Using design of experiment to diagnose analog blocks geometrical defects: Application to current reference circuits

The purpose of this paper is to present an automated diagnosis methodology that targets analog blocks. An application example is given for a Current Reference (CR) circuit. The presented methodology focuses on speeding up the diagnosis process of anomalous variations of a CR outputs (i.e. the output current I
_REF
, the consumption current I
_SUNK
, the temperature dependency factor β
_T
and the supply voltage dependency factor β
_V
). This diagnosis methodology is based on a CR mathematical model which links specific CR design parameters to CR outputs. This model is generated thanks to a “Design Of Experiment” (DOE) technique. The DOE technique takes as input electrical simulation results of a CR circuit for different component geometries. DOE generates polynomial equations of the current reference outputs. Using these equations, the root cause of an anomalous CR output is detected in terms of CR design parameters.