Insertion Loss and Filter Capacitor Performance

Capacitors are used in both analog and digital circuits to remove unwanted high-frequency signals from supply and signal lines. The filtering performance of a capacitor or a filter circuit is commonly described by its insertion loss, the attenuation it provides between a source and a load over frequency. Insertion loss is not a fixed property of the component: it depends on the filter configuration, the parasitic inductance of the capacitor and its connections, the source and load impedances, the load current and the stability of the dielectric. This article explains how insertion loss is defined and measured, how to calculate it for a shunt capacitor, and how to read datasheet curves for real capacitors and filters.

How to Measure Filter Performance: Insertion Loss, VNA Fixtures and CM/DM Separation

Passive Filter Technologies Compared: LC, Ferrite, Feedthrough, Ceramic, Cavity, Microstrip, SAW and BAW

Designing 800 V DC EMC Filters: Calculation, Simulation and Measurement

Key Takeaways

Insertion loss compares the voltage or power at the load without and with the filter; datasheets specify it in a 50 Ω source and 50 Ω load system.
A shunt capacitor is effective only when the impedance it works against is high. The same 1 nF capacitor gives about 24 dB at 100 MHz in a 50 Ω system and less than 0.5 dB between 1 Ω terminations.
Above its self-resonant frequency (SRF), a capacitor behaves as an inductance, and its insertion loss is set by its equivalent series inductance (ESL), not by its capacitance.
Three-terminal and feedthrough capacitors reduce the inductance in the ground path and extend the useful frequency range; a feedthrough capacitor mounted in a shield wall comes closest to ideal behaviour.
Configuration should follow the impedances: capacitors face high impedances, inductors face low impedances. C and Pi filters suit high source and load impedances, T filters suit low ones.
Datasheet curves are measured at room temperature with no load current. DC bias of class 2 dielectrics and ferrite saturation can reduce the insertion loss substantially in the application.

Filtering EMI in Circuits

Electrical disturbances, both natural and man-made, can affect the operation of an electronic circuit. These unwanted signals are collectively called electromagnetic interference (EMI). Typical sources include lightning, power lines, motors, ignition systems, radar transmitters, power amplifiers, switching converters and digital clocks. Filter circuits built from capacitors and inductors are used in most analog and digital designs to keep these signals out of sensitive circuits or to prevent a circuit from emitting them. The simplest filter, the C filter, is a single capacitor connected between the line and ground (a shunt capacitor). In EMI filter products, this capacitor is usually a feedthrough capacitor because of its low self-inductance, but the C configuration itself can be built from any capacitor type [1], [2]. Combining capacitive and inductive elements gives L-C, Pi and T filters with steeper attenuation, as described in the EMI filtering basics.

Insertion Loss Definition and Measurement

Insertion loss (IL) describes how much a filter attenuates a signal between source and load. In voltage form, it compares the voltage at the load without and with the filter:

IL = 20\log_{10}\left(\frac{V_\mathrm{without}}{V_\mathrm{with}}\right)\ \mathrm{dB} \tag{1}

In power form, the same quantity is:

IL = 10\log_{10}\left(\frac{P_\mathrm{without}}{P_\mathrm{with}}\right)\ \mathrm{dB} \tag{2}

Both forms give the same result because the load resistance is unchanged. The Syfer filter catalogue defines insertion loss as the ratio of the generator voltages needed to keep a constant output voltage with and without the component, which is equivalent to Eq. (1) [3]. With a network analyser, the instrument keeps the source level constant and records the ratio directly in decibels.

Datasheet curves of feedthrough capacitors and EMI filters are measured in a 50 Ω source and 50 Ω load system, following the method of MIL-STD-220 [3], [4]. The measurement is made at room temperature and with no load current [2]. When comparing components, check that the test impedance, fixture and configuration are the same; otherwise the differences may come from the measurement set-up rather than from the components.

Shunt Capacitor Insertion Loss

A real capacitor is a series combination of capacitance C, equivalent series resistance (ESR) and equivalent series inductance (ESL). Its impedance is:

Z_C = \mathrm{ESR} + j\left(\omega\,\mathrm{ESL} – \frac{1}{\omega C}\right),\quad \omega = 2\pi f \tag{3}

The reactive parts cancel at the self-resonant frequency, where the impedance falls to the ESR:

f_\mathrm{SRF} = \frac{1}{2\pi\sqrt{\mathrm{ESL}\cdot C}} \tag{4}

When the capacitor is connected as a shunt element between a source with resistance R_S and a load R_L, the insertion loss follows from the voltage divider:

IL_C = 20\log_{10}\left|1 + \frac{R_P}{Z_C}\right|,\quad R_P = \frac{R_S R_L}{R_S + R_L} \tag{5}

For an ideal capacitor, the corner frequency at which the insertion loss reaches 3 dB is:

f_c = \frac{1}{2\pi R_P C} \tag{6}

In a 50 Ω / 50 Ω system, R_P is 25 Ω, so an ideal 1 nF capacitor has a corner frequency of 6.4 MHz. Above it, the insertion loss rises by 20 dB per decade. A series inductor L behaves in the opposite way: it is effective against low impedances, as the series-element expression shows:

IL_L = 20\log_{10}\left|1 + \frac{j\omega L}{R_S + R_L}\right| \tag{7}

These two expressions explain the impedance rules for filter configuration. A shunt capacitor needs a high impedance to work against, and a series inductor needs a low one. For example, a 1 µH inductor gives only 1.4 dB at 10 MHz in a 50 Ω system but 29.9 dB between 1 Ω terminations, while an ideal 1 nF capacitor gives 23.9 dB at 100 MHz in a 50 Ω system and only 0.4 dB between 1 Ω terminations.

Insertion Loss Variation with Frequency

The insertion loss of an ideal capacitor rises without limit as frequency increases. A real capacitor follows the ideal curve only up to its SRF. At resonance, the insertion loss reaches a peak limited by the ESR, and above the SRF the capacitor behaves as an inductance, so the insertion loss falls as frequency rises. The difference from the ideal curve is therefore not small: at high frequencies, it can amount to tens of decibels.

Table 1 shows values calculated with Eq. (5) for a 50 Ω / 50 Ω system. An ESR of 20 mΩ is assumed for the real capacitors.

Frequency	Ideal 1 nF	1 nF, ESL 1 nH (SRF 159 MHz)	1 nF, ESL 0.1 nH (SRF 503 MHz)	10 nF, ESL 1 nH (SRF 50 MHz)
1 MHz	0.1 dB	0.1 dB	0.1 dB	5.4 dB
10 MHz	5.4 dB	5.4 dB	5.4 dB	24.3 dB
100 MHz	23.9 dB	28.3 dB	24.3 dB	34.5 dB
At own SRF	–	61.9 dB	61.9 dB	61.9 dB
1 GHz	43.9 dB	12.5 dB	34.5 dB	12.3 dB

Table 1. Calculated insertion loss of a shunt capacitor in a 50 Ω / 50 Ω system, Eq. (5), ESR 20 mΩ for real capacitors.

Three effects are visible. Below resonance, the real capacitors follow the ideal curve, with a small gain just below the SRF where the inductance partly cancels the capacitive reactance. At resonance, the peak of about 60 dB is set by the ESR, which agrees with the observation from Murata that a single capacitor reaches only about 60 dB at some frequencies regardless of its capacitance [5]. At 1 GHz, the 1 nF and 10 nF capacitors with the same 1 nH ESL give practically the same insertion loss of about 12 dB.

Above resonance, therefore, the insertion loss does not change with capacitance if the ESL stays the same. A larger capacitance only moves the resonance to a lower frequency. For noise suppression at high frequencies, the important parameter is the ESL of the capacitor together with the inductance of its connections, not the nominal capacitance. Reducing the ESL from 1 nH to 0.1 nH raises the 1 GHz insertion loss from 12.5 dB to 34.5 dB in the example.

Three-Terminal and Feedthrough Capacitors

In a two-terminal capacitor, the residual inductance is high because the leads, terminations and electrodes carry the current to ground and act as inductors. The inductance grows with the length of the current path and is higher for narrow electrodes. Changing the capacitor architecture reduces this inductance.

In a three-terminal capacitor, the signal current passes through the capacitor from the input to the output terminal, and separate ground terminals connect the electrodes to ground. The inductance that remains in series with the capacitance is mainly in the ground path, which is short. Murata documentation shows that this structure gives a higher self-resonant frequency and a better insertion loss than two-terminal capacitors of the same 1,000 pF value, both for leaded and chip parts [6]. Murata, an EPCI member, also notes that the leads of two-terminal capacitors act as inductors and limit the high-frequency performance.

Feedthrough capacitors take the principle further. The ground electrode surrounds the dielectric, and the signal conductor passes through its centre. When the ground electrode is soldered or bolted directly to a shielding plate, there is practically no residual inductance on either side, and the insertion loss approaches that of an ideal capacitor [6]. Discoidal ceramic capacitors have the lowest inductance and are used in panel-mount filters and filtered connectors. Tubular capacitors are cheaper but are typically limited to about 20 nF [2]. Plastic film feedthrough capacitors are used where higher voltages or self-healing behaviour are required.

A three-terminal feedthrough chip carries the signal current through its electrodes, so its current rating and performance are limited by the electrode conductivity and architecture. An X2Y capacitor uses opposing currents to cancel internal inductance and is connected as a bypass, so the load current does not flow through it [2].

Surface-mount filters lose insertion loss in the range of hundreds of megahertz to 1 GHz, above the component resonance, because the PCB ground path adds inductance and the input and output tracks couple to each other. A grounded shield over the clean side of the circuit, or a buried-layer layout with the output track on the opposite side of the board and an internal ground plane between the tracks, brings the performance closer to that of panel-mount filters [2].

Filter Configuration and Impedances

Although single elements can remove unwanted signals, most filters use a combination of capacitive and inductive elements. The most common configurations are C, L-C (or C-L), Pi and T. Figure 1 shows these basic sections together with the single series inductor (L) and the resistive RC and CR variants.

Figure 1. Basic low-pass filter sections: C, L, LC, CL, RC, CR, Pi and T.

Theoretically, each reactive element adds 20 dB per decade to the slope of the insertion loss curve: about 20 dB per decade for a C filter, 40 dB per decade for an L-C filter and 60 dB per decade for Pi and T filters [4]. Multi-element filters, such as L-C-L-C-L, give still steeper slopes. Combining a capacitor with an inductor also overcomes the roughly 60 dB limit of a single capacitor; an LC filter can provide 80 dB, or more than 100 dB depending on the conditions [5].

The choice of configuration should follow the source and load impedances. Capacitors should face high impedances and inductors should face low impedances [1], [2].

Configuration	Elements	Ideal slope	Source and load impedance	Typical use
C	1 C (shunt)	20 dB/decade	High source and high load	Simple filtering, high-current lines where inductors would saturate
L-C or C-L	1 L, 1 C	40 dB/decade	One low, one high; inductor faces the low impedance	Supply inputs fed from a low-impedance source
Pi	2 C, 1 L	60 dB/decade	High source and high load	Signal and control lines, panel-mount filters
T	1 C, 2 L	60 dB/decade	Low source and low load	Low-impedance power lines

Table 2. Filter configurations, ideal insertion loss slopes and the matching source and load impedances.

A filter specified in a 50 Ω system can behave very differently between a low-impedance power stage and a high-impedance load. The datasheet curve should therefore be treated as a comparison tool, not as a prediction of the attenuation in the application [1]. Matching the topology to the real impedances is often more effective than increasing the nominal capacitance or inductance.

A rough estimate is useful before simulation. A single-pole filter with corner frequency f_c provides about 20 dB at 10 × f_c and 40 dB at 100 × f_c under the test impedances; each additional reactive element adds about 20 dB per decade. The estimate is valid only below the self-resonant frequencies of the components and for the impedances used in the calculation.

Factors That Determine Insertion Loss

The main factors are the electrical configuration, the source and load impedances, the load current, the dielectric material, the earthing impedance and the shielding integrity, as discussed in Insertion Loss and Performance in EMI Filtering and in the Knowles filter guidance [1]. Knowles is an EPCI member and manufactures panel-mount and surface-mount EMI filters.

Load current

Surface-mount and small panel-mount filters form their inductive elements with ferrite beads of about 50 nH to 500 nH. Ferrite saturates as the current increases, and typically about 6 A through the pin leads to total saturation; the rated current of many filters is set by pin heating rather than by the ferrite [2]. When the ferrite saturates, the insertion loss approaches that of a C filter. T filters suffer more than Pi filters because they use a smaller capacitance. For high-current lines, a high-capacitance C filter is preferable [2]. See also the impact of DC bias on SMT EMI suppression ferrites.

Dielectric material

EMI filters use C0G/NP0 (ultra-stable, dielectric constant about 10 to 100) and X7R (stable, about 2,000 to 4,000) dielectrics. X7R capacitance can change by ±15 % over the temperature range with no voltage applied and ages by about 1 % per time decade [3]. Applied voltage reduces the capacitance further, as explained in DC bias characteristics of ceramic capacitors. General-purpose Z5U, Y5V and X7W dielectrics can lose most of their capacitance under voltage [1].

The Syfer filter guidance compares two 10 nF filters at 50 V and 125 °C: a 500 V-rated X7R part keeps about 8.6 nF, a 50 V-rated Z5U part only about 1.5 nF [2]. With Eq. (5) in a 50 Ω system at 1 MHz, this corresponds to about 4.5 dB versus 0.2 dB of insertion loss. The capacitance loss also moves the corner frequency up and shifts the SRF.

Earthing impedance and shielding

Any impedance between the capacitor ground and the reference adds in series with the capacitor and acts like additional ESL. In a MIL-STD-220 test fixture, slackening the mounting nut of a 680 nF panel-mount filter by a quarter turn visibly reduced the insertion loss, and an oversized mounting hole let high-frequency noise radiate through the bulkhead to the output side [2]. A filter can only be as good as the shield wall or ground plane it is mounted in.

Factor	Effect on insertion loss	Datasheet condition	Design action
Source and load impedance	Can increase or decrease the attenuation	50 Ω / 50 Ω	Choose the configuration by the real impedances
ESL and connections	Insertion loss falls above the SRF	Test fixture with short connections	Use three-terminal, feedthrough or X2Y parts; short ground paths
Load current	Ferrite saturation, IL approaches C filter	No load current	Check DC bias curves; use C filter for high currents
Dielectric	Capacitance falls with voltage, temperature and age	Room temperature, no bias	Prefer C0G or X7R with voltage margin
Earthing and shielding	Leakage around the filter, lower high-frequency IL	Ideal fixture	Bond the filter to the shield or ground plane with low impedance

Table 3. Factors that reduce the insertion loss in the application compared with the datasheet curve.

Layout Guidelines

Connect shunt capacitors to ground with the shortest and widest path; use several vias directly at the ground pads.
Place the filter at the point where the line enters the shielded area or the board, and keep the unfiltered and filtered tracks apart.
Use three-terminal capacitors with both ground terminals connected, or feedthrough capacitors bolted or soldered into the shield wall.
Route the output track on another layer than the input track, with a continuous ground plane in between, when high-frequency attenuation is critical.
Avoid long leads or narrow traces in series with the capacitor; a few millimetres of trace add inductance comparable to the ESL of the part.

Selection Checklist

Frequency range and required attenuation defined from the measured noise and the applicable limit.
Source and load impedances estimated; configuration chosen so that capacitors face high and inductors face low impedances.
Capacitor SRF placed above or within the critical band; ESL of the part and connections included in the calculation.
Capacitance checked at the operating voltage, temperature and end of life.
Ferrite elements checked at the maximum DC load current.
Grounding and shielding of the filter defined and verified.
Datasheet curve compared only at identical test conditions; final performance confirmed by measurement in the application.

Conclusion

Capacitors are key elements of EMI filters, and their effectiveness is best described by insertion loss over frequency. The insertion loss of a shunt capacitor depends on the impedance it works against, and above the self-resonant frequency it is set by the ESL rather than by the capacitance. Three-terminal and feedthrough capacitors reduce the inductance in the ground path and extend the useful frequency range, provided they are well connected to ground or a shield.

Datasheet curves measured in a 50 Ω system with no load current are a comparison tool. In the application, the real impedances, DC bias of the dielectric, ferrite saturation and the quality of the ground connection determine the attenuation. Choosing the configuration by the source and load impedances and checking these factors gives a filter that reaches its calculated performance.

FAQ about Insertion Loss and Filter Capacitors

What is insertion loss in EMI filtering?

Insertion loss is the ratio of the voltage or power at the load without and with the filter, expressed in decibels over frequency. Datasheets specify it in a 50 ohm source and 50 ohm load system, usually measured with a network analyser according to MIL-STD-220. The attenuation in an application with different impedances can be higher or lower than the published curve.

Is a C filter always a feedthrough capacitor?

No. A C filter is any single capacitor connected between the line and ground. In EMI filter products the capacitor is usually a feedthrough or discoidal type because of its very low self-inductance, but an MLCC, a three-terminal chip or a film capacitor can also form a C filter. The configuration describes the circuit, not the capacitor construction.

Why does the insertion loss of a real capacitor fall at high frequencies?

A real capacitor has an equivalent series inductance. At the self-resonant frequency the capacitive and inductive reactances cancel and the impedance falls to the ESR, which gives the highest insertion loss. Above this frequency the capacitor behaves as an inductance, its impedance rises with frequency and the insertion loss decreases.

Does a larger capacitance improve high-frequency filtering?

Not above the self-resonant frequency. There the insertion loss is set by the ESL, so a 1 nF and a 10 nF capacitor with the same 1 nH inductance give practically the same attenuation of about 12 dB at 1 GHz in a 50 ohm system. A larger capacitance only moves the resonance to a lower frequency.

Why do three-terminal and feedthrough capacitors give better insertion loss?

They pass the signal current through the capacitor and connect the electrodes to ground by a short, wide path, so only a small inductance remains in series with the capacitance. This raises the self-resonant frequency. A feedthrough capacitor mounted directly in a shield wall has practically no residual inductance and comes closest to an ideal capacitor.

How do source and load impedances affect filter selection?

A shunt capacitor works against high impedances and a series inductor against low impedances. C and Pi filters therefore suit high source and load impedances, T filters suit low ones, and in an L-C filter the inductor should face the lower impedance. The same 1 nF capacitor gives about 24 dB at 100 MHz with 50 ohm terminations but below 0.5 dB with 1 ohm terminations.

How does load current change the insertion loss?

Filters with inductive elements usually use ferrite beads, which saturate as the DC current increases. The inductance falls and the insertion loss approaches that of a C filter at full load. T filters are affected more than Pi filters because they contain a smaller capacitance. High-current lines are therefore often filtered with a high-capacitance C filter.

Why can a filter give less attenuation in the application than in the datasheet?

Datasheet curves are measured at room temperature, with no load current and in a 50 ohm system. In the application, class 2 dielectrics lose capacitance under voltage and temperature, ferrites saturate, the impedances differ from 50 ohms and the ground connection adds inductance. Each of these effects can reduce the insertion loss.

How to Select Filter Capacitors for Optimal Insertion Loss

Step-by-step procedure for selecting filter capacitors and a filter configuration that reach the required insertion loss in the real application.

Define the noise problem and frequency range
Identify the main noise sources, such as switching converters, motors or clocks, and determine the frequency range and the attenuation needed to meet the limit with margin.
Determine source and load impedances
Estimate or measure the real source and load impedances seen by the filter instead of assuming 50 ohm terminations.
Choose the filter configuration
Select C or Pi for high source and load impedances, T for low impedances, and L-C with the inductor facing the lower impedance; add elements where a steeper slope is needed.
Select capacitor technology and value
Calculate the insertion loss with the real impedances, place the self-resonant frequency in or above the critical band, and use three-terminal, feedthrough or X2Y capacitors when low ESL is required.
Check dielectric and load current
Verify the capacitance at the operating voltage, temperature and end of life, and check that ferrite elements do not saturate at the maximum DC load current.
Optimize layout and grounding
Connect the capacitors to ground or the shield wall with the shortest path, separate input and output tracks, and use a ground plane or shield between them.
Verify insertion loss in the application
Compare datasheet curves only at identical test conditions and confirm the attenuation by network analyser or EMI measurement in the real circuit; adjust the configuration or components if needed.

References

P. Matthews, Insertion Loss and Performance in EMI Filtering, Knowles Precision Devices blog, 2019.
EMI Filters – Hints and Tips, Syfer Technology (Knowles Precision Devices).
EMI Filters and Capacitors – Filter Catalogue, Syfer Technology (Knowles Precision Devices).
EMI Filters – Panel Mount EMI Filters, catalogue section, Knowles Precision Devices.
Basics of Noise Countermeasures, Chapter 6 – EMI suppression filters, Murata Manufacturing.
Noise Suppression by EMIFIL – Three-terminal capacitor structure and feedthrough capacitors, Murata Manufacturing.
Knowles – EPCI premium supplier page, Passive Components Blog.
Murata – EPCI premium supplier page, Passive Components Blog.

Article revised and technically reviewed in November 2026: corrected the C filter definition and the description of real versus ideal capacitor behaviour, added insertion loss, impedance, self-resonance and series-element equations, calculated insertion loss examples, impedance guidance for filter configurations, effects of load current, dielectric, earthing and shielding with source data, layout guidelines, checklist, FAQ and How-to, and added referenced sources.