IO information-based component relevance estimation in a mixed-signal ASIC analog circuit

With increasing demand for active sensor surface and compact modularization, monolithic active pixel sensors are gaining track in high-energy physics experiments. The analog circuits in these mixed-signal ASICs are often black boxes where the components’ function and the circuitry’s behavior can only be evaluated through simulation or educated guesses. Further, parameters acting on these components are often present, allowing the tuning of the behavior towards optimal performance of the circuitry while in operation. These parameters span a multidimensional feature space that is highly intertwined, making precise predictions of their influence on the sensor performance challenging. An approach to evaluate the interplay and importance of these parameters based on Shapley values is presented, which can provide information about the implemented circuit on the ASIC. We show how input–output information can be used, to assess the importance and influence of the components in the analog circuit on the performance of an exemplary pixel sensor. The resulting increase in information can help to evaluate the best tuning parameters of the circuit respective to the present requirements of the users as well as generate valuable feedback for designers guaranteeing the optimal performance of the sensor.