New methods for determining the number of effective bits of A/D converters based on eigendecomposition

For calculating the number of effective bits, we measure the rms noise by fitting the ADC´s output signal to a curve. Thus, a nonlinear equation system is obtained and several Newton type iterative algorithms are used to resolve it. We established that a good estimation of the frequency for the first computation of an iterative process is necessary. To remove this problem, we proposed new methods based on the eigendecomposition of the autocorrelation matrix in the input signal. The first, using the MUSIC method gives an accurate prime estimation of the frequency for the model which fits the output data. Then, this value is introduced in a Newton type algorithm for a good estimation of the noise. The second, more attractive, provides a direct accurate estimation of the noise. Condition testing and simulations will be given to validate these new methods