Harmonizing Circuit Design and EMC DesignPart 20 EMC Calculation Methods and EMC Simulations (5): Trial Calculation of Conducted Immunity (CI)

Hello! I’m Inagaki, at ROHM.

This 20th article is again about electromagnetic compatibility (EMC) calculation methods and simulations. This is the fifth article focusing on the trial calculations of conducted immunity (CI). The discussion relates to electromagnetic compatibility (EMC) characteristics of equipment for automobiles, and so involves the Harness Excitation (HE) method stipulated in the ISO 11452-4 standard.

The written standards for the HE method describe the Bulk Current Injection (BCI) method and the Tubular Wave Coupler (TWC) method. In both cases electromagnetic noise is applied to a wire harness (a model of electrical wiring in an automobile), and the malfunction level of the DUT (device under test) is judged. The BCI method is a testing method in which electrical current noise ranging from 0.1 MHz to 400 MHz is applied via a current injection probe; in the TWC method, electrical power noise at 400 MHz to 3 GHz is applied via a coupler. In this article, I will explain calculated predictions using the BCI method.

The BCI method as in ISO 11452-4 is a testing method that is very different from the DPI method of IEC 62132-4, explained in the previous article. In the DPI method, the extent of noise tolerance of the DUT (device under test) is determined by precisely measuring the power value while raising and lowering the traveling-wave power. On the other hand, in the BCI method a current noise of for example 200 mA is applied, and tests are conducted to determine whether malfunctions occur for all frequency bands. Hence test results yield pass/fail judgments for each test frequency.

When performing calculated predictions, the DPI method is probably easier to use. When pass/fail results are measured, as with the BCI method, it is troublesome to decide on the representation to use in circuit analyses (SPICE simulations). The following explanation will include these matters as well. Incidentally, these BCI method calculated predictions are performed without using electromagnetic field analysis. Phenomena that can be solved for in circuit analyses require less time for calculated predictions (the CPU time in this case was about two minutes).

The objects of calculations include automobile batteries, line impedance stabilization networks (LISNs), wire harnesses (3-wire harnesses: power supply wire, ground wire, output wire), current injection probes, loads, components for EMC measures (here, capacitive elements C), DUTs (LSI models), current noise sources, standard conformance testers, and the like. Here as well, a method is explained for creating computer models (simulation models) based on measured values.

The method is explained in order below. In trial calculations, two-stage processing is performed, with (shell) scripts used to automate each of the stages—the IB (malfunction threshold) model extraction in the first stage, and the calculated prediction in the second stage. The IB (malfunction threshold) model extraction of the first stage is performed using the following calculation procedure.

■First Stage: IB (Malfunction Threshold) Model Extraction

1. ① Similarly to the 19th article, the calculation circuit diagram is first created from the above-described object of calculation. The measurement circuit is connected as-is for use as the calculation circuit. The current injection probe and the wire harness are represented by a circuit analysis transformer coupling. This is a 3-wire representation, and so by using three dependent connections for the transformer, current noise can be applied uniformly to each wire. The impedance across LSI pins of the DUT (device under test) is measured using an LCR meter, and the electrical characteristics are used to create an LCR (passive element) circuit (this is important!).
2. ② Next, the measurement results are converted into numerical values so as to facilitate circuit analysis. For example, suppose that, when 200 mA of current noise is applied, there is a pass frequency and a fail frequency. Suppose that the measured value for the pass frequency is set to 200 mA, and the measured value for the fail frequency is set to 100 mA (one-half that of the pass frequency, a value that can be modified as appropriate according to empirical results). By doing so, a file of numerical values that makes circuit analysis easier can be created (this too is important!).
3. ③ The current value obtained in ② is applied as a signal source, and circuit analysis (transient analysis) is performed to calculate the IB (malfunction threshold value) (the current reaching the LSI pin) for one frequency. If a current greater than the calculated current value arrives, the DUT (device under test) malfunctions, and if a smaller current arrives there is no malfunction; in other words, this is the threshold value for malfunctioning.
4. ④ Step ③ is repeated for all frequencies. If execution is set up to enable repetition using (shell) scripts and macros, even when there are many analyses to perform, there should be no problem performing them in a single execution. The result is saved in a file, and plotted along the frequency axis to obtain a graph like the following.

Example of IB (malfunction threshold) model calculation

It should be noted that the IB (malfunction threshold) model has limited, intrinsic current values that depend on the calculation circuit diagram and the LSI model (impedance characteristics). Used together with the calculation circuit diagram and LSI model, it becomes a computer model (simulation model) that can reproduce malfunctions at the time of measurement. (Here, in response to customer requests, frequencies from 1 MHz to 1 GHz are supported.)

In calculated predictions in the second stage, the following procedure is used.

■Second stage: Calculated Prediction

1. ⑤ A circuit for calculated prediction is created. The difference with the circuit for IB (malfunction threshold) model extraction is the addition of a malfunction tester (comparator). The current reached in the LSI and the IB (malfunction threshold) are compared by the malfunction tester (comparator).
2. ⑥ Next, the signal source for current noise is set to a damped oscillation waveform. In SPICE, creating a damped oscillation waveform is relatively simple. It is the same as the current-converted waveform used in the DPI method of the preceding article.
3. ⑦ In the circuit analysis (transient analysis), when the analysis is executed for one frequency, we see that the LSI transitions from, for example, a malfunctioning state to a non-malfunctioning state. This is repeated for all the frequencies.
4. ⑧ The current value that is the threshold for malfunctioning using a damped oscillation waveform becomes the calculated prediction value to be sought. The graph on the lower left shows calculated predictions using the same circuit that was used in IB (malfunction threshold) extraction prior to EMC countermeasures. The measured and calculated values essentially coincide. There are small calculation errors at lower frequencies; these occur because of the extremely small values for the IB (malfunction threshold) model. At high frequencies, the calculated value of 100 mA that is the fail criterion is accurately reproduced.
5. ⑨ The figure on the lower right shows the results of calculated predictions with an EMC countermeasure circuit added (here, two 3-lead capacitors C are used; these capacitors feature smaller parasitic inductances L compared with ordinary 2-lead capacitors). The calculated prediction values, at 200 mA or higher, indicate that there should be conformance to standards. In particular, it should be noted that the portion that was at 100 mA prior to the EMC countermeasure has been improved to be equal to the 200 mA of other parts of the plot. We see that, as explained above, the added capacitors are effective in improving the LSI malfunction level.

This example confirms malfunction resistance at 200 mA or greater over the frequency range from 1 MHz to 1 GHz. The results verify improvement of malfunction resistance over all frequencies, regardless of the blue lines in the graphs (standard limit values).

(In the original ISO 11452-4 standard, out of the entire range from 1 MHz to 400 MHz, the limit values are somewhat lower for 1 MHz to 3 MHz and for 200 MHz to 400 MHz. Limit values are also set for 0.1 MHz to 1 MHz.)

Thank you for your kind attention.