110214-0836 EST
AAEB:
You have to define your goal.
Generally from a practical point of view you want to sample something in excess of 2 times the maximum frequency component you want to observe.
There are tricks to employ to get short time events over long time periods.
If you want a clean look at transformer inrush current at 60 Hz, then I would want to sample at 8/256 milliseconds, or about 100,000 samples per second. Note: 8.3 milliseconds is 1/2 cycle at 60 Hz. See my scope photo P6 at
http://beta-a2.com/EE-photos.html . Sixteen samples might be sufficient for useful information, but I generally like clean waveforms.
When you collect massive quantities of data you need ways to work with the data and extract the desired information.
At home on my power monitor the maximum sample rate is about once per second. In a day this is 86,400 samples. Or for both voltage and power at 4 + 6 + 4 + 10 ASCII characters the total file size per day is about 2 megabytes. This is the maximum rate this commercial product provides (TED System). The rate is too slow to easily distinguish the startup of two different freezers. However, if your are not trying to distinguish between different motors for signature analysis, then the rate is more than adequate for most purposes.
.