On the use of signals with a constant signal-to-noise ratio in the frequency-domain

It is pointed out that most estimation methods, when applied to linear time-invariant systems, do not take into account the noise on the input signal. It is well known that these methods generate biased estimates when input noise is present. It is demonstrated that unbiased pole-zero estimates of the device under test (DUT) can still be obtained when an input signal with a constant signal-to-noise ratio (SNR) in the frequency domain is used. By means of a practical problem, it is experimentally demonstrated that the multisine technique gives the necessary flexibility to create optimal signals with a constant SNR, even when the noise is strongly colored.<
>