Filter Capacitors in EMC Design

Capacitors play a vital role in electromagnetic compatibility (EMC) by helping to control and manage electromagnetic interference (EMI) and improve the immunity of electronic devices. The following are some of the main applications of capacitors in EMC.

Filtering

Capacitors are key components in filters. In electronic equipment, placing capacitors on signal or power lines can effectively filter out high-frequency noise and EMI, ensuring that the device's power and signal integrity are not compromised by external electromagnetic waves.

Power Supply Decoupling

In electronic circuits, capacitors are used for power supply decoupling to ensure that components receive a stable power source during operation. This helps prevent noise on the power lines from propagating to critical electronic components.

RF Interference Suppression

Radio Frequency (RF) interference is a high-frequency disturbance that often affects wireless communication equipment and other high-frequency electronic devices. Capacitors can be used to absorb and suppress these RF signals, preventing them from entering or exiting the device.

Electrostatic Discharge (ESD) Protection

In certain environments, electrostatic discharge can damage equipment. Capacitors can be used to absorb and dissipate electrostatic energy, thereby reducing the impact of ESD on the device.

Differential-Mode Noise Filtering

In analog circuits, capacitors are commonly used for filtering differential-mode signals, helping to reduce the impact of noise on the signal.

Common-Mode Rejection

Capacitors are also used in common-mode rejection circuits to prevent common-mode signals (interference signals that act on both conductors of a circuit simultaneously) from affecting the device.

In EMC design, the selection and layout of capacitors are critical. Proper capacitor selection can significantly improve a device's electromagnetic compatibility, prevent mutual interference between different parts, and ensure stable operation in its intended electromagnetic environment.

Capacitor Self-Resonance

The capacitors we use for filtering are not ideal. In a system, they behave as an ideal capacitor in series with an inductor and a resistor.

A multi-layer ceramic capacitor (MLCC) mounted on a PCB will have a parasitic inductance of nearly 5 nH, plus a lead resistance of about 30 m?. Its frequency characteristic is not that of an ideal low-pass filter. Instead, its actual insertion loss characteristic behaves like a band-stop filter centered at its self-resonant frequency (SRF).

When two capacitors are connected in parallel, an anti-resonance problem can occur due to their Equivalent Series Inductance (ESL) and Equivalent Series Resistance (ESR).

In a wide frequency band from approximately 15 MHz to 175 MHz, the impedance of the parallel capacitors can be greater than that of a single large capacitor. Due to the resonance between the two capacitors, an impedance peak occurs at 150 MHz. In this frequency range, only a small portion of the energy generated by other parts of the system can be shunted to the ground plane.

When designing general circuits, engineers typically focus on parameters like capacitance, voltage rating, package size, operating temperature range, and temperature drift. However, in high-speed circuits, power systems, or critical clock circuits, a capacitor is more than just a capacitor; it is a circuit composed of an equivalent capacitance, equivalent resistance, and equivalent inductance.

The performance of this equivalent circuit is affected by many factors. When selecting such capacitors, it is necessary to consider not only the aforementioned parameters but also the equivalent parameters at specific frequencies. Taking a 1 µF Murata capacitor as an example, at its resonant frequency, the equivalent capacitance is 602.625 nF, the equivalent resistance is 11.5356 m?, and the equivalent inductance is 471.621 pH. The performance of an ideal capacitor and a real capacitor will differ.

In practice, many engineers copy capacitor arrangements from reference designs or other boards, assuming it will work for their product. This is not always the case. Different product applications and structures can lead to different PCB layer stacks and current-carrying planes, all of which cause inconsistencies in the power delivery network (PDN).

Impact of ESR on Parallel Capacitor Frequency Response

The impedance peak at anti-resonance is inversely proportional to the capacitor's ESR. As board design levels and component performance improve, the ESR of capacitors decreases, which can increase the anti-resonance impedance peak. The shape and position of this parallel resonance peak depend on the PCB design and capacitor selection.

Key principles to understand:

1. As ESR decreases, the impedance at the series resonance point decreases, but the impedance at the anti-resonance point increases.
2. When n identical capacitors are used in parallel, the minimum impedance can be lower than ESR/n.
3. When multiple capacitors are in parallel, the lowest impedance point does not necessarily occur at the self-resonant frequency of any single capacitor.
4. For a given number of capacitors, a better choice is to spread the capacitance values over a wide range with moderate ESR for each. A poor choice is to use only a few capacitance values, all with very low ESR.

Impact of ESL on Parallel Capacitor Frequency Response

ESL varies with capacitor package and construction. 

The ESL, along with the capacitance value, determines the resonant frequency of a single capacitor and the anti-resonant frequency range of parallel capacitors. In practical design, you should try to select capacitors with low ESL.

Capacitor Selection

For RF design, ceramic, polyester, and polystyrene film capacitors are good choices. For EMI filters, the requirements for the dielectric material are not strict; common dielectrics like X7R, Y5V, and Z5U are suitable. The absolute capacitance value, temperature coefficient, and voltage coefficient are generally not critical. Different types and values of capacitors have different filtering ranges. 

For the same 0805 package, a 0.01 µF ceramic capacitor has better high-frequency filtering characteristics than a 0.1 µF capacitor. For boards with operating frequencies above 50 MHz, it is recommended to use 0.01 µF filter capacitors instead of the commonly used 0.1 µF.

Power Supply Input and Output Capacitors

Capacitors on the input and output loops of a power module are typically called filter capacitors. Their role is to ensure a stable input and output voltage. In a power module, the principle for placing filter capacitors is "large to small." The capacitors should be placed in order from largest to smallest capacitance value in the direction of current flow.

When designing the power supply, ensure that traces and copper pours are sufficiently wide and that there are enough vias to handle the required current. The width and number of vias should be evaluated based on the current magnitude.

Power Supply Input Capacitor

The power supply input capacitor forms a current loop with the switching circuit. This loop experiences large current changes, with an amplitude of Iout at the switching frequency. These rapid current changes (high di/dt) are generated during the DC-DC chip's switching process.

For a synchronous buck converter, the freewheeling current path passes through the chip's GND pin. The input capacitor should be connected between the chip's GND and Vin pins with the shortest and widest path possible.

The smaller the area of this current loop, the less it will radiate.

Decoupling and Bypass Capacitors

1. Select capacitors based on the self-resonance characteristics provided in the supplier's datasheet to meet the design's clock speed and noise frequency requirements.
2. Add as many capacitors as possible within the required frequency range. For example, a 22 nF capacitor has a self-resonant frequency near 11 MHz and a useful impedance (Z < 1 ?) range from 6 MHz to 40 MHz. You can add multiple capacitors to achieve the necessary decoupling level in this band.
3. Place at least one decoupling capacitor as close as possible to each power pin of an IC to minimize parasitic impedance.
4. Place bypass capacitors on the same PCB layer as the IC whenever possible. A key point: in both layouts, the Vcc net connects to the Vcc plane at only one point. This forces noise from inside and outside the IC to pass through this single via to the power plane. The via's added impedance helps prevent noise from spreading to the rest of the system.
5. For multi-clock systems, the power plane can be partitioned. Each section can use a capacitor of the correct value. The slot separating the power planes isolates noise from sensitive components in other parts of the board while providing mid-value capacitance separation.
6. For systems with clock frequencies that vary over a wide range, selecting bypass capacitors is difficult. A good solution is to place two capacitors with a capacitance ratio of approximately 2:1 in parallel. This provides a wider low-impedance area and a broader bypass frequency range. The impedance peak still occurs but is less than 15 ?, and the usable frequency range (impedance < 15 ?) extends from about 3.25 MHz to 100 MHz. This multi-capacitor method should only be used when a single IC requires a wide bypass frequency range that a single capacitor cannot cover. The capacitance values must be kept within a 2:1 ratio to prevent the impedance peak from exceeding the usable range.

High-speed IC power pins require a sufficient number of decoupling capacitors, ideally one for each pin. In practice, if space is limited, some can be omitted judiciously.

The capacitance values for IC power pin decoupling are usually small, such as 0.1 µF or 0.01 µF, with corresponding small packages like 0402 or 0603. When placing decoupling capacitors, note the following:

1. Place them as close as possible to the power pin; otherwise, they may not provide effective decoupling. In theory, capacitors have a limited decoupling radius, so the proximity principle must be strictly followed.
2. The traces from the decoupling capacitor to the power pin should be as short and wide as possible, typically 8-15 mils (1 mil = 0.0254 mm). Widening the trace reduces inductance and ensures power integrity.
3. After fanning out from the capacitor pads, the power and ground traces should have vias placed nearby to connect to the power and ground planes. These traces should also be widened, and larger vias should be used (e.g., use a 10-mil via instead of an 8-mil one if possible).
4. Keep the decoupling loop area as small as possible. Examples of common decoupling capacitor placements for SOP and QFP packages are shown in various figures.
5. For BGA packages, decoupling capacitors are typically placed on the underside of the BGA. Due to the high pin density of BGAs, the number of decoupling capacitors is often limited, but as many as possible should be placed.

Energy Storage Capacitor Design

Energy storage capacitors (also known as bulk capacitors) ensure that the supply voltage does not drop during sudden heavy load transitions. They can be categorized as board-level and device-level storage capacitors:

A. Board-level storage capacitors: Ensure that the supply voltage across the board does not drop when the load changes rapidly. On high-frequency, high-speed boards, it is recommended to distribute a certain number of larger value tantalum capacitors (1 µF, 10 µF, 22 µF, 33 µF) to maintain a consistent voltage level across the board.

B. Device-level storage capacitors: Ensure that the local supply voltage around a device does not drop during rapid load changes. For devices with high operating frequencies, high speeds, and large power consumption, it is recommended to place 1-4 larger value capacitors (1 µF, 10 µF, 22 µF, 33 µF) around them.

The design of storage capacitors should be distinct from that of decoupling capacitors. Here are some design suggestions:

1. When a board has multiple supply voltages, use only one capacitance value for the storage capacitors for a single voltage rail. Surface-mount tantalum capacitors are generally chosen, with values like 10 µF, 22 µF, or 33 µF selected as needed.
2. Group chips by their supply voltage, and distribute the storage capacitors evenly within each group.

The purpose of a storage capacitor is to provide energy in the shortest possible time when an IC needs it. Storage capacitors typically have large capacitance values and larger packages. In a PCB layout, storage capacitors can be placed further from the device than decoupling capacitors, but not too far. The common fanout method for storage capacitors is also illustrated.

Capacitor fanout principles are as follows:

1. Keep traces short and wide to minimize parasitic inductance.
2. For storage capacitors or components with high current, use multiple vias.
3. The best electrical performance is achieved with via-in-pad, but practical manufacturing constraints must be considered.

Using Capacitors in Filter Circuits

An EMC filter is typically a low-pass filter composed of inductors (L) and capacitors (C). The difference between various LC filter structures lies in how the components are connected. The effectiveness of an LC filter depends not only on its structure but also on the impedance of the connected network. For example, a single-capacitor filter works well in a high-impedance circuit but poorly in a low-impedance one. Traditionally, filter characteristics are described under the condition of 50-ohm terminations at both ends. However, in practice, the source impedance (Zs) and load impedance (Zl) are complex and may be unknown at the frequencies to be suppressed. If one or both ends of the filter are connected to reactive components, resonance can occur, turning insertion loss into insertion gain at certain frequencies.

In a signal path, L and C form a low-pass filter. Since the source and load impedances at a specific frequency are unknown, we must avoid parameter combinations that filter out useful frequency components. In many cases, engineers tend to use 1 nF (102) or 100 nF (104) capacitors without calculation, which can sometimes be counterproductive.

Capacitor resonance is typically caused by the capacitance in conjunction with the equivalent inductance of its own leads or connecting traces. The resonant frequency is given by the formula:

F = 1 / (2 * π * √(LC))

In a series LC circuit at resonance, the impedance is at a minimum, acting like a short circuit. In a parallel LC circuit at resonance, the impedance is at a maximum, acting like an open circuit. The L and C components are in a shunt configuration, which forms a series resonant circuit to ground. If the resonant frequency of this LC circuit matches the interference frequency we want to filter, the path to ground acts as a short circuit, effectively filtering the noise.

For example, consider a signal path with a useful frequency of 5 MHz and an inductor (L) of 1 µH. We want to filter out a 10 MHz interference signal. We must avoid adding a filter capacitor (C) that creates a resonance near 5 MHz, which would filter out the useful signal. If we empirically choose a 1000 pF capacitor, the resonant frequency calculates to 5.03 MHz. At this frequency, the LC combination acts as a short to ground, shunting our useful signal and causing the circuit to fail. Instead, we should select the capacitor value based on the interference frequency we want to remove. Using the resonance formula, the required capacitance for a 10 MHz resonance is 253.3 pF; we can select the closest standard value. It is also important to use short leads for through-hole components, or preferably, use surface-mount devices to minimize ESL. This shows that correctly selecting the filter structure and component values is crucial. While experience is valuable, relying on empirical values can be problematic, especially when dealing with harmonics of useful frequencies. Always estimate the required values using proper calculations.

When filtering noise on a wire harness, a low-cost capacitor might be chosen. However, this can sometimes divert interference to other paths, creating an antenna effect and increasing radiated emissions. When selecting a capacitor, understand that it only transfers energy; it does not dissipate it. A capacitor is only effective for filtering when connected to a low-impedance network. In practice, this energy-transfer characteristic is often overlooked by engineers, who may assume that connecting to ground is a universal solution.

Common-Mode Capacitors

"Common-mode capacitor" typically refers to the capacitance to ground in a differential signal path. It is an important parameter, especially in differential amplifiers and communication systems.

A differential signal consists of two components:

• Differential-Mode Signal: The difference between the two input signals.
• Common-Mode Signal: The average or common part of the two input signals.

A common-mode capacitor is the capacitance of the common-mode portion of the signal to ground. This capacitance can be detrimental, especially in differential amplifiers. Ideally, a differential amplifier only amplifies the differential-mode signal and rejects the common-mode signal. However, real-world circuits are imperfect, and common-mode capacitance is one such imperfection.

Common-mode capacitance can cause problems such as:

• Common-Mode Noise: If there is common-mode noise on the input signals, the common-mode capacitance can cause this noise to be amplified, affecting circuit performance.
• Common-Mode Rejection Ratio (CMRR): This is a key metric that measures a differential amplifier's ability to reject common-mode signals. The presence of common-mode capacitance can degrade the CMRR.

Filters generally do not use separate differential-mode inductors because a common-mode choke will inherently have some differential-mode inductance due to winding imbalances. If differential-mode interference is severe, a dedicated differential-mode inductor should be added.

Differential-Mode Capacitors

Capacitors exhibit high impedance at low frequencies and low impedance at high frequencies. A filter utilizes the low impedance of a capacitor at high frequencies to short-circuit differential-mode interference. For example, at 50 Hz, a capacitor's impedance is nearly infinite, providing no attenuation. At 500 kHz, its impedance is very low. A differential-mode current at 500 kHz with a 50-ohm load and a capacitor impedance of 0.05 ohms will be almost entirely shunted by the capacitor.

In this scenario, the capacitor shunts 99.9% of the differential-mode interference current, while the load receives only 0.1%. This means the capacitor provides about 30 dB of attenuation for the 500 kHz differential-mode interference.