Device and Circuit Level EMI Induced Vulnerability: Modeling and Experiments

Thumbnail Image


Publication or External Link





Electro-magnetic interference (EMI) commonly exists in electronic equipment containing semiconductor-based integrated circuits (ICs). Metal-oxide-semiconductor field-effect-transistors (MOSFETs) in the ICs may be disrupted under EMI conditions due to transient voltage-current surges, and their internal states may change undesirably. In this work, the vulnerabilities of silicon MOSFETs under EMI are studied at the device and the circuit levels, categorized as non-permanent upsets (Soft Errors'') and permanent damages (Hard Failures'').

The Soft Errors, such as temporary bit errors and waveform distortions, may happen or be intensified under EMI, as the transient disruptions activate unwanted and highly non-linear changes inside MOSFETs, such as Impact Ionization and Snapback. The system may be corrected from the erroneous state when the EMI condition is removed. We simulate planar silicon n-type MOSFETs at the device level to study the physical mechanisms leading to or complicate the short-term, signal-level Soft Errors. We experimentally tested commercially available MOSFET devices. Not included in regular MOSFET models, exponential-like current increases as the terminal voltage increases are observed and explained using the device-level knowledge. We develop a compact Soft Error model, compatible with circuit simulators using lumped (or compact-model) components and closed-form expressions, such as SPICE, and calibrate it with our in-house experimental data using an in-house extraction technique based on the Genetic Algorithm. Example circuits are simulated using the extracted device model and under EMI-induced transient disruptions.

The EMI voltage-current disruptions may also lead to permanent Hard Failures that cannot be repaired without replacement. One type of Hard Failures, the MOSFET gate dielectric (or ``oxide'') breakdown, can result in input-output relation changes and additional thermal runaway. We have fabricated individual MOSFET devices at the FabLab at the University of Maryland NanoCenter. We experimentally stress-test the fabricated devices and observe the rapid, permanent oxide breakdown. Then, we simulate a nano-scale FinFET device with ultra-thin gate oxide at the device level. Then, we apply the knowledge from our experiments to the simulated FinFET, producing a gate oxide breakdown Hard Failure circuit model.

The proposed workflow enables the evaluation of EMI-induced vulnerabilities in circuit simulations before actual fabrication and experiments, which can help the early-stage prototyping process and reduce the development time.