ARRUDA, B. W. S.; DE SOUZA ARRUDA, BRUNO WILLIAN.; ARRUDA, BRUNO WILLIAN DE SOUZA.; http://lattes.cnpq.br/5547681254311718; ARRUDA, Bruno Willian de Souza.
Resumo:
In this thesis, we propose calibration methods for analog-to-information converters (AIC) based
on the random modulation pre-integrator architecture. In a state-of-the-art search about AIC
calibration, it has been found that implementations of this kind of converter frequently suffer
the impact of hardware non-idealities in the signal reconstruction step. In addition, it was
verified that the solutions addressed by several works make use of blind calibration methods,
which is a procedure in which the calibration coefficients are acquired together with the sparse
signal to be recovered, by means of a reconstruction algorithm. However, the reconstruction
performed by these algorithms does not always guarantee a satisfactory result, since the input
parameters influence the reconstruction of the signal. Therefore, to avoid the influence of
reconstruction algorithms in the calibration process, the methods proposed in this thesis use
only the measurements performed by the AIC’ hardware and those obtained by its dynamic
model, so that the calibration coefficients (gain, offset and delay errors) are known before
the reconstruction step, thus removing the influence of the algorithm on the accuracy of the
signal reconstruction. Experiments and simulations were performed with one, two and three
tone signals, and evaluation metrics were used to evaluate the measurement performed by the
converter and the signal reconstruction (performed by two different reconstruction algorithms).
As an example of metrics, we used the signal-to-noise and distortion ratio (SINAD), the mean
square error and the reconstruction signal-to-noise ratio. From the results, the efficiency of
the methods was verified by means of an improvement of 82.2%, 28.2% and 48.7% of the
SINAD value of calibrated measurements of one, two and three tone signals, respectively, taking
simulated measurements as ideal values. As a result of improvements in calibrated measurements,
there was also a significant improvement over signal reconstruction metrics.