IEC TR 62014-3: ICEM Models for EMI Behavioural Simulation of Integrated Circuits

A Technical Report on Modelling IC Internal Activities for Electromagnetic Compatibility Analysis

IEC TR 62014-3:2002 introduces the Integrated Circuit Emission Model (ICEM), a behavioural modelling framework for simulating the electromagnetic interference (EMI) characteristics of integrated circuits. As ICs integrate more gates on faster silicon, predicting electromagnetic behaviour becomes essential for avoiding costly EMC failures during product development. This technical report, developed by IEC technical committee 93 on design automation, provides the foundational model for evaluating conducted and radiated emissions from IC internal activities and represents the state of the art in IC-level EMC modelling at the time of publication.

The ICEM model fills a critical gap in the design flow: while IBIS and IMIC models focus on I/O interface behaviour for signal integrity analysis, ICEM specifically addresses emissions from internal chip activities, which are the dominant source of EMI in modern high-speed digital ICs.

The ICEM Model: Three Coupling Mechanisms

ICEM identifies three distinct mechanisms through which internal IC activities generate electromagnetic emissions, all addressed within a unified framework. The origin of parasitic emission lies in the current flowing through all IC gates during logic transitions. When hundreds of thousands of gates switch simultaneously at the clock edge, massive current spikes are created inside the die, inducing voltage drops in the internal voltage references. A modern IC integrating one million transistors can generate current peaks exceeding several amperes during clock edge transitions, with rise times as fast as 100 ps, creating broadband noise from DC to several gigahertz.

Coupling Mechanism Model Component Emission Path Frequency Range
Conducted via supply lines Power-supply line model Vdd/Vss pins to PCB power distribution network DC – 1 GHz+
Conducted via I/O lines Input/output model I/O pins to PCB traces and cables DC – 1 GHz+
Direct radiated emissions Direct radiation model IC package and die to free space 30 MHz – 1 GHz+
Simultaneous switching of thousands of gates creates current spikes that can exceed 1 A with sub-nanosecond rise times. This generates broadband noise from DC up to several GHz, making EMC prediction essential for high-speed digital design. Without proper modelling, these emissions can cause costly redesign cycles.

Power-Supply Line Model and Resonance Effects

The core of the ICEM model is the power-supply line model, consisting of a current generator (Ib) representing internal switching activity combined with RLC elements modelling the package, bonding wires, and on-chip interconnect. Primary resonance occurs between package inductance (LpackVdd, LpackVss) and parasitic capacitance (Cd) between Vdd and Vss pins, typically in the 10-100 MHz range. Secondary resonance involves on-chip series inductances (Lvdd, Lvss) and internal die capacitance (Cb), typically in the 100-1000 MHz range. A second-order model including both resonances provides significantly better correlation with measurements, particularly in the 100-1000 MHz range where secondary resonance can produce emission peaks 10-20 dB higher than first-order predictions.

Parameter Symbol Description Typical Range
Package inductance LpackVdd/Vss Lead frame or substrate inductance 1 nH – 10 nH
Parasitic capacitance Cd Package pin-to-pin capacitance 10 pF – 100 nF
Supply series resistance Rvdd/vss Bonding wire and die interconnect resistance 0.1 Ohm – 10 Ohm
Supply series inductance Lvdd/vss Bonding wire and on-chip inductance 1 nH – 20 nH
Internal die capacitance Cb On-chip decoupling capacitance 10 pF – 100 nF

I/O Coupling, Direct Radiation, and Parameter Extraction

The I/O coupling model addresses disturbances caused by internal voltage drops propagating through direct connections, parasitic coupling, and common impedance paths. PCB traces connected to I/Os act as antennas. The model supports both single-supply and multiple-supply structures, with substrate coupling impedance (Zsub) and I/O decoupling capacitance (Cio) as additional parameters. The direct radiation model is based on equivalent dipole representation, with parameters including internal currents, die size, internal loop areas, and package characteristics.

The ICEM model has been validated against actual measurements using IEC 61967-4 1 Ohm/150 Ohm direct coupling and TEM cell methods (IEC 61967-2). The excellent correlation between simulation and measurement confirms the model’s practical utility for EMC-aware design in real-world applications.

Parameter extraction can be performed through time-domain reflectometry (TDR) or network analyser measurements, or predicted from IC design tools that compute RLC parameters from geometrical and electrical characteristics. The current source Ib is described in piece-wise linear (PWL) format with amplitudes from several mA to 1 A, durations of 0.1 to 5 ns, and periods of 500 ps to 50 ns. These values depend closely on the software being executed on the IC. The standard provides typical value ranges and measurement methods for all model parameters.

Q1: How does ICEM differ from IBIS models?
IBIS focuses on I/O buffer behaviour and signal integrity parameters like cross-talk and overshoot, while ICEM targets EMI simulation by modelling internal chip activity and its coupling to power supplies, I/Os, and direct radiation.
Q2: What causes primary and secondary resonances?
Primary resonance arises from package inductance interacting with pin-to-pin parasitic capacitance. Secondary resonance involves on-chip interconnect inductance and die decoupling capacitance. Together they determine the IC emission spectrum.
Q3: How are ICEM parameters extracted?
Through TDR or network analyser measurements, or predicted from IC design data. The current source Ib is derived from measured external supply current using reverse engineering with known package parasitics.
Q4: Is ICEM applicable to modern deep-submicron ICs?
Yes, the model structure remains valid for advanced technologies. Parameter values shift with scaling but the modelling approach accommodates these changes through updated parameter extraction.

Leave a Reply

Your email address will not be published. Required fields are marked *