What is the difference between bulk decoupling and local bypass?

Bulk decoupling uses large capacitors (10-100 µF electrolytic or tantalum) placed near the power supply connector or regulator output to provide energy storage for low-frequency transients (DC to ~1 MHz) and filter switching regulator ripple. Local bypass uses small capacitors (10-100 nF ceramic) placed directly at each IC's power pins to provide high-frequency decoupling (1 MHz to 1 GHz) with minimal inductance. Both are needed: bulk capacitors cannot respond to fast transients due to their ESL (equivalent series inductance), while small ceramics cannot store enough energy for sustained current demands. A typical RF board uses 3-4 bulk caps per power rail plus one local bypass cap per IC power pin.

How do you calculate bulk decoupling capacitance?

The minimum bulk capacitance is determined by the maximum allowed voltage droop during a transient current demand: C = I_transient × t_response / ΔV_max. For a GaN PA that draws 2 A pulsed current with 1 µs rise time and maximum 100 mV voltage droop, C = 2 × 1µs / 0.1 = 20 µF. In practice, 2-3× this value (47-100 µF) is used to account for capacitor derating, ESR voltage drop, and temperature effects. The capacitor's ESR must also be low enough that I × ESR < ΔV_max: for 2 A and 100 mV, ESR < 50 mΩ.

Why is ESL important for decoupling capacitors?

ESL (equivalent series inductance) determines the capacitor's self-resonant frequency (SRF = 1/(2π√(LC))) and its impedance at high frequencies. Above the SRF, the capacitor becomes inductive and its impedance rises, making it ineffective for decoupling. A 100 µF electrolytic with 5 nH ESL has SRF = 225 kHz, meaning it only provides effective decoupling below about 100 kHz. A 100 nF ceramic with 0.5 nH ESL has SRF = 22 MHz, covering the MHz range. This is why multiple capacitor values are paralleled in a PDN design: each covers a different frequency decade.

What is Bulk Decoupling? | Power Supply Filtering for RF

Q: Why is ESL important for decoupling capacitors?

ESL (equivalent series inductance) determines the capacitor's self-resonant frequency (SRF = 1/(2π√(LC))) and its impedance at high frequencies. Above the SRF, the capacitor becomes inductive and its impedance rises, making it ineffective for decoupling. A 100 µF electrolytic with 5 nH ESL has SRF = 225 kHz, meaning it only provides effective decoupling below about 100 kHz. A 100 nF ceramic with 0.5 nH ESL has SRF = 22 MHz, covering the MHz range. This is why multiple capacitor values are paralleled in a PDN design: each covers a different frequency decade.

Definition & Role in PDN

Bulk decoupling is the practice of placing large-value capacitors (10-100+ µF) near the power supply entry point or voltage regulator output on an RF PCB to provide low-frequency energy storage and filtering. These capacitors serve as a local charge reservoir that supplies transient current demands before the voltage regulator can respond, preventing voltage droops that would modulate the RF circuit's operating point and degrade performance.

In the power distribution network (PDN) impedance profile, bulk decoupling capacitors dominate the impedance from DC to approximately 1 MHz, while local bypass capacitors (100 nF MLCC at each IC pin) cover 1 MHz to 1 GHz. The combined PDN must maintain impedance below the target Z_target = ΔV_max/I_transient across the entire frequency range. For a sensitive RF IC drawing 500 mA with 50 mV tolerance, Z_target = 100 mΩ, requiring careful selection of bulk capacitor values, ESR, ESL, and PCB plane geometry.

Key Formulas

Minimum Bulk Capacitance:

C_bulk = I_transient × t_response / ΔV_max

2 A pulse, 1 µs, 100 mV droop: C = 20 µF (use 47-100 µF with margin)

Self-Resonant Frequency:

SRF = 1 / (2π√(L_ESL × C))

100 µF, 5 nH ESL: SRF = 225 kHz | 100 nF, 0.5 nH: SRF = 22 MHz

Target PDN Impedance:

Z_target = ΔV_max / I_transient

Decoupling Capacitor Comparison

Parameter	Bulk Electrolytic	Bulk Tantalum	Bulk MLCC	Local Bypass MLCC
Capacitance	10-1000 µF	10-100 µF	10-100 µF	10-100 nF
ESR	50-500 mΩ	20-200 mΩ	2-10 mΩ	5-50 mΩ
ESL	5-15 nH	2-5 nH	0.5-2 nH	0.3-1 nH
SRF (typical)	50-500 kHz	200 kHz-2 MHz	1-5 MHz	20-200 MHz
Effective Range	DC-100 kHz	DC-1 MHz	DC-5 MHz	1 MHz-1 GHz
Placement	Near regulator	Near regulator	Near regulator	At IC power pin
Size (typ.)	Large (can)	Medium (SMD)	Medium (1210)	Small (0402)

Practical Application

A pulsed S-band radar transmitter module draws 8 A from a 28 V rail during the 10 µs transmit pulse at 10% duty cycle. The GaN PA's drain voltage must remain within ±100 mV during the pulse. Bulk decoupling requires C = 8 × 10µs / 0.1 = 800 µF minimum. The design uses four 330 µF tantalum capacitors (total 1320 µF) with 30 mΩ parallel ESR, supplemented by ten 10 µF X7R MLCCs for mid-frequency decoupling and twenty 100 nF C0G MLCCs at each PA drain pin. The PDN impedance simulation confirms Z < 12.5 mΩ (100 mV / 8 A) from DC to 100 MHz, ensuring the PA drain voltage remains stable throughout each pulse.

Frequently Asked Questions

Bulk decoupling vs local bypass?

Bulk: large caps (10-100 µF) near regulator for DC-1 MHz. Local bypass: small caps (10-100 nF) at each IC pin for 1 MHz-1 GHz. Both needed because bulk caps have too much ESL for high frequencies, and small caps lack energy storage for sustained demands.

How to calculate bulk capacitance?

C = I_transient × t_response / ΔV_max. For 2 A, 1 µs, 100 mV: C = 20 µF. Use 2-3× margin. ESR must satisfy I × ESR < ΔV_max.

Why does ESL matter?

ESL sets the self-resonant frequency: SRF = 1/(2π√(LC)). Above SRF, the cap becomes inductive and ineffective. 100 µF electrolytic: SRF ~225 kHz. 100 nF ceramic: SRF ~22 MHz. Multiple values are paralleled to cover all frequency decades.

Bulk Decoupling