amplitude of the signal up or down.
My problem is the Gaussian bit. This means that there is no limit to the range of the signal, just that signals spanning much more than 2V become increasingly rarer. Some times they might span 3Vp-p on rare occasion. More on even rarer occasions.
How do I set the input amplitude? Do I go low and only use 5 bits most of the time ensuring that I will never miss an out lying reading until the Universe cools? Or do I go high, use 9 bits then often miss higher amplitude signals? Err...
I'm not sure if this is actually an electrical or statistics question. However, this must be a common situation in signal acquisition so I'm hoping for some best practice advice. How would a commercial industrial acquisition device be configured specifically for this use case? I can only think of stupid analogies such as an accurate scientific beaver weighing device. How would I get the best use from my 10 bits to ensure that I can weigh all beavers with a good resolution? Please try to see this from a commercial perspective so that you're actually trying to design and sell these devices.
P.S. Having written the title lastly, my problem seems to boil down to "How to measure the amplitude of noise?"