ABSTRACT
When measuring a small but steady signal in the presence of random noise we can often improve the accuracy of the result by making a number of measurements and taking their average. This approach has the great advantage that it is easy to do - given enough time - but it cannot overcome all of the practical problems which arise when making real measurements. In particular, there are two sorts of problem which simple averaging copes with rather poorly: 1/fonoise, and the presence of Background effects.
