Tuning of a capacitorless bandpass biquad through sequentially trained ANN

The sequential trained artificial neural network (ANN) based on updated training sets is successfully deployed to tune a capacitorless all-OTA bandpass biquad. The training set contains less than a few tens samples which are selected from predefine bias points that are closed to the desired biquad requirement. To limit training time, the less complex ANN is recommended. Feasibility of a biquad requirement is easily indicated by observing the maximum error of the worst element in an initial training set. A second-order bandpass requirement, centered at 406.2 MHz, is successfully tuned as a sample. The proposed feasibility analysis and tuning process are tested with one hundred random bandpass requirements. As there is no indication of type-I and type-II errors, the proposed process is considered very efficient.