Harmonizing Circuit Design and EMC DesignPart 18 EMC Calculation Methods and EMC Simulations (3): Trial Calculation of Radiated Emissions (RE)

Hello! I’m Inagaki, at ROHM.

In this 18th column, I will explain a calculation method and simulation for electromagnetic compatibility (EMC), which is the trial calculation of radiated emission (RE). This is the “ALSE method” in the CISPR25 standard relating to electromagnetic compatibility (EMC) characteristics for automotive products”.

CISPR (pronounced “sis-per”) is an acronym for “Comité International Spécial des Perturbations Radioélectriques” (English: International Special Committee on Radio Interference), which is subordinate to the IEC. ALSE is an acronym for Absorber-Lined Shielded Enclosure, and refers to a method of measuring electromagnetic noise radiated from a DUT (device under test: a semiconductor integrated circuit (LSI)), a wire harness, or the like.

Targets for calculations may be automotive batteries, artificial mains networks (ANs), wire harnesses, DUTs (devices under test), ground planes, and the like. Calculation concepts conform to IEC 62433 standards, supporting data assimilation and noise reduction. As analysis methods, circuit analysis, electromagnetic field analysis, and numerical analysis are employed.

Calculations are explained in order. In trial calculations, two-stage processing is performed, with (shell) scripts used to automate both optimization and prediction. The first stage, optimization, employs the following calculation procedure.

1. ① IA models (electromagnetic interference models) are created using PWL (piecewise linear) waveforms for the power supply current and load current of the semiconductor integrated circuit (LSI). These are found by circuit analysis (transient analysis).
2. ② Since data assimilation techniques are used, CISPR25 ALSE method measurement values are obtained for a state in which EMC measures have not been implemented.
3. ③ Noise elimination processing (upper-limit envelope processing) is performed for both ① the IA models (PWL waveforms) and for ② measurement values. These are found by numerical analysis.
4. ④ Automotive batteries and artificial mains networks (ANs) are described as SPICE circuit networks, wire harnesses and ground planes are subjected to electromagnetic field analyses, and CAD data is created.
5. ⑤ By using electromagnetic field analysis (MoM, Method of Moments) for ③ and ④, the radiated emission (electric field dBμV/m) is found for one frequency.
6. ⑥ Up to this point, only calculated values have been obtained; the differences between these and the measurement values of ③ are taken, and calculated values are corrected.
7. ⑦ The same calculations are repeated for the required frequencies (for example, the N harmonics of a switching frequency, and so on), and the transient analysis results are graphed (frequency axis); in addition, limiting values are also plotted (same format as for AC analysis results). Hence the optimization calculation results will result in substantially complete agreement of the calculated values and measurement values.

In the second-stage prediction calculations, the following calculation procedure is used.

1. ⑧ IA models (PWL waveforms) are found for the semiconductor integrated circuit (LSI) for which EMC measures have been implemented. When a silicon chip is being redesigned, or when an application circuit is being modified, the object is to predict the extent to which radiated emission (RE) is reduced as a result of the changes to the power supply current, load current, or the like.
2. ⑨ The IA models (PWL waveforms) of ⑧ are subjected to noise elimination processing (upper-limit envelope processing).
3. ⑩ The IA models (PWL waveforms) of ③ are substituted for the IA models (PWL waveforms) of ⑧, and the radiated emission (electric field dBμV/m) for one frequency is determined.
4. ⑪ Using the difference found in ⑥ optimization, the calculated values of ⑩ radiated emission are corrected.
5. ⑫ Calculations similar to those of ⑦ are performed for multiple frequencies. Using the results, it is possible to judge whether or not the prediction is in conformance with the CISPR25 ALSE method.

The above is a summary of the calculations; below are graphical illustrations of a representative example of such calculations.

In actual design processes, circuit design of semiconductor integrated circuits (LSIs) is performed on an extremely tight schedule, and it is very difficult to devote much time to EMC measurement and verification or to EMC predictive calculations. Here, by using a comparatively simple calculation model, the time required for calculations is shortened, and by obtaining conformity with measurement values, practicable calculation precision is achieved. Moreover, by using (shell) scripts, electromagnetic field analyses and numerical analysis can be run in the background, rendering complex operations unnecessary. In terms of hardware also, a powerful PC is not required (in this trial calculation example, the CPU required about 15 minutes, and memory used was under 100 MB). On the other hand, there are also methods in which more detailed device modeling is performed and supercomputers or the like are used freely to perform electromagnetic field analyses. Both methods are effective, but when limiting the scope to semiconductor integrated circuits (LSIs), I think that these trial calculations are advantageous.

Thank you for your kind attention.