What People (and AI Agents) Get Wrong About Common-Mode Filtering
If you ask an AI chatbot how to design a filter to reduce common-mode emissions from your product, the advice you get back is often incomplete, and sometimes simply wrong. The same is frequently true of the guidance in filter-component application notes and on manufacturers’ websites.
There’s no shortage of advice on common-mode filter design, but much of it leaves out the context that determines whether a given approach will actually work. Common-mode filters, for instance, are often characterized by insertion-loss measurements in 50-ohm systems. Those measurements are easy to make and to reproduce, but a 50-ohm system looks nothing like most real-world applications. “Before and after” demonstrations can be misleading in a different way: if you start with a poorly laid-out board, almost any new layout with a few added components will look like an improvement.
It also helps to remember what application notes are for. They’re written to help engineers apply a particular component, not to teach filter design from first principles, so the surrounding design context is often simplified or left out. That material then works its way into AI training data and gets repeated on social media, and the nuances tend to fall away. The result is a subject that feels far more confusing than it needs to be. Four questions come up again and again:
- When is a common-mode choke appropriate? It’s easy to come away with the impression that a choke is almost always called for. In practice, chokes do play an important role in many balanced, unshielded power and signal interfaces, but they’re a poor fit for a lot of the applications where you’ll see them recommended.
- Should Y-capacitors go on both sides of the choke, or on one side only? You’ll find both configurations recommended, sometimes within the same source, usually without a clear explanation of which one is right in a given situation.
- Should the ground plane extend underneath the choke? Should it extend everywhere? Should a hole be cut in the plane beneath the choke? Or should the plane stop where the choke begins? You’ll find each of these advocated somewhere.
- Should the choke use a low-loss or high-loss ferrite? Is loss in the ferrite material desirable or not? And if it is, why don’t data sheets tell you how much loss a given choke provides?
Despite all of this, common-mode filtering is actually fairly straightforward once a few ideas are in place. Most of the confusion traces back to three points that often get left out:
- There are two different definitions of common-mode current. In one, the current flows out on the signal or power conductors and returns on a separate ground or return conductor. In the other, it flows out on all of the conductors together and returns as displacement current through the surrounding environment. Both can be reduced by common-mode filtering, but the filter design is very different in each case.
- It’s commonly assumed that a common-mode choke impedes common-mode current and passes differential-mode current. That’s true in a balanced system, but not in general. Common-mode chokes are balanced components. In an unbalanced system, a choke promotes mode conversion: it can put common-mode noise into the signal path and turn signal currents into a common-mode voltage that drives conducted and radiated emissions.
- As noted above, common-mode chokes are usually evaluated in 50-ohm systems. Designing a filter that looks good in a 50-ohm system is very different from designing one that will attenuate conducted or radiated emissions in a real product.
If you’d like to see how common-mode filtering should be done, I’d recommend our website tutorial at learnemc.com/cm-filtering. Once you understand the nature of the common-mode source in a particular application, designing an appropriate filter is relatively straightforward.
So how do you decide what to trust in all the available guidance? Let’s revisit the four questions from the start of the article.
When is a common-mode choke appropriate?
In most cases, a common-mode choke belongs in a system where the source, load, and transmission path are all electrically balanced. Good examples include most AC power inputs and high-speed differential signaling. A choke is generally not the right tool for an inherently unbalanced interface such as unbalanced DC power or a single-ended signal. In an unbalanced system, a choke converts some of the differential signal into common-mode noise and injects common-mode noise back into the signal path. There is one important exception: when a strong common-mode source (such as a large heatsink) drives a board that has no metal chassis, you can use Y-capacitors to make the input look balanced at the noise frequencies and then add a lossy choke to reduce the common-mode current. (More on the lossy choke below.)
Should Y-capacitors go on one side of the choke or both?
If you’re filtering common-mode current that returns on a ground plane, the Y-capacitors belong on both sides of the choke. This creates a pi-filter that can be very effective, and it’s a common arrangement on AC (or isolated-DC) power inputs. If instead you’re filtering common-mode current that returns as displacement current, the Y-capacitors belong on the product side of the filter only. They should not be placed on the cable side.
Should the ground plane extend underneath the choke?
If there are Y-capacitors on both sides, the answer is yes. The ground plane is needed to help the common-mode currents return to their source. If the common-mode current is returning as displacement current, the answer is no. In that case, a continuous ground plane can carry the current around the choke and keep it from doing its job.
Should the ferrite be low-loss or high-loss?
It depends on what the choke is being asked to do. In a power-line filter, where the choke acts as an inductor in a pi-filter, a relatively low-loss ferrite is generally fine, though some loss in the tens-of-megahertz range is welcome because it helps damp resonances. For a filter that has to block common-mode current returning as displacement current, high-frequency loss is essential. The choke’s inductive reactance doesn’t block that current; it only creates or shifts a resonance. A choke works in this situation only when it provides a significant amount of high-frequency resistance, much as a ferrite core or clamp does.
The bottom line
How and when to use a common-mode choke depends heavily on the application and on the nature of the common-mode source. When you come across general guidance that doesn’t account for these factors, treat it as a useful starting point rather than a finished recipe, and take the time to work out what the common-mode source in your design actually is.