Verarbeitungsverfahren und Systeme für Künstliche Neuronale Netze zur Bereitstellung von Halbleiterbauelementparametern

Details

Veröffentlichungs-Nr.

EP4749510

Anmeldetag

3. November 2025

Veröffentlichung

27. Mai 2026

Rechtsraum

EP

Abstract

A method, comprising: providing a set of measured characteristic curves values (MS) of an electronic device having a control node and a current flow path therethrough between a first current node and a second current node, the measured characteristic curves values (MS) comprising measured electric current values through the current flow path as a function of electric voltage values across first and second current nodes with a set of control voltage levels applied at said control node of the semiconductor device (10); applying an artificial neural network, ANN processing pipeline (30) to the set of measured characteristic curves values (MS), obtaining a set of reconstructed physical parameter values of the semiconductor device (10) as a result. The ANN processing pipeline (30) comprises a plurality of ANN processing stages (340, 342, 343, 344, 346, 347, 348) each having at least one respective set of ANN processing weights (W1, W2, W3, W4, W5, W6; WG). The method further comprises: computing reconstructed characteristic curves values of the semiconductor device (10) based on the set of reconstructed physical parameter values and a set of device modeling equations; performing a comparison (36) of reconstructed characteristic curves values and measured characteristic curves values (MS), and based on the performed comparison, providing an indicator (D) of a variation of physical values of the semiconductor device (10).

Anmelder

Firma

STMicroelectronics International N.V.

Land

🇨🇭 Schweiz