Why Power Distribution Networks Combine Bulk and MLCCs

In high-speed electronics, maintaining a stable power supply requires a power distribution network (PDN) that exhibits low impedance across a vast range of frequencies. Because no single capacitor can perform effectively at both very low and very high frequencies due to inherent physical limitations, engineers combine bulk electrolytic capacitors with arrays of low-inductance multilayer ceramic capacitors (MLCCs). This strategic combination bridges the gap between different frequency domains, ensuring that slow, large current demands and ultra-fast transient switching currents are simultaneously satisfied without causing destructive voltage droops or electromagnetic interference.

The Goal of Broadband Decoupling

High-speed integrated circuits (ICs), such as processors and FPGAs, contain millions of transistors switching on and off in picoseconds. This rapid switching creates sudden, dynamic demands for current across a wide frequency spectrum—from near-DC up to several gigahertz.

If the PDN has high impedance at any of these frequencies, the sudden draw of current will cause the supply voltage to drop, which can trigger logic errors or system crashes. Broadband decoupling aims to keep the PDN impedance below a specific target value across the entire operational bandwidth of the device by placing local charge reservoirs (capacitors) right next to the load.

The Physical Limitations of Real Capacitors

An ideal capacitor decreases its impedance infinitely as frequency rises. However, real-world capacitors possess parasitic elements: Equivalent Series Resistance (ESR) and Equivalent Series Inductance (ESL).

Because of ESL, a capacitor only acts like a capacitor up to its Self-Resonant Frequency (SRF). Above the SRF, the parasitic inductance dominates, and the component begins to act like an inductor, causing its impedance to rise with frequency. Therefore, a capacitor is only effective at decoupling frequencies up to and slightly beyond its SRF.

The Role of Bulk Electrolytic Capacitors

Bulk capacitors, typically aluminum electrolytic or tantalum, offer large capacitance values ranging from tens to thousands of microfarads. Their primary role is to handle low-frequency power demands, generally from DC up to a few hundred kilohertz.

When a circuit powers up or shifts into a high-power processing state, it draws a massive amount of current. The distant voltage regulator module (VRM) cannot respond instantly due to the inductance of the circuit board traces. Bulk capacitors act as large energy reservoirs on the board, supplying this substantial charge until the VRM can adjust. However, because bulk capacitors are physically large, they have high ESL, making them completely ineffective at responding to high-frequency transients.

The Role of Low-Inductance MLCC Arrays

To handle high-frequency transients—ranging from a few megahertz to several gigahertz—engineers use MLCCs. These capacitors have much smaller capacitance values (nanofarads to picofarads) and are housed in tiny surface-mount packages, which gives them extremely low ESL.

Low ESL allows MLCCs to push their self-resonant frequencies into the megahertz and gigahertz ranges, providing a fast, low-impedance path for high-speed switching currents. Engineers use arrays of these MLCCs in parallel for two specific reasons:

1. Lowering Total Inductance: Placing multiple MLCCs of the same value in parallel divides their collective ESL and ESR, driving the impedance floor even lower.
2. Expanding Frequency Coverage: Using an array of different capacitance values (e.g., 1 µF, 0.1 µF, and 0.01 µF) staggers their self-resonant frequencies. Each smaller value takes over as the larger value becomes inductive, creating a continuous, flat, low-impedance profile across a wide frequency band.

Managing Anti-Resonance

Combining bulk capacitors with MLCCs requires careful balance. When two different capacitor types are placed in parallel, there is a frequency zone where the larger capacitor acts inductively (above its SRF) while the smaller capacitor still acts capacitively (below its SRF).

This creates a parallel LC tank circuit, resulting in a sharp impedance spike known as an anti-resonance peak. Engineers mitigate this by selecting bulk capacitors with a controlled amount of ESR, which dampens the resonance peak, and by using closely spaced capacitor values in the MLCC array to smooth out the impedance curve.