This article briefly covers the pitfalls of proper filtering. We’ll look at some of the reasons why the filters you may think will work often do not when they’re put into real circuits. So… let’s talk about why filters fail.
Have you ever spent countless hours researching and locating what you believed would be the best possible filter with the best possible performance for your specific needs? And then discovered that once installed, it barely suppresses any RF emissions? One reason for this lack of performance could be that the filter manufacturer has tested the attenuation characteristics for common-mode (CM) or differential-mode (DM) noise and it’s not clear from their specifications which one they used. If your emissions problems are mainly CM, yet the filter’s attenuation is specified for DM, you’ll have issues with successful implementation of the filter.
Non-ideal Test Method
Another problem could be the related to the standard used to test the performance of the filter (usually MIL-STD-220). Typically, filters are characterized by their insertion loss (IL), expressed in dBs. It’s a measure of the load reduction at the given frequency due to the insertion of the filter. The IL of a filter is dependent on the source and load impedances and shouldn’t be stated independently of terminal load/source impedances, but often is according to MIL-STD-220. The measuring instruments, source and load impedances, input attenuator and other components are specified as having an ideal 50Ω characteristic impedance. It’s rare to have something like a power supply input circuit having this same ideal 50Ω impedance. The load impedance a filter really sees isn’t going to precisely match 50Ω. Also, the input attenuator has a series impedance which can dampen out any resonances. This is a problem because the attenuator used in testing isn’t present in the end-product.
Test Current versus Use Current
Applied current during the test is another issue. The test method doesn’t require current to flow in the filter during testing, Therefore, it won’t match the circuit the filter is intended for, no matter what. The value of inductance in the filter may be different if DC current is flowing. When used outside of its specified current range, a choke can saturate, leaving it unable to supply its original intended impedance.
For these reasons, a perfect filter testing situation is one that doesn’t necessarily follow the standard method but one that is tailored for the specific EMI test supply impedance and utilizes the actual switching power supply planned for the product and operated at the expected current draw. A filter’s insertion loss or attenuation characteristics should be established at no load and full load current levels to provide the best results and information to potential users.
Unshielded Filter Elements
Unshielded filter elements can also cause problems. When filter components are unshielded and mounted on a PCB containing noise sources like switching power supplies or fast rise-time digital logic circuits, the noise will often couple to both filter components and the input connections to the filter. This unwanted crosstalk reduces the filter’s attenuation capabilities partially or even completely. A similar situation can occur when input/output power lines to the filter are run too close together. This problem can be mitigated by shielding the power line filter and mounting it on the equipment enclosure wall with the input power connector mounted to the filter enclosure. Keeping input/output connections separated far apart from one another will also help.
Filtering I/O Signal Lines
When adding low-pass filters to I/O signal lines, you may notice that the filter does not reduce emissions as you had expected. The problem could be that CM noise is present on each line and the ground (return) path. Note that CM noise current flows equally on all lines, including the ground path. In this scenario, if a capacitor is used to suppress CM noise it may only make things worse because it will carry the noise from the dirty digital ground to the clean signal lines. For better results, try removing the capacitor or connect the noisy digital ground to a clean (chassis) ground. In this instance, a common-mode choke may be a better solution than a capacitor to ground solution.
Don’t forget about the possibility of poor filter high-frequency response due to parasitics. We often overlook the fact that a low-pass filter’s nearly ideal low and mid-frequency response won’t continue upward in frequency. Because of parasitic capacitance, a filter’s real attenuation can drop off significantly at higher harmonics of switching frequencies (on the order of several megahertz). Look for ways to reduce these parasitics. Select components with low equivalent series resistance (ESR) and keep leads short, fat, and flat.
In real life, filter components exhibit tolerance, saturation, parasitics, and coupling issues. With careful forethought, planning and proper knowledge, these issues can be considered and mitigated so that you’ll no longer wonder why you filter isn’t working as expected. Wishing you the best of luck your future filtering efforts!
References and Further Reading
- In Compliance Magazine. (2018, November 12). What Every Electronics Engineer Needs to Know About: Filters.
- Have You Considered Everything in the Design of Your EMI Filter?, http://www.schaffner usa.com
- Electronic Design. (2016, July 26). 11 Myths About EMI/EMC.
- Eadie, A., EMC Fast Pass. The Top EMC Failures and Tips from 5 EMC Consultants.