A Novel LBIST Signature Computation Method for Automotive Microcontrollers using a Digital Twin

LBIST has been proven to be an effective measure for reaching functional safety goals for automotive microcontrollers. Due to a large variety of recent innovative features, every customer can adjust LBIST settings in a way that fits their use case. The downside of these user-defined configurations is the handling of their golden signatures: Traditionally, they can be computed only with access to the gate-level netlist. This is typically not possible for MCU customers because a netlist contains protected IP, which cannot be disclosed to third parties.This paper proposes a digital twin of the LBIST functionality that can overcome this drawback. It is an executable model that can be delivered together with the product. As a result, for the first time, a customer can compute a golden signature without knowledge of the netlist or other support of the supplier. We prove the efficacy of the digital twin in an industrial environment on an automotive microcontroller.