In-Vehicle EMC Simulation

To put it simply, a vehicle's "electromagnetic environment is far more severe" than that of home appliances. First, the voltage systems are different. Home appliances use just one AC system of 100V/200V, but vehicles have a 12V battery, a 48V mild hybrid system, and 400V/800V high-voltage drive batteries coexisting within the same vehicle body.

Furthermore, there are over 100 ECUs (Electronic Control Units), and the total length of wire harnesses exceeds 2km. All of these are emitting and receiving electromagnetic waves within a single metal body enclosure. It's like 100 people, each holding an instrument, crammed into a 3-tatami mat room.

The automotive CISPR 25 Class 5 has limit values that are more than 10dB stricter compared to the home appliance CISPR 32. Moreover, the ISO 11452 immunity tests require withstanding electric field strengths of 200V/m. Compared to home appliances, which typically only need to handle 3-10V/m, you can see the difference is orders of magnitude.

The reason is simple: if a vehicle malfunctions due to EMC issues, people can die. In fact, there were reported cases in the 1990s of cruise control malfunctioning near radio towers. That's why ISO 11452-2 requires 200V/m at the vehicle level and over 100V/m at the component level. Because it's directly linked to safety, the requirements are orders of magnitude stricter than for home appliances.

CISPR 25 is a standard for "protecting vehicle-mounted receivers from the vehicle's own noise." It covers the frequency range from 150kHz to 2.5GHz and sets limits for both conducted emissions (noise traveling through harnesses) and radiated emissions (noise radiated into space).

There are Classes 1 through 5, with higher numbers being stricter. Almost all major OEMs require Class 5. Here are some specific values:

Exactly. A common failure in the field is not meeting Class 5 in the AM band. Why? Because harmonics from the PWM switching frequency of DC-DC converters or motor drives often fall right into the AM band. For example, the 50th harmonic of a 20kHz PWM is 1MHz, right in the middle of the AM broadcast band. With the increased inverter output in EVs, this problem has become more severe.

That's correct. ISO 11452 is the international standard for immunity testing, specifying test methods by part:

• ISO 11452-2: Radiated immunity in an anechoic chamber (20-2000MHz, up to 200V/m)
• ISO 11452-4: BCI method (Bulk Current Injection, injecting current into harnesses via a clamp)
• ISO 11452-5: Stripline method (board-level testing)
• ISO 11452-8: Magnetic field immunity
• ISO 11452-9: Near field from portable transmitters

BCI (Bulk Current Injection) is actually a very efficient test method. A ferrite core is attached to the harness as a coupling clamp, and RF power amplifiers inject current from 1MHz to 400MHz through it. While the radiated method requires a huge anechoic chamber to apply an electric field to the entire vehicle, BCI allows you to identify harness vulnerabilities on a bench.

In simulation, this BCI test is reproduced using an MTL (Multi-conductor Transmission Line) model. The common mode and differential mode currents flowing through each conductor of the harness are separated and evaluated.

The basics are indeed Maxwell's equations themselves. In time-domain form:

However, what's characteristic of automotive EMC is that the frequency range handled is extremely wide. It spans over five orders of magnitude, from 150kHz (wavelength 2km) to several GHz (wavelength 10cm). Therefore, a single solution method cannot cover everything. Quasi-static approximations can be used at low frequencies, but full-wave analysis is essential in the GHz range.

Good question. Think in terms of the ratio of wavelength to vehicle size:

• 150kHz (λ=2000m): Vehicle is 1/400 of the wavelength → can be treated as a lumped-parameter circuit
• 30MHz (λ=10m): Vehicle is 1/2 wavelength → resonance region, most troublesome
• 1GHz (λ=30cm): Vehicle is 17 times the wavelength → optical region, ray-tracing-like approximations are possible

Especially around 30MHz, the vehicle body acts as a half-wavelength resonator, and harnesses radiate efficiently as antennas, making it the most difficult frequency band for EMC countermeasures.

Ideally, yes, but in reality, the vehicle body is full of doors, windows, vents, and harness penetration holes, making it a very leaky shield. Shielding Effectiveness SE (dB) is:

Here, $A$ is absorption loss, $R$ is reflection loss, and $B$ is the multiple reflection correction term. The absorption loss of a metal plate uses the skin depth $\delta$:

For example, for a 0.8mm thick steel plate ($\sigma = 6.99 \times 10^6$ S/m):

• 1MHz: $\delta$ = 0.19mm → $A$ = 36dB → sufficient shielding
• 100kHz: $\delta$ = 0.60mm → $A$ = 11.5dB → somewhat insufficient

However, the real problem is not the plate thickness but the apertures. If there is a slot (gap) longer than λ/2, the shielding effectiveness drops sharply. Window glass is completely open unless it has a conductive coating, and door gaps become continuous slots several mm long. Accurate modeling of these apertures is key in simulation.

In automotive EMC, a single solver cannot cover the entire frequency range, so selecting the appropriate solver for the problem is fundamental. Let's compare the four main methods:

FDTD is the most popular for vehicle-level radiated EMC analysis. It updates Maxwell's equations on a Yee grid in time steps:

The strength of FDTD is that a single time-domain calculation yields S-parameters for all frequencies via FFT. Since CISPR 25 covers 150kHz to 2.5GHz, FDTD, which can excite in the time domain and obtain results for the entire band at once, is efficient. However, accurately representing thin wires like harnesses in an FDTD grid requires either making the mesh extremely fine locally or using an approximation like the Thin Wire Model.

Because a vehicle's wire harness is the most efficient antenna. There are 2km of conductors snaking through the vehicle body. They function as "unintentional antennas" across all frequency bands. It's said that over 70% of CISPR 25 failures are due to noise via harnesses.

Harness modeling uses MTL (Multi-conductor Transmission Line) theory. The telegrapher's equations for an n-conductor harness bundle are:

Here, $[\mathbf{R}]$, $[\mathbf{L}]$, $[\mathbf{G}]$, $[\mathbf{C}]$ are n×n distributed parameter matrices. $\mathbf{V}_F$, $\mathbf{I}_F$ are forcing terms from external electromagnetic fields (Agrawal formulation). Important points are:

• Crosstalk: Off-diagonal elements of [L] and [C] represent inter-conductor coupling.
• Shielded wires: Modeled by the transfer impedance $Z_T$ of the outer conductor (shield).
• Twisted pair: Mode conversion according to twist pitch is considered.

Transfer Impedance $Z_T$ (Transfer Impedance) is a parameter that indicates how much noise voltage is induced inside a shield by current flowing on its outside.