Passive low pass filters allow DC and lower-frequency signals to pass while attenuating unwanted higher-frequency components.
This article explains their operating principles, RC calculations, L-type, π and T topologies, component selection and practical EMI design for engineers working with signal lines, power supplies and cable interfaces.
Key Takeaways
A useful filter must preserve the intended waveform and reduce interference in the assembled product. Its schematic is only the starting point.
- Passive filters use resistors, capacitors, inductors and magnetic components without an external source of amplification. They provide no power gain, although resonant networks can exhibit voltage magnification.
- The familiar RC cut-off calculation assumes negligible source resistance and a high-impedance load. Real source resistance, receiver capacitance and load resistance change the response.
- L-type, π and T describe circuit arrangements, not guaranteed filter orders. Attenuation depends on the actual components, their parasitics and the source and load impedances.
- Select capacitors and magnetic components using impedance curves and operating-condition data, not nominal capacitance, inductance or a single quoted impedance alone.
- Differential-mode and common-mode noise require different current-path analysis. A common-mode choke is not an ideal component and can also affect differential signals.
- Short return paths, separation of filtered and unfiltered wiring, resonance control and measurements under realistic operating conditions determine whether a filter remains effective in practice.
Passive low pass filters: fundamentals
A filter is a frequency-dependent network. In electromagnetic compatibility (EMC) work, a passive low-pass network reduces the propagation of unwanted radio-frequency (RF) currents while preserving the required DC, analogue or digital behaviour. Capacitors provide shunt paths, series components impede propagation, and resistive losses dissipate part of the unwanted energy. [1, 2]
Noise current does not disappear into an ideal ground. It returns to its source through a complete circuit that may include a PCB reference plane, chassis, cable shield, protective earth or parasitic capacitance. Effective filtering therefore starts with identifying the noise source, its coupling path and its return path.
Filter types and ideal response curves
Filters are named according to their frequency response. The four amplitude-selective responses in Figure 1 provide the basic terminology; practical filters have a finite transition region and finite stopband attenuation rather than an abrupt boundary. [1]
- Low-pass: passes DC and lower frequencies, with increasing attenuation above the passband.
- High-pass: attenuates DC and lower frequencies while passing a higher-frequency band.
- Band-pass: passes a selected frequency band and attenuates frequencies below and above it.
- Band-stop: attenuates a selected frequency band while passing frequencies on either side. A narrow band-stop response is commonly called a notch.
- All-pass: changes phase while ideally maintaining constant amplitude. It is not shown among the four amplitude-selective responses in Figure 1.

Passive versus active filters
Passive filters contain no powered amplifying devices. Active filters use devices such as operational amplifiers or transistors and can provide buffering or gain. The absence of active amplification does not mean that every voltage in a passive network must be lower than the source voltage: a lightly damped LC or ferrite-capacitor network can produce resonant voltage peaking. It still cannot deliver net power gain. [1, 3]
A simple RC low-pass network is often suitable when signal current is modest and some series resistance is acceptable. LC and ferrite-based networks are useful when DC voltage drop must be limited or a different attenuation characteristic is required. There is no universal frequency boundary that requires changing from RC to RLC construction; current, impedance, loss, component size and parasitic behaviour determine the appropriate implementation.
RC low-pass filter operation and equations
The basic RC low-pass filter places a resistor in series with the signal and a capacitor from the output node to the signal return. The output is measured across the capacitor. Reversing the positions of the resistor and capacitor produces the familiar RC high-pass arrangement. [1]

Capacitive reactance and the cut-off frequency
For an ideal capacitor, the magnitude of capacitive reactance XC, in Ω, falls as frequency f, in Hz, increases. With capacitance C in F and angular frequency ω in rad/s, Eq. (1) gives the relationship. At low frequency the capacitor draws little current; at higher frequency its lower impedance increases the voltage drop across the series resistor. [1, 4]
For an ideal voltage source, series resistance R in Ω and a load whose impedance is much higher than the filter output impedance, the voltage transfer function H is given by Eq. (2). Here VIN and VOUT are the input and output voltage phasors, and j is the imaginary unit. The cut-off frequency fc is the frequency at which the magnitude is 1/√2 of its DC value. [1]
At fc, XC equals R and the magnitude is approximately 0.707 of the low-frequency value, corresponding to approximately 3.01 dB attenuation. This is a first-order corner, not resonance: the ideal RC circuit contains only one energy-storage element. The response changes continuously below and above the corner rather than suddenly beginning to attenuate there.
The positive voltage attenuation A, expressed in dB, follows Eq. (3). Well above the corner, the ideal first-order response approaches 20 dB per decade, or approximately 6 dB per octave. These are asymptotic slopes, not the attenuation at every frequency above cut-off. [1]
Source resistance and load impedance
The simple cut-off formula is not independent of the surrounding circuit. A driver has output resistance, the receiver may have finite resistance, and package, trace and input capacitance add to the installed shunt capacitance. These effects change both the low-frequency gain and the pole frequency.
For a resistive source RS, an added series resistor R and a resistive load RL connected in parallel with an ideal capacitor C, direct voltage-divider analysis gives Eq. (4). Here VS is the source voltage before RS, H0 is the DC voltage ratio, RT is the total series resistance, s is the complex frequency variable and ∥ denotes a parallel combination. The pole is 3.01 dB below H0, not necessarily 3.01 dB below VS. [1]
Eq. (4) applies to this lumped, resistive termination model. It is not a substitute for a transmission-line model on an electrically long cable or PCB trace, or for a frequency-dependent converter input model. For a high-impedance receiver, the source resistance adds to R, while receiver and PCB capacitance increase the effective C.
A single shunt capacitor can sometimes provide sufficient noise reduction because source and interconnection impedance already supply the other part of the voltage divider. It is not automatically effective: a low-impedance source may drive substantial capacitor current while maintaining almost the same node voltage. Check driver loading, transient current and the return-current path before removing or adding a series element.
Selecting cut-off without damaging the signal
For analogue signals, specify the highest useful frequency, acceptable amplitude error, phase error and settling time. For digital signals, also specify rise and fall times, receiver thresholds, timing margin and the relevant interface requirements. Repetition frequency alone does not describe the bandwidth needed to preserve a fast edge. [1]
Preserving several harmonics of a periodic waveform can be a useful initial check, but no universal fifth-harmonic or tenth-harmonic rule guarantees signal integrity. A filter chosen to preserve clock harmonics may provide little rejection where emissions are failing. Conversely, a lower corner may suppress emissions while making edges too slow for the receiver. Use the actual waveform and interface limits to resolve this trade-off.
For the unloaded ideal RC network driven by an instantaneous voltage step, the time constant τ is RC. Its 10% to 90% rise time tr follows Eq. (5), derived from the exponential charging response. A real driver has its own edge shape, so the final waveform must be simulated or measured rather than assuming this is the complete system rise time. [8]
Basic filter topologies
L-type, π and T describe the arrangement of series and shunt branches. They help organise a design, but the actual transfer function depends on the components and terminations. A resistor, an inductor and a ferrite bead are not interchangeable merely because each can occupy a series position. [1, 2, 3]
L-type networks
An L-type low-pass network has a series branch and a shunt branch. An RC version normally has one independent energy-storage element and is first order. An LC version can have two independent energy-storage elements and a second-order response. The orientation of the branches should be selected using the impedances seen from both ends at the noise frequencies.
π filters
A low-pass π network has a shunt capacitor, a series element and another shunt capacitor. The series element may be a resistor, inductor or ferrite bead. This arrangement can provide useful isolation between two circuit regions, but both capacitor currents and the return path must be considered. [2, 3]
A C-L-C network can realise a third-order response under suitable terminations. A C-R-C network does not become third order because it contains three components. An input capacitor connected directly across an ideal voltage source does not create an additional voltage-transfer pole. Determine the actual poles rather than assigning order from the outline of the schematic.
T filters
A low-pass T network has a series element, a shunt capacitor and another series element. An L-C-L implementation can provide a higher-order response, while an R-C-R network still has only one independent energy-storage element. T networks can be useful when series isolation is needed at both ports, but their voltage drop, losses and interaction with the load must be checked.
Filter order and attenuation slope
For an ideal all-pole low-pass transfer function, each pole contributes an eventual 20 dB per decade attenuation slope. Second-, third- and fourth-order responses therefore approach 40, 60 and 80 dB per decade, respectively, or approximately 12, 18 and 24 dB per octave. These slopes apply where the intended poles dominate and before parasitic feedthrough changes the response. [1]
Filter order is the order of the transfer-function denominator after any cancellations. Counting reactive parts is only a preliminary guide: some parts do not create independent states, some are constrained by the terminations, and real components introduce additional poles and zeros. Cascading unbuffered sections also changes their loading, so their individual cut-off frequencies cannot simply be assumed to remain unchanged.
Choosing between L-type, π and T filters
Table 1 compares the original three topology families while distinguishing circuit shape from filter order. Select the simplest arrangement that meets passband, stopband, current and layout requirements across operating conditions.
| Topology | Elements | Typical use case | Key advantages | Practical considerations |
|---|---|---|---|---|
| L-type | One series branch and one shunt branch; commonly RC or LC | Signal-line filtering or a basic power-input filter | Low component count, compact layout and straightforward analysis | RC is normally first order; LC can be second order. Check orientation, loading, damping and available attenuation. |
| π filter | Shunt capacitor, series element, shunt capacitor | Supply-rail or connector filtering where both ends need local RF return paths | Can provide stronger isolation with suitable components and terminations | Order depends on implementation. Check capacitor current, inrush, resonant peaking and return-path coupling. |
| T filter | Series element, shunt capacitor, series element | Applications requiring series isolation at both ports | Flexible distribution of series impedance and shunt filtering | Check total voltage drop, magnetic coupling, parasitic feedthrough and load interaction. R-C-R is not third order. |
Component selection for EMI low-pass filters
Real component impedance varies with frequency, temperature, voltage, current and mounting arrangement. Select parts using the impedance required in the failing emission band and verify that their behaviour remains acceptable under worst-case operating conditions. Nominal values alone are insufficient. [3, 4, 5]
Capacitor types
Shunt capacitors should present a sufficiently low installed impedance between the line and the intended RF return. Compare capacitance tolerance, effective capacitance, equivalent series resistance (ESR), equivalent series inductance (ESL), voltage rating and allowable AC current, not capacitance alone.
- Multilayer ceramic capacitors: compact surface-mount parts can provide low ESL and low impedance over a useful RF band. C0G/NP0 is suitable where stability is important; high-permittivity dielectrics such as X7R provide larger capacitance but require bias- and temperature-aware selection. Do not assume a less stable dielectric such as Y5V is an equivalent replacement in a controlled filter response.
- Film capacitors: useful in power and mains-filter applications where AC voltage, ripple current, loss and long-term behaviour matter. Lead length and construction affect high-frequency impedance. Use a specifically approved safety series where an X or Y classification is required.
- Aluminium electrolytic capacitors: useful for bulk energy storage and lower-frequency filtering. Their ESR can contribute damping, but lifetime, ripple-current capability and temperature must be considered. Do not assume that adding bulk capacitance will solve an RF problem.
High-permittivity MLCC capacitance and ESR can change with applied voltage and temperature. Use manufacturer characteristics or models at the intended operating point, and include ageing where relevant to the selected dielectric. A nominal capacitor value measured under catalogue conditions may not represent the installed filter. [5]
A capacitor reaches minimum impedance near its self-resonant frequency, where its capacitance and ESL cancel and ESR limits the minimum. Above this frequency it becomes inductive. It can still provide useful attenuation if its installed impedance remains sufficiently low relative to the surrounding circuit; self-resonance is not an absolute on/off boundary for filtering. [4]
Smaller packages, shorter connections and suitable electrode arrangements can reduce inductance. However, package size alone does not guarantee the best choice because capacitance under bias, current capability, assembly robustness and the mounting loop also matter. Safety-certified surface-mount ceramic capacitors are available, so mains suppression is not limited to leaded film constructions. [4, 7]
Inductors
A series inductor provides increasing inductive reactance only over the frequency range in which its inductive behaviour dominates. Winding capacitance, core loss and self-resonance limit the useful band. Compare the complete impedance characteristic with the required noise spectrum. [1, 2]
- Inductance and tolerance at the specified measurement frequency, operating current and temperature.
- Saturation-current definition, including the permitted inductance reduction used by the manufacturer.
- RMS or temperature-rise current rating, which addresses heating and is distinct from saturation current.
- DC resistance (DCR), winding loss, core loss and the resulting voltage drop and temperature rise.
- Self-resonant frequency, impedance-versus-frequency data, shielding and coupling to nearby components.
Magnetic shielding can reduce coupling to neighbouring circuitry, but it does not eliminate the need for suitable placement and orientation. Core material should be selected for the required impedance and loss characteristics rather than assuming that every ferrite-core inductor behaves similarly at RF.
Ferrite beads
A ferrite bead has inductive, resistive and capacitive regions. In its useful lossy region it dissipates part of the noise energy as heat. It can provide RF impedance with a lower DC voltage drop than a comparable resistor, but its DCR is not zero and its effect on the circuit is not limited to unwanted frequencies. [3]
Compare the resistive and reactive parts of impedance across the emission band. A single catalogue impedance point does not describe broadband attenuation. In a supply rail, use impedance data under DC bias: substantial impedance reduction may occur below the thermally permitted current rating. That current rating is not a guarantee of filtering performance.
Ferrite beads are not categorically unsuitable for high-current paths. Suitability depends on retained impedance at the actual current, temperature rise, DCR and the required attenuation. Similarly, a bead is not inherently a low-impedance component in every application; what matters is its impedance relative to the source and load at the frequency of interest.
Below its lossy region, a bead can behave mainly as an inductor and resonate with low-ESR ceramic capacitance. The resulting peaking can increase noise instead of reducing it. Include the bead model, source impedance and load in simulation, and provide damping where necessary. [3]
Common-mode chokes
A common-mode choke uses coupled windings arranged so that the magnetic flux from normal differential current largely cancels, while common-mode current produces reinforcing flux. It can therefore impede common-mode noise while allowing the intended differential current to pass with relatively little disturbance. [6]
The cancellation is not perfect. Leakage inductance, winding resistance and parasitic capacitance affect differential-mode transmission. In power filters, leakage inductance may contribute useful differential-mode filtering. In data interfaces, the same behaviour may degrade the wanted signal. Compare common-mode attenuation and differential-mode transmission, not common-mode inductance alone.
Balanced load current does not magnetise a current-compensated core in the same way as current through a single-winding series inductor. Nevertheless, winding heating, imbalance, common-mode current and transient conditions must remain within the component’s capabilities. Verify insulation and voltage ratings as well as current. [2, 6]
Impedance, resonance and application limits
An EMI filter works as part of a system, not as an isolated collection of nominal values. Source and load impedance determine current division, voltage division, damping and the effect of each component. These impedances differ between common-mode and differential-mode paths and vary with frequency. [2, 3, 6]
Impedance contrast rather than universal matching
For suppression, a series element should provide substantial impedance in the unwanted-current path, while a shunt element should provide a lower-impedance local return path. This deliberate impedance contrast is different from matching a transmission line for efficient signal-power transfer. Both requirements may exist in the same product, but they must not be confused.
A series impedance is often useful next to a low-impedance termination, and a shunt capacitor can be effective at a higher-impedance node. These are starting points, not universal rules. There is no general 100 Ω boundary separating low and high impedance for every EMI problem. Evaluate the actual impedances and the direction in which noise propagates. [2]
For signal interfaces, preserve required termination and differential balance while adding filtering. For power circuits, include the supply, wiring and converter input behaviour. A topology that works well in a laboratory impedance environment may provide much less attenuation in either application.
Damping LC and ferrite-capacitor resonances
Low-loss inductance and capacitance can form a high-Q resonance. Source resistance, load resistance and component ESR may supply damping, but relying on unspecified parasitic loss makes the result sensitive to component substitution and operating conditions. Check both frequency-domain peaking and time-domain ringing. [3]
Damping options include a suitable lossy component or a resistor-capacitor branch connected across the filtered supply node. The series capacitor in that branch blocks DC while allowing the resistor to dissipate oscillatory energy around resonance. Select the damping network from the actual circuit model; adding resistance can reduce high-frequency attenuation or increase loss if placed in the wrong branch. [3, 9]
Power and SMPS input-filter stability
A regulated DC/DC converter can exhibit negative incremental input impedance within part of its control bandwidth: if input voltage falls while output power is maintained, input current rises. An added LC filter can interact with this behaviour, reduce damping and cause oscillation or a poor transient response. This concern is specific to the powered converter system, not merely the passive filter on its own. [9]
Check the filter output impedance against the converter’s frequency-dependent input impedance and verify the combined system’s stability. Keeping filter output impedance well below converter input-impedance magnitude over the interaction band is a useful design objective, but a full assessment must consider frequency, phase and operating conditions. Validate startup, minimum input voltage, maximum load and load steps, as well as emissions. [9]
Differential-mode and common-mode noise
Differential mode (DM) and common mode (CM) describe current relationships and return paths, not whether a signal is useful. A circuit may carry useful differential data and unwanted differential ripple at the same time. Both DM and CM mechanisms can contribute to emissions. [2, 6]
Differential-mode filtering
Differential-mode current flows out through one conductor and returns through the other. Unwanted DM components can be reduced with capacitance across the pair and suitable series impedance. For a power circuit this may be an LC network; for a data interface, the permissible capacitance and series impedance depend on signal-integrity requirements.
Keeping outgoing and return paths close reduces loop area and external field coupling, but equal and opposite currents do not imply perfect field cancellation everywhere. Layout discontinuities, imbalance and unequal parasitic coupling can convert differential energy into common-mode energy. A balanced filter must preserve the symmetry needed by the interface. [2, 6]
Common-mode filtering
Common-mode currents flow in the same direction on the conductors of a pair, with their return path outside that pair. The return may involve chassis, earth, shields or parasitic capacitance. Even a relatively small CM current can create a significant radiated-emissions problem when it flows on an external cable. [2]
A common-mode choke adds impedance to this path. Capacitors from the conductors to an appropriate RF reference can provide a controlled return, provided the reference connection is effective and electrically safe. Circuit return, chassis and protective earth are not interchangeable labels; the connection must be selected for the product architecture. [2, 6]
A real CM filter can also load the differential signal through choke leakage inductance and line-to-reference capacitance. Do not assume that a failed DM filter proves the problem must be CM, or vice versa. Insufficient impedance, parasitic bypassing, poor placement and mixed-mode noise can produce the same result.
Parasitics, layout and placement
At RF, the installed interconnect is part of the filter. Trace inductance, capacitor mounting loops, winding capacitance and coupling between the two sides can dominate performance. A technically correct schematic can therefore fail when the physical current paths differ from the assumed model. [2, 4, 6]
Parasitic effects and useful bandwidth
A real capacitor includes ESR and ESL; a real inductor includes winding resistance, core losses and parasitic capacitance. Capacitor impedance rises above self-resonance, while an inductor or choke can lose its intended high-frequency impedance as capacitive behaviour becomes significant. Use measured impedance curves, equivalent circuits or suitable S-parameter models across the complete band of interest. [4, 6]
Surface-mount construction often reduces lead inductance, but pads, vias and the return plane remain important. Combining different capacitance values can introduce additional resonances rather than simply extending bandwidth. Verify the assembled network instead of assuming that more capacitance or more components always improves attenuation.
Placement and return paths
For cable emissions and immunity, place the interface filter at the connector or enclosure boundary so that an unfiltered internal trace does not continue to drive the cable. For local supply isolation, place the network at the boundary between the noisy and sensitive circuit regions. These are different placement objectives; a filter need not always be at the noise source or always at the connector. [2]
- Keep shunt-capacitor connections and their RF return paths short and low inductance.
- Separate unfiltered and filtered traces, wires and component terminals so that capacitive or magnetic coupling does not bypass the network.
- Minimise loop area and avoid forcing high-frequency return current through a long shared path.
- Route both conductors of a differential pair consistently through the filter and avoid unnecessary asymmetry.
- Follow component-specific guidance for copper beneath chokes and other magnetic parts. Do not create a capacitive bypass across the intended isolation boundary.
- Keep safety clearances and isolation requirements intact when shortening RF paths.
A ground symbol is not a zero-impedance sink. Identify where each RF current closes its loop and whether that path contaminates the supposedly clean side of the filter. This analysis is more useful than adding capacitors to an undefined ground reference.
Worked example: 33 Ω and 100 pF on a digital line
Consider the original example: a 5 V digital line has useful spectral content up to 10 MHz and unwanted cable-related energy around 100 MHz. A series resistor of 33 Ω and a shunt capacitor of 100 pF are proposed at the connector. The following calculations assume an ideal voltage source, a high-impedance receiver, ideal components and no cable transmission-line effects.
Using Eq. (2), the cut-off frequency is 48.23 MHz. The original rounded value of about 48 MHz is correct. The time constant is 3.30 ns, and Eq. (5) gives an ideal step rise time of 7.25 ns. These are calculated results for the stated network, not guaranteed measurements of the complete interface.
Table 2 applies Eq. (3) to selected frequencies. It shows why the asymptotic 20 dB per decade slope should not be confused with 20 dB attenuation at 100 MHz.
| Frequency | Output/input magnitude | Calculated attenuation |
|---|---|---|
| 10 MHz | 0.979 | 0.18 dB |
| 30 MHz | 0.849 | 1.42 dB |
| 50 MHz | 0.694 | 3.17 dB |
| 100 MHz | 0.434 | 7.24 dB |
The network introduces little amplitude loss at 10 MHz but appreciably attenuates higher-frequency components that form sharp digital edges. Whether a 7.25 ns filter step rise time is acceptable depends on the driver, receiver and timing budget. A 5 V logic swing does not establish that requirement, and the calculated cut-off is slightly below 50 MHz rather than above it.
The calculated 7.24 dB attenuation at 100 MHz may be useful, but it does not guarantee the same reduction in radiated emissions. The actual cable current can include common-mode components that this single-ended RC network does not control effectively. Driver resistance, receiver capacitance, the capacitor return path and the installed cable must be included in the final evaluation.
If greater rejection is required, first confirm the emission mode and available signal-integrity margin. Then consider revised RC values, an appropriately damped higher-order network or a common-mode solution. Retest the waveform and emissions after each change rather than assuming that a steeper nominal slope resolves the system problem.
Reliability, safety and selection checklist
A filter must meet its electrical targets throughout production variation and service conditions, not only on a room-temperature prototype. Include effective component values, heating, mechanical robustness and the product’s insulation architecture in the selection process. Safety approval of an individual component does not establish compliance of the complete product. [5, 7]
Mains-connected filters and safety capacitors
Use appropriately approved X capacitors for across-the-line suppression positions and Y capacitors where the application requires their line-to-earth or insulation-bridging safety function. IEC 60384-14 is the relevant component standard identified in safety-capacitor documentation. The required subclass and voltage rating depend on the circuit position and product requirements; a generic high-voltage capacitor is not an equivalent substitute. [7]
Check the selected part’s actual certificates, permitted working voltage, impulse capability and environmental ratings. Capacitor approval does not remove PCB creepage and clearance requirements. Line-to-earth capacitance also contributes to leakage or touch current, so capacitance cannot be increased solely to improve EMI attenuation. Apply the limits and test methods required by the product’s safety standard. [2, 7]
Where required by the product safety assessment, provide suitable discharge arrangements for stored charge. Also assess inrush, abnormal operating conditions, protective-earth connections and insulation in the magnetic components. Treat mains-filter changes as safety-relevant design changes, not merely component substitutions.
Design and purchasing checklist
- Requirements: record the wanted signal bandwidth or power waveform, permissible voltage drop, edge-rate and timing limits, failing emission frequencies and required attenuation margin.
- Noise paths: identify DM and CM contributions, their source and return paths, and the reference used by each shunt component.
- Capacitors: compare effective capacitance under operating bias, tolerance, temperature, ageing, ESR, ESL, impedance curves, AC current and safety classification where applicable.
- Inductors and beads: compare operating-current impedance, DCR, saturation characteristics, thermal rating, self-resonance, core losses and mounting constraints.
- Common-mode chokes: compare CM attenuation, DM transmission or leakage inductance, winding balance, current, voltage, insulation and parasitic capacitance.
- Power-converter interaction: verify resonance damping, filter output impedance, converter stability, startup and load-transient behaviour.
- Catalogue versus custom: use catalogue parts when their documented electrical, mechanical and approval limits meet the requirement. Define custom magnetics or assemblies by measurable impedance, current, temperature, insulation and mechanical requirements rather than nominal inductance alone.
- Qualification: match environmental and application qualifications to the product. Do not treat a reliability qualification or a supplier’s general compliance statement as a substitute for the required safety approval.
- Substitutions: qualify alternatives using impedance and bias data, thermal behaviour, certificates and installed-system tests. Equal nominal values do not establish interchangeability.
- Production control: retain approved part numbers, model revisions, test conditions and assembly constraints so that later purchasing changes do not silently alter attenuation or damping.
Measurement and troubleshooting
Component models and bench measurements reduce design uncertainty, but compliance depends on the complete product in its specified test configuration. Measure both the intended signal and unwanted emissions, using repeatable operating conditions and an identified noise mode. [2, 6, 9]
Insertion loss compares the received level without and with the filter under otherwise defined conditions. It is not automatically identical to the voltage transfer ratio of an unloaded circuit. Catalogue curves measured with specified source and load impedances do not directly predict attenuation in an arbitrary supply rail or cable interface.
For conducted-emissions work, use the line impedance stabilisation network (LISN), receiver settings and arrangement required by the applicable test method. A vector network analyser is useful for component and network characterisation, but its port impedance and fixtures form part of the measurement. Account for fixture parasitics and protect test equipment from operating voltage and current. [2, 6]
- Establish a baseline with the original cable, load, supply and operating mode recorded.
- Identify whether the dominant path is differential mode, common mode or a combination before changing topology.
- Check the signal waveform at the receiver, including edge rate, ringing, settling and threshold margin.
- Measure or model the filter across load current, supply voltage and temperature, including low-load conditions that can expose resonant peaking.
- If attenuation is weak, check the installed return path, coupling around the network, bias-reduced component impedance and actual source and load impedances.
- If noise increases after adding the filter, investigate resonance or converter interaction before adding further stages.
- Repeat conducted and radiated tests after the final layout and component selection are fixed. Confirm that an improvement in one band has not created a new problem elsewhere.
Conclusion
Passive low-pass filtering is a practical way to reduce unwanted high-frequency currents, but successful design depends on more than selecting a nominal cut-off frequency. Define the wanted waveform, identify the DM and CM paths, choose an appropriate topology and evaluate the real component impedances and circuit terminations.
The RC example demonstrates the central trade-off: 33 Ω and 100 pF give a calculated 48.23 MHz corner, but only 7.24 dB ideal attenuation at 100 MHz and a 7.25 ns ideal step rise time. For power filters, add damping and converter-stability checks; for mains filters, preserve the required safety architecture. Final validation must include the installed layout, cables, operating conditions and intended signal performance.
FAQ
These answers summarise the main design and component-selection questions addressed in the article.
A passive low pass filter uses resistors, capacitors, inductors or magnetic components to pass DC and lower-frequency signals while attenuating unwanted higher-frequency components. It contains no powered amplification. Its installed performance depends on source and load impedance, component parasitics and layout, so a nominal cut-off frequency alone does not establish its EMI effectiveness.
For an ideal source, a series resistor and a shunt capacitor feeding a high-impedance load, the cut-off frequency is the reciprocal of two times pi times resistance times capacitance. The output magnitude is then approximately 3.01 dB below its DC value. Source resistance, load resistance and additional input capacitance change the result. This RC corner is not resonance.
No. The names describe the arrangement of series and shunt branches, not the number of independent poles. A resistor-capacitor-resistor T network has only one independent energy-storage element. An inductive and capacitive implementation can provide a higher-order response, but the source, load and parasitic elements determine the actual transfer function and attenuation.
A ferrite bead can be useful when RF impedance is needed with less DC voltage drop than a resistor would produce. Select it from impedance curves at the operating current and temperature, not a single catalogue impedance value. Check heating, DC resistance and resonance with the shunt capacitance. The rated current does not guarantee retained filtering impedance.
Choose the response from the receiver timing requirements, permissible edge rate, waveform distortion and noise frequencies that require attenuation. Clock frequency alone is insufficient because fast edges contain higher-frequency energy. Preserving a fixed number of harmonics is only an initial check. Verify the complete driver, filter, interconnect and receiver with simulation and measurements.
Differential-mode filtering targets unwanted current flowing out through one conductor and back through its paired return, commonly using capacitance across the pair and series impedance. Common-mode filtering targets same-direction currents whose return is outside the pair, commonly using a common-mode choke and an appropriate RF return. Real components can affect both modes and must preserve the intended signal.
An LC or ferrite-capacitor network can increase noise through resonant peaking, and a power-input filter can interact with a regulated converter. Weak attenuation can also result from unsuitable source and load impedances, reduced component impedance under bias, parasitic bypassing or an incorrect noise-mode diagnosis. Check current paths, layout, damping and operating conditions before adding more components.
Compare operating-condition impedance, effective capacitance or inductance, DC resistance, thermal capability, parasitics, mechanical requirements and relevant qualifications. For mains positions, verify the required safety class and certificates. Equal nominal values do not establish equivalence. Qualify substitutions in the installed circuit because changes in losses or bias behaviour can alter attenuation, damping and converter stability.
How to design and verify a passive low-pass EMI filter
Use the following procedure to connect the initial requirements, component selection and final system measurements. Apply the signal-interface, power-converter and mains-safety checks that correspond to the actual application.
Design and verification procedure for a passive low-pass EMI filter.
- Define the wanted waveform and EMI target
Record the useful bandwidth, edge-rate or ripple requirements, permissible voltage drop and failing emission frequencies. Identify the applicable product test method and the attenuation margin required.
- Identify the noise mode and current paths
Determine whether the dominant interference is differential mode, common mode or mixed mode. Trace the source, outgoing path and return path, including chassis, cable and parasitic connections.
- Select topology using the actual impedances
Choose an RC, LC, ferrite-based or common-mode network using frequency-dependent source and load impedances. Treat L-type, π and T as arrangements rather than guaranteed filter orders.
- Calculate the response and preserve the signal
Calculate cut-off, attenuation and transient behaviour with source resistance and receiver loading included. For digital lines, verify edge rate and receiver timing rather than relying on clock frequency or a fixed harmonic rule.
- Select components at operating conditions
Compare effective capacitance, magnetic impedance under current, DC resistance, thermal ratings and parasitic behaviour. Check common-mode and differential-mode characteristics separately where a coupled choke is used.
- Control resonance and converter interaction
Simulate peaking and ringing with realistic losses and add damping where needed. For a regulated converter input, verify the interaction between filter output impedance and converter input impedance across operating conditions.
- Implement the physical filtering boundary
Place cable-interface filtering at the connector or enclosure boundary and local supply filtering at the relevant circuit boundary. Keep return paths short and separate filtered wiring from unfiltered wiring to prevent bypass coupling.
- Complete safety and qualification checks
For mains or insulation-bridging positions, verify capacitor safety class, certificates, working voltage, insulation, creepage, clearance and leakage-current requirements. Record approved parts and the tests needed to qualify substitutions.
- Measure, iterate and freeze the design
Compare emissions and receiver waveforms before and after filtering under repeatable conditions. Check startup, load changes and operating extremes where relevant. Resolve remaining mode or layout problems, then retain the final part numbers, layout constraints and verification results.
Further Reading
These related articles develop the underlying circuit concepts and filter specifications in more detail.
- Basic Filter Circuits Explained: introduces the frequency-dependent voltage-divider view of RC, RL and LC networks.
- Five Key Filter Specifications: explains how passband, rejection and selectivity requirements guide filter selection.
- Filter Q Factor Explained: provides further background on selectivity and resonant behaviour.
- Filter Bandwidth Explained: develops the bandwidth terminology used when specifying the wanted signal range.
- Filter Poles and Zeros Explained: extends the transfer-function interpretation of filter order and response.
- RF Inductors and Filters: broadens the component context for frequency-selective passive networks.
References
The numbered citations identify the technical sources used for the circuit principles, component behaviour and implementation guidance. Numerical results in the worked example are calculated from the stated component values and assumptions.
- Analog Devices, Basic Linear Design, Chapter 8: Analog Filters, Analog Devices, undated chapter PDF.
- Raf Vleugels, EMI Debugging: Designing Your Ideal Filter, Würth Elektronik, 2024.
- Jefferson Eco and Aldrick Limjoco, AN-1368: Ferrite Bead Demystified, Analog Devices, undated online application note.
- Murata Manufacturing, What are impedance/ESR frequency characteristics in capacitors?, Murata Capacitor Guide, 2013.
- Seiji Hidaka, MLCC Dynamic Model Supports Circuit Simulations, Murata Manufacturing, 2016.
- Heinz Zenkner, ANP146: Theoretical Insights and Practical Applications of the WE-CMDC Series Common Mode Chokes, Würth Elektronik, 2025.
- KEMET, Safety Certified CAS Surface Mount X1/Y2 and X2 Capacitors, YAGEO Group, datasheet revision 25 June 2026.
- Analog Devices, Activity: Transient Response of an RC Circuit, Analog Devices Wiki, 2022 revision.
- Charles Zhang, Analysis and Design of Input Filter for DC-DC Circuit, Texas Instruments, application report SNVA801, 2017.
Latest technical review and revision: 5 November 2026. RC calculations, topology descriptions, component-selection guidance, EMI implementation, safety considerations and supporting references reviewed and updated.





















