CISPR EMC Standards and Compliance Simulation

It's determined by the product category. Home appliances follow CISPR 14, IT equipment follows CISPR 32 (formerly 22), automotive follows CISPR 25, and industrial equipment follows CISPR 11. The key at the design stage is to predict the response of the Quasi-Peak (QP) detector through simulation. Adding filters after failing measurements in real tests incurs significant rework costs.

Comité International Spécial des Perturbations Radioélectriques, the International Special Committee on Radio Interference. It's a sub-organization of the IEC (International Electrotechnical Commission), established in 1934. It originated to protect radio broadcasting from electrical equipment noise and now forms the foundation of EMC standards for all electronic devices.

The standards issued by CISPR are broadly divided into two series:

• Basic Standards (CISPR 16 series): Define measurement instrument specifications, measurement methods, and statistical evaluation techniques. The IF bandwidth and charge/discharge time constants of the QP detector are also specified here.
• Product Standards (CISPR 11/14/25/32, etc.): Define emission limits and measurement conditions for each product category.

Yes, CISPR 22 (emission for IT equipment) and CISPR 24 (immunity for IT equipment) were integrated into CISPR 32 and CISPR 35 respectively. CISPR 32 is the successor standard that expanded its scope to include multimedia equipment. Old certification reports might say CISPR 22, but new certifications use CISPR 32. For example, Wi-Fi routers and game consoles need to comply with CISPR 32 Class B.

Not simple at all. Limits are determined by four axes: "Frequency Band" × "Detection Method" × "Class" × "Distance". For example, the radiated emission limits for CISPR 32 are as follows:

The key point is that Class B is about 10dB stricter than Class A. Residential environments require lower radiation levels to protect radio and TV reception. A 10dB difference is about 3.16 times in electric field strength and 10 times in power.

Yes, they are. Conducted emission measures the noise voltage leaking out through the power line. For CISPR 32:

It's important to note that limits are set for both QP and Average values. This means you must clear the limits for both detection methods. Simulation also needs to predict both.

QP (Quasi-Peak) detection is a unique CISPR detection method. Its biggest characteristic is that the output changes according to the signal's repetition rate (PRF: Pulse Repetition Frequency). It reflects the human auditory characteristic that "continuous noise is perceived as more annoying than occasional pulses."

Comparing the three detection methods:

They are specified in CISPR 16-1-1. They vary by frequency band:

The charge time constant is much shorter than the discharge time constant. This means "fast charge, slow discharge". If pulses come frequently, the discharge can't keep up and the output rises. Conversely, if the PRF is low, discharge progresses and the output drops. This asymmetric response is the essence of QP detection.

The input-output relationship of the QP detector can be written mathematically as follows. For the output envelope $V_{env}(t)$ of the IF stage:

Exactly. The starting point for EMC simulation is Maxwell's equations:

Two points are particularly important in the context of EMC:

• Radiated Emission: Solve Maxwell's equations from the current distribution on the board to find the electric field strength at 10m (or 3m). For example, if a board trace is 30cm long, it matches the wavelength of 1GHz (30cm) exactly and can act as an efficient antenna.
• Conducted Emission: Solve for common-mode/differential-mode currents on power lines using a combination of transmission line theory and Maxwell's equations.

Scary? This is the majority of EMC problems. Designers lay out traces intending to "transmit signals," but for high-frequency components, they can become "radiating antennas." Especially square waves like clock signals have rich harmonic components, with spectra spreading up to 10-20 times the fundamental frequency. A 100MHz clock can cause issues up to 1GHz-2GHz.

It depends on the nature of the problem. Roughly speaking:

For EMC, FDTD (or its generalization, FIT) is overwhelmingly common. There are three reasons:

• Broadband analysis in one go: CISPR standards cover a wide band from 9kHz to several GHz. FDTD yields results for all frequencies at once by performing an FFT on a single time-domain simulation.
• Strong with large-scale structures: Can handle models including the entire enclosure + board + cables. FEM or MoM would be memory-intensive at this scale.
• Direct transient response: Can naturally handle non-stationary excitation sources like switching power supply ON/OFF.

However, FDTD is based on structured meshes, so it has limitations in approximating curved surfaces. For modeling fine connector shapes, FEM might be more suitable in some cases.

It's actually quite involved; we need to implement the following processing chain:

1. Time waveform → FFT: Perform FFT on the simulation result's electric field (or voltage) time waveform to obtain the frequency spectrum.
2. IF Bandwidth Filter: At each measurement frequency point, apply a bandpass filter with the CISPR-specified IF bandwidth (e.g., 9kHz for Band B).
3. Envelope Detection: Calculate the amplitude envelope of the filtered signal.
4. Charge/Discharge Simulation: Process the envelope with an asymmetric RC circuit using charge time constant $\tau_c$ and discharge time constant $\tau_d$.
5. Meter Time Constant: Smooth the final QP reading value with a critically damped response.

Step 4 discretization is sufficient with the forward Euler method. The recurrence formula with sampling time $\Delta t$: