A decoupling capacitor (also called a bypass capacitor) is a capacitor placed between a power supply rail and ground, close to an IC's power pins, to suppress high-frequency noise on the supply and provide a local reservoir of charge for transient current demands.
In practice
In digital and mixed-signal PCB design, decoupling capacitors are generally placed as close as possible to each IC's VCC/VDD and GND pins, though specific placement constraints vary by part, package, and vendor recommendation. The short trace length minimizes parasitic inductance, which would otherwise limit how quickly the capacitor can respond to a sudden load step. A common rule of thumb is to place the capacitor before the via, not after, with the goal of minimizing the loop inductance between the capacitor and the IC's power pin rather than simply controlling current path order. Typical values for high-frequency decoupling are often 100 nF ceramic (X5R or X7R dielectric) per power pin, though the appropriate value varies by IC, package, and PDN requirements, and many designs use different values or multiple capacitors per rail.
A single capacitor value is often not enough. Because a real capacitor has series inductance (ESL), it becomes inductive above its self-resonant frequency and stops suppressing noise effectively. Pairing a small ceramic (e.g., 100 nF) with a larger bulk capacitor (e.g., 10 uF ceramic or electrolytic) covers a wider frequency range. The blog post "Two Capacitors Are Better Than One" explores this interaction in detail, including the anti-resonance dip that can appear between two capacitors of different values and how to manage it.
For power stages such as H-bridges and motor drivers, decoupling requirements are more demanding. High-current switching transients can exceed what small ceramics can supply, and the inductance of long power wiring makes the problem worse. In these designs, low-ESR, low-ESL film or electrolytic capacitors are added as DC link capacitors to absorb the bulk energy. The blog post "Lost Secrets of the H-Bridge, Part IV: DC Link Decoupling and Why Electrolytic Capacitors Are Not Enough" covers why electrolytic capacitors alone are often insufficient at high switching frequencies, and why a parallel ceramic is typically needed.
A less obvious application is supply filtering for sensitive analog or RF sections, where a decoupling network is combined with a small series resistor or ferrite bead to form an LC or RC filter that isolates the analog rail from digital noise. The blog post "The Other Kind of Bypass Capacitor" addresses this use case specifically. Choosing the wrong dielectric (e.g., Y5V ceramic, which loses most of its capacitance under DC bias) is a common pitfall: always verify capacitance under actual operating voltage, especially for ceramic capacitors in small packages where DC bias derating can reduce effective capacitance by 50-80%.
Discussed on EmbeddedRelated
Frequently asked
How close is 'close enough' when placing a decoupling capacitor?
As close as physically possible to the IC's power pin, ideally within 1-2 mm. The goal is to minimize the loop inductance formed by the
capacitor, its traces, and the IC's power and ground pins. At 100 MHz, even a few nH of parasitic inductance significantly reduces the capacitor's effectiveness. On dense boards, placing the capacitor on the back side of the PCB directly under the power pin (connected through a via) can be effective, though whether it is beneficial depends on the return path and via placement for the specific package and board stackup.
Why use multiple capacitor values instead of one large capacitor?
Each
capacitor value has a self-resonant frequency (SRF) above which it becomes inductive and stops decoupling. A 100 nF 0402 ceramic might have an SRF around 100-200 MHz, while a 10 uF ceramic resonates in the low MHz range. Using both in parallel covers a broader frequency span. Be aware that two capacitors in parallel can produce an anti-resonance (impedance peak) between their individual resonant frequencies, which can actually worsen noise at that specific frequency. The blog post 'Two Capacitors Are Better Than One' covers how to predict and mitigate this.
What is the difference between a decoupling capacitor and a bulk capacitor?
The terms describe the same component used in two different roles. A decoupling (or bypass)
capacitor is typically a small ceramic (commonly in the range of 1 nF to 100 nF, though larger values and other technologies are also used) placed directly at an IC pin to handle fast, high-frequency transients. A bulk capacitor (typically 1 uF to hundreds of uF, often electrolytic or tantalum) is placed at the board or module level to handle slower, larger transient loads and to maintain voltage stability across multiple ICs. In practice, a well-designed power distribution network uses both.
Does the capacitor dielectric type matter?
Yes, significantly. X7R and X5R ceramics are common choices for decoupling due to their stability over temperature, though their capacitance can still derate substantially under DC bias, particularly in small packages and high-voltage-rated parts. C0G (NP0) ceramics are very stable but are only practical in small values (up to a few nF). Y5V and Z5U dielectrics can lose 50-80% of their rated capacitance at operating voltage and temperature, making them unreliable for decoupling despite having attractive rated values. Always check the manufacturer's DC bias derating curves, particularly for small-package, high-capacitance parts.
Do microcontrollers with internal voltage regulators still need external decoupling capacitors?
Typically yes. Many MCUs (for example, STM32 and nRF52 series) have internal LDOs or reference regulators whose datasheets explicitly require external
bypass capacitors on VCAP, VDDIO, or VDDA pins to ensure regulator stability and correct analog performance. Omitting these
capacitors can cause oscillation in the internal
LDO, degraded
ADC accuracy, or unexpected resets. Always follow the recommended decoupling network in the device datasheet and reference schematic.
Differentiators vs similar concepts
The terms 'decoupling
capacitor' and '
bypass capacitor' are used interchangeably in most contexts and refer to the same physical component placed at a power pin. Some engineers use 'bypass' specifically for high-frequency noise shunting and 'decoupling' for isolating one circuit section from another, but this distinction is not universally observed. A bulk capacitor serves a related but distinct purpose: it handles lower-frequency, higher-energy transients at the supply input or across a power bus rather than at individual IC pins.
