ABSTRACT
It goes without saying that any device has to be tested before it can be used for measurements. Generally speaking, the testing procedure is quite simple: A known or specially generated signal Iin is applied to the device input and the output signal Iout is compared against its expected value, which is GIin = U. Let σ be the measurement error, which is determined by the level of noise in the measuring system. If the condition I Iout in
2 − ≤ σG is fullled, then everything goes well, the device works
correctly, and it can be safely used for measurements. is looks really simple, provided we know the operator G and the value of σ, which describe, respectively, the properties of the measuring system and the noise acting in it. As soon as the operator G is known, one can always estimate the output signal U = Iout = GIin or reconstruct the input signal from the output signal with the help of the equation U I N= +G in , where N represents the noise in the system. e success of this latter procedure depends on the magnitude of the signal-to-noise ratio (S/N) GI N.in is equation can also be used to obtain the operator G itself. is option is oen much simpler to realize since one can always use the input signals of fairly high amplitude so that the relation GI Nin >> is fullled. We start our discussion by illustrating these methods and related problems for linear measuring-recording systems.
