When designing an EMI/EMC filter, the orientation relative to the source and victim is critical for high effectiveness. Can you save components in your filters?
If you have an EMI problem or you are failing EMC-conducted or radiated emissions tests, you will try to design a filter to solve the problem.
First question: Are you sure?
Remember: if you try to design a filter, it is because you will not try to kill the source. Please be sure this is your final decision. Usually to kill the source is my preferred option for any EMI/EMC problem.
But sometimes, as in our example, we cannot remove the source of the problem so we will try a filter. In EMI/EMC, we use low-pass filters to avoid high-frequency energy arriving to the victim.
Usually, you will start with one component (cheaper, smaller, etc.) filter: a shunt capacitor or a series inductor (Figure 1).
Note from general theory that you will obtain -20dB/dec (or -6dB/oct) in attenuation for each component included in your filter.
But, when to use the capacitor instead of the inductor?
With reactive components you are trying to filter using mismatching: you choose the capacitor to be low impedance and the inductor to be high impedance at EMI frequencies COMPARED with ZS and ZL terminal impedances.
So, the capacitor option will be especially effective if terminal impedances are high, and the inductor will be effective if terminal impedances are low.
Let me show you this idea with an example.
Think your terminal impedances are resistive ZS=5ohm and ZL=100ohm. You need to design a low pass filter with order 2 (-40dB/dec), so you will use an inductor and a capacitor.
How to orientate the filter?
Following the previous idea, you will use the shunt component (capacitor) in parallel with the higher impedance (ZL) and the series component (inductor) in series with the smaller impedance (ZS).
In Figure 2, we compare insertion losses with good vs bad orientation. There is more than 20dB in difference, while the slope in the attenuation band is -40dB/dec for both solutions!
I have seen the importance of this idea many times while reviewing designs in real products for companies or teaching.
For example, in Figure 3, the conducted emissions from a coffee machine were measured without a filter between 100 kHz and 30 MHz (red color). The original filter from this company was an LC network with 30uH on the LISN side and 100nF shunt capacitance on the DUT side.
The results were not as expected because EMI was over the limit in all the frequency ranges of interest (green color) on the left side of Figure 3.
When obtaining the result on the left, a typical decision is to add components to the filter (additional inductors and capacitors), trying to increase the insertion losses (remember 20dB/dec for each added component). In doing this, the filter will be bigger, more expensive, etc. (sometimes this is not possible because of available space and limited budget).
But, note what happens when I rotate the filter 180º as shown on the right side of Figure 3. With the same components and a good orientation, the product is now complying with regulations. Great!
Have you tried this idea in your products?
Perhaps your filter is more complex than necessary?
Finally, an important idea: for EMI engineers, terminal impedances are usually unknown so if you have no experience with the product, perhaps you will need to try both orientations to find the appropriate solution.
Homework:
- In our example, which is larger, the coffee machine’s output impedance or the LISN’s input impedance?
- When will you use a T-network (LCL) versus a PI-network (CLC)?
