I still remember the first time I spent three days chasing a 60Hz hum in an audio preamp, only to realize the problem wasn't the power supply itself — it was a missing high-pass filter at the input stage. The noise had always been there; I just didn't have the right filter in place to block it.

That experience taught me more about AC filters than any textbook did. Filters aren't just theory — they're the difference between a circuit that works and one that works reliably.

This guide is for engineers and buyers who've been burned by filter-related problems. I'll skip the textbook definitions and go straight to the practical stuff: how to pick the right filter, how to calculate what you actually need, and how to troubleshoot when things go wrong.

An AC filter is a circuit that decides which frequencies get through and which get stopped. It uses the frequency-dependent behavior of capacitors and inductors to do this:

A capacitor lets high frequencies through easily (low impedance at high frequency) but blocks low frequencies
An inductor does the opposite — it chokes off high frequencies while letting low frequencies pass

You combine these with resistors to control exactly where the cutoff happens.

The three things that actually matter

Most engineers memorize these three parameters and move on. But understanding why they matter is what separates a working design from a marginal one.

Cutoff frequency (fc) is where the filter starts working — the output power drops to half of the input. At that point, the voltage is down to about 71%. If you're designing an audio filter and your cutoff is too high, you'll be cutting off the low notes.

Roll-off slope tells you how fast the filter works past the cutoff. A first-order filter drops at 20 dB per decade. That sounds steep until you actually calculate what it means at 10x the cutoff frequency — you're only down 20 dB, which is often not enough in practice. Most real designs need at least second-order (40 dB/decade) or they end up adding extra stages.

Passband ripple is the variation in the passband. A Butterworth filter is completely flat — no ripple. A Chebyshev filter has ripple in exchange for a sharper roll-off. If you're filtering audio, flat is usually better. If you're selecting a channel in a communication system, you might tolerate some ripple for steeper selectivity.

One thing many people get wrong

A lot of people think DC filters and AC filters are the same thing. They're not. A DC filter (like a bypass capacitor) smooths ripple on a fixed voltage rail. An AC filter has to behave correctly at every frequency in its operating range — it's not just a smoothing capacitor tacked on somewhere.

Low-Pass Filter (LPF)

Passes everything below the cutoff. Stops everything above.

You see this everywhere after a rectifier stage, before an ADC, or in any circuit where switching noise needs to go away but the baseband signal needs to stay.

The simplest LPF is just a resistor and capacitor in series:

fc = 1 / (2*pi*R*C)

Quick example: if your resistor is 1k and your capacitor is 100nF, your cutoff is around 1,592 Hz. Frequencies above that get attenuated. Below it, they're unaffected.

Common mistake: people pick a cutoff frequency and forget to check what happens at 10x that frequency. If you need 40 dB attenuation there, a first-order filter won't cut it — you'll need two stages.

High-Pass Filter (HPF)

Does the opposite. Passes high frequencies, blocks low frequencies and DC.

The classic use case is removing 50/60Hz power line hum from audio signals. It also shows up in AC coupling between circuit stages — you want the AC signal through, but not the DC bias from the previous stage.

Band-Pass Filter (BPF)

Only lets a specific range through. Everything outside gets blocked.

Band-pass filters are where things get interesting. The resonant frequency of an LC circuit determines the center frequency:

f0 = 1 / (2*pi*sqrt(L*C))

A BPF used in a radio receiver has to be precise — if your channel spacing is 25 kHz, you can't afford a filter that's too wide or you'll bleed into adjacent channels. But if it's too narrow, you'll lose the signal edges and distort modulation.

Band-Stop / Notch Filter

The opposite of a band-pass. It blocks a narrow range and lets everything else through.

The 60Hz notch filter is one of the most practical filters in audio work. Power line hum is stubborn — it couples through everything: power supply, ground loops, even the air. A well-designed notch filter at 60Hz (or 50Hz, depending on your mains frequency) can pull that hum down by 30-40 dB without touching the music.

EMI Line Filter

This one is different from the other four. While the others are typically designed as RC or LC networks, an EMI line filter is almost always a pre-built module installed at the AC power inlet.

It targets electromagnetic interference flowing into or out of your equipment through the mains wiring. And unlike signal filters, EMI filters have regulatory implications — most markets require EMC compliance that can't be achieved without them.

If you're building anything that plugs into AC mains, this is the filter you need to understand.

The formula is simple: fc = 1 / (2*pi*R*C)

But most engineers get into trouble not because they can't calculate — it's because they don't think about what happens past the cutoff.

Filter Order What It Means Real-World Implication
1st order -20 dB/decade Only useful if your noise is far above fc
2nd order -40 dB/decade Good for most applications
3rd order -60 dB/decade When you need serious attenuation
4th order -80 dB/decade Two 2nd-order stages cascaded

Example 1: Removing 60Hz hum from audio

You want an audio circuit that passes everything above 80Hz and blocks 60Hz hum.

Using the HPF formula, pick C = 100nF:
R = 1 / (2*pi*80*100e-9) ≈ 19,900 ohm

Use 20k (standard value). Your cutoff ends up at 79.6Hz. At 60Hz, you're getting meaningful attenuation. At 100Hz (your audio), you're essentially unaffected.

That's a clean design. One capacitor, one resistor, and the hum is gone.

Example 2: Blocking 100kHz switching noise

You have a switching power supply that's clean up to 20kHz but noisy up past 100kHz. You want the audio to pass through undisturbed.

Pick your cutoff at 30kHz (leaves some margin so audio doesn't roll off too early). With C = 10nF:

R = 1 / (2*pi*30000*10e-9) ≈ 530 ohm

Use 560 ohm (E24 standard). Actual cutoff: 28.4kHz.

At 100kHz, a first-order filter gives you about -10.5 dB of attenuation. That might be enough. If it's not, add a second identical stage — that gets you to roughly -21 dB, which is usually plenty for this kind of noise.

I worked on a motor drive project once that failed EMC testing three times before someone realized the EMI filter was undersized. The filter looked correct on paper, but it was saturating under load current and doing almost nothing above 200kHz. We had to replace it with a filter rated for 50% more current. It passed on the first try after that.

That's the lesson: EMI line filters have to be sized for real operating conditions, not just the datasheet spec.

Common-mode vs. differential-mode noise

Before you pick a filter, you need to know what kind of noise you're fighting.

Common-mode noise travels on both conductors at the same time, returning through ground. Think of it as noise picked up equally on both lines. Y capacitors and common-mode chokes (CMC) handle this.

Differential-mode noise travels on the hot-neutral pair, following the normal current path. X capacitors and the leakage inductance of a common-mode choke deal with this.

Most switching power supplies generate both. If you only address one type, you'll only solve half the problem.

What an EMI filter actually contains

A proper EMI line filter typically has:

1.X capacitors (X1/X2 class) sit across the line-neutral connection. They catch differential-mode noise. They're safety-rated — when they fail, they fail open rather than shorting. Don't substitute standard film caps here.
2.Y capacitors (Y1/Y2 class) connect line-ground and neutral-ground. They handle common-mode noise. These have strict limits on leakage current — too much and you'll fail safety testing even if the EMC passes. Handheld devices typically need under 0.25 mA of leakage current through Y caps.
3.Common-Mode Choke (CMC) is the heart of the filter. Two windings on a shared magnetic core, wound in opposite directions. Differential current (your normal load current) produces equal and opposite flux, which cancels out — so the choke presents almost zero impedance to your circuit. But common-mode noise sees the full inductance of both windings and gets blocked.

The selection parameters that actually matter

Don't just pick a filter based on voltage rating. Here's what I check:

Rated current is non-negotiable. A filter rated for 10A that runs at 12A will get hot, age faster, and eventually stop working properly. Always size for at least 1.25x your expected continuous current. If the load is inductive (motors, transformers), size for 1.5x.

Insertion loss specs can be misleading. Manufacturers measure this in a 50 ohm system. Your actual circuit impedance is almost never 50 ohm. Ask for measured data in your specific configuration, or at least be conservative in your selection.

Operating temperature matters more than people think. A filter rated for 85°C ambient might be fine in an office environment, but inside an enclosed industrial enclosure running near full load, ambient can easily hit 60-70°C. That doesn't leave much margin.

Audio equipment

Audio is one of the most demanding applications for filters because the ear is extremely sensitive to both noise and frequency response errors.

The most common audio filter tasks: remove DC offset (use HPF with fc around 5-15 Hz), block 50/60Hz hum coupling (HPF set above the hum frequency but below the lowest audio frequency), and prevent ultrasonic noise from aliasing in ADCs (LPF at or below the Nyquist frequency).

For hum removal specifically: don't try to solve a grounding problem with a filter. Fix the ground first. A notch filter at 60Hz will work, but if the hum is getting in through a bad ground connection, you're just treating a symptom.

Power supplies

After a full-wave rectifier, you get 100/120Hz ripple (depending on your mains frequency). The capacitor after the rectifier smooths this. For a 10A load with 100mV ripple target: C = I / (2*f*DeltaV) = 10 / (2*120*0.1) = 417 uF. Use 470 uF.

The high-frequency switching noise is a different problem. That's where EMI filters or LC filters on the switching stage come in.

Variable Frequency Drives (VFD)

VFDs are notoriously dirty. They generate conducted EMI across a wide frequency range, from a few kHz up to 30 MHz or more.

The fix is straightforward: install an EMI line filter rated for the drive's full-input current, with at least 60 dB insertion loss at 150kHz. But don't forget the output side — the fast voltage edges from the inverter switching can damage motor winding insulation over time. A dV/dt filter on the output protects the motor.

Automotive
Every automotive electronics engineer learns this the hard way: automotive EMC requirements (CISPR 25, ISO 7637) are aggressive. Typical measures include ferrite beads on signal lines, X/Y capacitors at DC/DC inputs, and LC filters on motor drive outputs. The EV charging inlet almost always needs a dedicated EMI filter.

Problem: Filter worked fine, now it doesn't
The most common cause: electrolytic capacitor aging. Electrolytics lose 20-30% of their capacitance over 5-10 years, especially in warm environments. When the capacitor drifts, the cutoff frequency drifts with it.

I once diagnosed a "new EMI problem" in a 7-year-old power supply that turned out to be three aged electrolytics. The filter was still there; it just wasn't working at the right frequency anymore.

Fix: Replace aged electrolytics with film capacitors where stability matters. Film caps don't age.

Problem: EMI got worse after adding a filter
This usually means one of two things. Either you have an LC resonance (the filter's inductance and capacitance are resonating at a frequency in your noise band, making the noise worse instead of better), or the filter is saturating under load current.

Diagnose it: Put a network analyzer on the filter and check the insertion loss curve. If it shows a peak instead of a smooth attenuation roll-off, you have a resonance problem.

Fix: Add a small series resistor to damp the resonance (typically 1-10 ohm). Yes, it adds insertion loss, but it's better than making noise worse.

Problem: Filter is running hot
If the filter body is noticeably warm to the touch, you're either at or over its current rating, or the core is saturating.

Diagnose it: Measure the actual RMS current through the filter. Compare it to the datasheet rating. If you're within spec, check for asymmetric loading conditions that could be causing DC bias on the common-mode choke.

Fix: Oversize the filter. If you're at 15A continuous, use a 20-25A filter.

Problem: Equipment fails EMC testing with a filter installed
This usually means the filter is handling the wrong noise type. You added differential-mode filtering but the noise is common-mode, or vice versa.

Diagnose it: Use a current probe to measure the noise spectrum on each conductor separately. If the noise appears equally on both, it's common-mode. If it appears differentially, it's differential-mode.

Fix: Add the missing filter stage. For CM noise, add more Y capacitance and a larger CMC. For DM noise, add more X capacitance.

Q: Can I just use a low-pass filter to remove 60Hz hum from audio?

Not if you want to keep the music. A low-pass filter that blocks 60Hz will also block everything around 60Hz — and that's squarely in the middle of the audio band. You need a notch filter that targets 60Hz specifically, or a high-pass filter set well below 60Hz if the hum is coupling through DC offset or ground loops.

Q: How do I calculate the right cutoff frequency for a power supply filter?

Start by identifying the noise frequency you want to block. Then pick a cutoff that's 3-5x below that frequency for a first-order filter (so you have some attenuation margin), or 1.5-2x below for a second-order filter. Always verify what happens at 10x the noise frequency — if you need significant attenuation there, plan for additional filter stages.

Q: What's the difference between X and Y capacitors in an EMI filter?

X capacitors go across the line (line-to-neutral). They're there to catch differential-mode noise — noise that appears differently on the two power conductors. Y capacitors go from each conductor to ground. They're for common-mode noise — noise that appears equally on both conductors. X caps are sized for maximum capacitance within safety limits. Y caps are limited by leakage current requirements — too much Y capacitance and your grounding becomes a shock hazard.

Q: Does filter order really matter that much?

Yes, and it's one of the most underestimated factors in filter design. A first-order filter gives you -20 dB/decade. If your noise is at 100 kHz and you want it attenuated by 40 dB, you'd need the cutoff at 10 kHz. But if your signal extends to 20 kHz (audio), you've just filtered your signal. A second-order filter (-40 dB/decade) would let you set the cutoff at 30-40 kHz and still get 40 dB of attenuation at 100 kHz. The math matters.

Q: What safety certifications do EMI filters actually need?

It depends entirely on your target market. UL 1283 is standard in North America. VDE 0565 or EN 60939 covers most of Europe. China requires CCC, Japan requires PSE, and Korea requires KC. For medical equipment, you additionally need to verify leakage current limits under IEC 60601-1 — these are stricter than general safety standards. Always check that the safety markings are physically printed on the component, not just listed in the datasheet.

Q: Why did my EMI filter pass bench testing but fail in the field?

Thermal management. Bench tests run in open air at room temperature. Field conditions can be 30-40°C hotter inside an enclosure, especially if other components are nearby. The filter's insertion loss degrades with temperature, and if it's running near its thermal limit, it may not meet spec in real conditions. Always derate components for the actual operating environment, not the datasheet conditions.

Q: Can I cascade filters to get better attenuation?

Yes — cascading two identical stages gets you approximately the sum of their individual roll-off rates (two 1st-order stages ≈ one 2nd-order stage, roughly). But be aware of impedance interaction between stages. If the second stage loads the first too heavily, the combined response won't be what you calculated. Use op-amp buffers between stages if precise control is needed.

Q: How do I choose between a passive and an active filter?

Passive filters are simple, reliable, and don't need power. Use them for RF applications, high-current power circuits, or anywhere you can't afford the noise and distortion that active components introduce. Active filters make sense when you need gain, when the signal source has high output impedance, or when you need a precise cutoff frequency that passive components can't hit due to tolerance variation. Active filters also let you build higher-order responses without inductors, which is useful at audio frequencies.

Here's what I want you to remember from this guide:

1.Start with the noise frequency, not the filter type. Before you pick any topology, measure or estimate what frequency you're trying to block. Everything else follows from that.
2.Calculate your cutoff frequency before you build anything. It takes five minutes with a calculator and saves hours of trial and error.
3.If it's AC mains connected, you need an EMI line filter. Not optional. Not something you add later. The filter needs to be there from the beginning because the EMC requirements affect your entire board layout.
4.Size your components for real conditions. Rated current means continuous rated current in your actual thermal environment, not the ideal conditions on the datasheet.
5.When things go wrong, check the basics first: Is the capacitor aged? Is the filter saturating? Is it the right filter type for the noise mode? Most filter problems turn out to be one of these three.

For component sourcing — EMI line filter modules, X/Y capacitors, common-mode chokes, or LC filter inductors — Welllink carries verified stock with full traceability. Contact us or search our catalog for your specific requirements.