Skip to main content
  • Created:
  • Updated:
  • Author:
    Takeshi Takatsudo

Linear Regulator Capacitor Selection

Understanding why linear regulators need specific capacitor values and placements for stable, low-noise operation.

Overview

Linear voltage regulators (like LM7812, LM7805, LM7912) require external capacitors for:

  1. Stability - Prevent oscillation
  2. Noise filtering - Remove high-frequency switching noise
  3. Transient response - Handle sudden load changes
  4. Decoupling - Provide local energy storage

The Two-Capacitor Strategy

Every linear regulator needs two types of capacitors on both input and output:

Complete Circuit (LM7812 example):

Input Capacitors: Regulator IC: Output Capacitors:
+13.5V ──┬────────────────┬───────────┬──────────────────┬───────────┬────────────────┬─── +12V
│ │ │ │ │ │
C20 C14 │ LM7812 │ C17 C21
470µF 470nF │ (TO-263-2) │ 100nF 470µF
electrolytic ceramic │ │ ceramic electrolytic
(farther) (CLOSE!) ┌──┤1 IN OUT 3├──┐ (CLOSE!) (farther)
│ │ │ │ │ │ │ │
│ │ ┌──┤2 GND │ │ │ │
│ │ │ └──────────────────┘ │ │ │
│ │ │ │ │ │
GND GND GND GND GND GND

Key points:

  • Input side: C20 (bulk) farther, C14 (ceramic) CLOSE to pin 1
  • Output side: C17 (ceramic) CLOSE to pin 3, C21 (bulk) farther
  • IC in the middle: Shows physical relationship between caps and IC pins

Why Two Different Capacitor Types?

Ceramic Capacitors (Small: 100nF, 470nF)

Characteristics:

  • Low ESR (Equivalent Series Resistance) < 10mΩ
  • Low ESL (Equivalent Series Inductance) < 1nH
  • Fast response to high-frequency noise
  • Small physical size

Purpose:

  • Filter high-frequency noise (1MHz - 100MHz+)
  • Handle fast transients (nanosecond response)
  • Provide local decoupling for IC

Why close to IC:

  • Even 1cm of trace adds ~10nH inductance
  • At high frequencies, inductance blocks current
  • Must minimize trace length for effectiveness

Electrolytic Capacitors (Large: 470µF)

Characteristics:

  • High capacitance (1000x larger than ceramic)
  • Higher ESR (~100mΩ typical)
  • Higher ESL (~10nH typical)
  • Slow response compared to ceramic

Purpose:

  • Provide bulk energy storage
  • Handle low-frequency ripple (100Hz - 10kHz)
  • Manage load transients (millisecond response)
  • Supply inrush current during startup

Why farther is OK:

  • Lower frequency operation is less sensitive to inductance
  • Large physical size makes close placement difficult
  • Bulk storage doesn't need ultra-fast response

Frequency Coverage

Together, the two capacitor types cover the full spectrum:

Frequency RangeHandled ByPurpose
DC - 10kHz470µF electrolyticBulk storage, load transients, ripple filtering
10kHz - 100kHzBoth working togetherMid-range filtering, switching noise
100kHz - 100MHz+100nF/470nF ceramicHigh-frequency decoupling, IC bypass

Why Different Values: Input vs Output?

Input Ceramic: 470nF (Larger)

DC-DC Switching → [470nF] → Linear Regulator
(Noisy!) (Heavy filtering)

Reasons:

  1. Input sees switching noise from upstream DC-DC converter (LM2596S)
  2. Switching frequency typically 50kHz - 500kHz generates harmonics
  3. Larger cap provides better attenuation at switching frequency
  4. Load transients - regulator draws varying current from input
  5. Datasheet recommendation: LM78xx typically specifies 0.33µF - 0.47µF

Example calculation:

Switching freq: 150kHz
Ripple current: 100mA
Required impedance: V_ripple / I_ripple = 10mV / 100mA = 0.1Ω

Ceramic impedance at 150kHz:
Z = 1 / (2π × 150kHz × 470nF) = 2.26Ω (too high!)

With 470nF: Provides some attenuation
Without it: Full switching noise reaches regulator → instability

Output Ceramic: 100nF (Smaller)

Linear Regulator → [100nF] → Clean Output
(Pre-filtered) (Light decoupling)

Reasons:

  1. Output already filtered by linear regulator's internal circuitry
  2. Main purpose is local high-frequency decoupling
  3. Smaller value sufficient for clean, regulated output
  4. Faster response at very high frequencies (smaller = lower ESL)
  5. Datasheet recommendation: LM78xx typically specifies 0.1µF

Why not larger?

  • Output is already low-noise from regulator
  • 100nF is optimal for HF decoupling (best impedance at 1-10MHz)
  • Larger caps can reduce phase margin (potential instability)

Physical Placement is Critical

Ceramic Capacitor Placement (CRITICAL)

IC Pin ──┤<-- 2mm max -->├── Ceramic Cap

✓ Trace length: 2-5mm
✗ Trace length: >10mm (inductance kills effectiveness)

PCB Layout Rules:

  1. Place RIGHT NEXT to IC pins (2-5mm max trace length)
  2. Short, wide traces (minimize inductance)
  3. Direct path to GND plane (via right next to cap)
  4. No other signals between cap and IC pin

Why so critical?

Trace inductance: L = 1nH/mm (typical)
10mm trace = 10nH

At 10MHz:
Z_inductance = 2π × 10MHz × 10nH = 0.63Ω

This impedance blocks high-frequency current!
The ceramic cap becomes useless if placed too far.

Electrolytic Capacitor Placement (Less Critical)

IC Pin ──┤<-- 10-50mm OK -->├── Electrolytic Cap

PCB Layout Rules:

  1. Can be placed 10-50mm from IC (still reasonable)
  2. Normal trace width (2-3mm copper)
  3. Connect to power plane (not critical if traces are adequate)
  4. Keep away from heat sources (electrolytics are temperature sensitive)

Why less critical?

  • Operates at lower frequencies where inductance matters less
  • Large physical size prevents very close placement anyway
  • Bulk storage function doesn't need ultra-fast response

Why Output Ceramic Must Be So Close: Preventing Oscillation

The Problem: Linear Regulators Can Oscillate

Short answer: The regulator oscillates, so kill the vibration near! 🎯

Linear regulators contain an internal feedback loop that can become unstable:

Internal Feedback Loop:
Output voltage → Error amp → Pass transistor → Output
↑ │
└────── Feedback ────────────┘

If phase shift occurs in this loop:
→ Positive feedback at certain frequencies
→ Oscillation! (typically 100kHz - 10MHz)

Without proper output capacitor:

Output voltage waveform:
╱╲╱╲╱╲╱╲╱╲╱╲╱╲
╱ ╲╱ (Oscillating at MHz frequency!)
╲╱

With ceramic cap VERY CLOSE:

Output voltage waveform:
──────────────── (Stable! ✅)

Why "CLOSE" is Critical: The Physics

Trace inductance blocks high-frequency current:

If ceramic cap is FAR (>5cm):

IC Output ──┬── 5cm trace (~50nH inductance) ── Ceramic cap ── GND
│ ↑
Oscillation Inductance blocks MHz currents! ❌
(1-10MHz) Cap can't "see" the oscillation
│ Vibration stays at IC output!
└──→ ╱╲╱╲╱╲╱╲╱╲ (Unstable!)

At MHz frequencies, even short traces act like inductors:

Trace LengthInductanceImpedance at 1MHzImpedance at 10MHz
1mm~1nH0.006Ω0.06Ω
1cm~10nH0.06Ω0.6Ω
5cm~50nH0.3Ω

At 10MHz with 5cm trace: 3Ω impedance blocks oscillation current from reaching the capacitor!

If ceramic cap is VERY CLOSE (<2mm):

IC Output ──┬── 2mm trace (~2nH) ── Ceramic cap ── GND
│ ↑
Oscillation Minimal inductance! ✅
(1-10MHz) Cap immediately shorts vibration to ground

└──→ ────────── (Stable! No oscillation)

Why it works:

  1. Oscillation current has very low impedance path to ground
  2. High-frequency vibrations are immediately damped
  3. Feedback loop remains stable
  4. Output stays clean and steady

Visual Analogy: Shock Absorber

Think of the output ceramic capacitor like a car shock absorber:

🚗 Bouncing Spring (Oscillation):
╱╲ Spring bouncing up/down
╱ ╲ (Like regulator oscillating)
╱ ╲
╱ ╲

🔧 Shock Absorber (Ceramic Cap):
Must be attached DIRECTLY to spring!

✅ Shock absorber attached directly:
Spring ── [shock absorber] ── chassis
(dampens vibration immediately)

❌ Shock absorber via long flexible cable:
Spring ── [5m rubber hose] ── [shock absorber] ── chassis
(too slow, spring keeps bouncing!)

Same principle for capacitors:

  • Regulator = Spring (can oscillate)
  • Ceramic cap = Shock absorber (dampens oscillation)
  • Trace inductance = Flexible cable (blocks effectiveness)
  • Solution: Attach directly! (minimize trace length)

The Numbers: Why <2mm Matters

PCB trace inductance rule of thumb: ~1nH per millimeter

Best practice trace lengths:

✅ Excellent: <2mm trace
- Inductance: ~2nH
- Impedance at 10MHz: 0.12Ω
- Result: Cap effectively shorts oscillation ✅

✅ Good: 2-5mm trace
- Inductance: ~5nH
- Impedance at 10MHz: 0.3Ω
- Result: Cap still effective, minor degradation

⚠️ Acceptable: 5-10mm trace
- Inductance: ~10nH
- Impedance at 10MHz: 0.6Ω
- Result: Reduced effectiveness, may work

❌ Poor: >10mm trace
- Inductance: >10nH
- Impedance at 10MHz: >0.6Ω
- Result: Oscillation likely! ❌

Input vs Output: Different Priorities

Why is output ceramic placement MORE critical than input?

SideWhat Happens If Cap Is FarConsequence
InputMore noise reaches ICRegulator filters it (PSRR helps) ✅
OutputOscillation can't be dampedRegulator oscillates!

Input capacitor far:

Switching noise → [far cap can't filter well] → Regulator IC

PSRR (Power Supply Rejection)
filters most of it ✅

Output (mostly OK)

Output capacitor far:

Regulator IC → Oscillation starts → [far cap can't damp] → Output
↑ │
└────────── Positive feedback ─────────────────────────────┘
(Oscillation continues! ❌)

Key insight:

  • Input: Regulator helps compensate for poor cap placement
  • Output: Nothing can save you if cap is too far! ⚠️

PCB Layout Checklist for Stability

Critical rules for output ceramic capacitor:

  • Distance: <2mm from IC output pin (ideal)
  • Trace width: As wide as possible (reduces inductance)
  • Via to ground: Place GND via right next to capacitor
  • No obstacles: Direct, straight path from IC pin to cap
  • Keep away from: High-speed signals, switching nodes

Example of GOOD layout:

        IC Output Pin

│ <── 1-2mm trace, 2mm wide

[Ceramic]

[Via] <── Ground via right next to cap

════╧════ (Ground plane)

Example of BAD layout:

        IC Output Pin

├── routes around other components

<5cm total trace length>


[Ceramic] <── TOO FAR! ❌

[Via]

Real-World Impact

What you'll see with improper placement:

Oscilloscope measurement (no load):

Bad placement (ceramic 3cm away):
┌─────────────────────────────┐
│ ╱╲╱╲╱╲╱╲╱╲╱╲╱╲╱╲╱╲╱╲╱╲╱╲ │ ← 500mV oscillation!
│╱ Expected 12.00V ╲│ ← Unstable
│ ╲│
└─────────────────────────────┘

Good placement (ceramic <2mm away):
┌─────────────────────────────┐
│─────────────────────────────│ ← Flat 12.00V
│ Stable output ✅ │ ← <1mV noise
│ │
└─────────────────────────────┘

Summary: The regulator oscillates at MHz frequencies. To kill this vibration, the ceramic capacitor must be physically close (<2mm) so trace inductance doesn't block the damping current. Think "shock absorber attached directly to spring" - distance kills effectiveness! 🎯

Common Mistakes and Fixes

❌ Mistake 1: Swapping Ceramic Values

Input: 100nF (too small for switching noise)
Output: 470nF (unnecessary, wastes space)

Result: Input switching noise gets through → regulator instability

Fix: Follow datasheet: Input 470nF, Output 100nF

❌ Mistake 2: Ceramic Too Far from IC

IC Pin ────── [20mm trace] ────── Ceramic Cap

Result: Trace inductance blocks high-frequency current → cap is useless

Fix: Place ceramic RIGHT NEXT to pin (2-5mm max)

❌ Mistake 3: Only Using Electrolytic Caps

Input: Only 470µF electrolytic
Output: Only 470µF electrolytic

Result: No high-frequency filtering → oscillation, instability

Fix: Always pair electrolytic with ceramic

❌ Mistake 4: Using Low-Quality Ceramics

Using Y5V dielectric ceramic (capacitance varies wildly)

Result: Capacitance drops 80% at rated voltage and temperature

Fix: Use X7R or X5R dielectric (stable across temperature/voltage)

❌ Mistake 5: Wrong Electrolytic Polarity (Negative Rails)

-12V rail: Negative terminal to GND (WRONG!)

Result: Electrolytic explodes or fails

Fix: Negative rail → Negative terminal to -12V, Positive terminal to GND

Practical Examples from This Project

Positive Rails (+12V, +5V)

LM7812 (TO-263-2):

Input:
- C14: 470nF ceramic X7R (RIGHT NEXT to pin 1)
- C20: 470µF electrolytic (10-20mm from pin 1)

Output:
- C17: 100nF ceramic X7R (RIGHT NEXT to pin 3)
- C21: 470µF electrolytic (10-20mm from pin 3)

Why this works:

  • DC-DC converter upstream generates 150kHz switching noise
  • C14 (470nF) filters this switching noise at input
  • C20 (470µF) provides bulk storage for load transients
  • C17 (100nF) decouples high-frequency noise at output
  • C21 (470µF) provides output bulk capacitance

Negative Rail (-12V)

LM7912 (TO-252-3):

Input:
- C16: 470nF ceramic (CLOSE to pin 1)
- C24: 470µF electrolytic (farther) ※ Negative to -13.5V, Positive to GND

Output:
- C19: 100nF ceramic (CLOSE to pin 2)
- C25: 470µF electrolytic (farther) ※ Negative to -12V, Positive to GND

Critical polarity note:

  • For negative voltage rails, electrolytic polarity is REVERSED
  • Negative terminal connects to negative voltage (e.g., -12V)
  • Positive terminal connects to GND (0V)

Advanced: ESR and Stability

Why ESR Matters

Linear regulators need some ESR (Equivalent Series Resistance) in the output capacitor for stability:

Too low ESR → Phase shift → Oscillation
Optimal ESR → Stable operation
Too high ESR → Poor transient response

Typical requirements (from datasheets):

  • LM78xx: Output cap ESR should be 0.1Ω - 10Ω
  • Pure ceramic (ESR < 10mΩ) can cause instability
  • Electrolytic + ceramic combination provides optimal ESR

Our design:

  • C21/C23/C25 (electrolytic 470µF): ESR ~100mΩ (provides damping)
  • C17/C18/C19 (ceramic 100nF): ESR < 10mΩ (provides HF decoupling)
  • Together: Optimal combination for stability and performance

Load Transient Response

When load current changes suddenly:

Load step: 0A → 1A in 1µs

Without capacitors:
- Output dips 2V (regulator can't respond fast enough)
- Takes 100µs to recover

With proper capacitors:
- Ceramic provides instant current (sub-µs response)
- Electrolytic provides sustained current (µs-ms response)
- Output dips only 50mV
- Recovers in 10µs

Testing and Validation

What to Check on PCB

  1. Ceramic placement: Measure distance from cap to IC pin
  • ✓ Goal: < 5mm
  • ✗ Problem: > 10mm
  1. Output ripple: Measure with oscilloscope (20MHz bandwidth)
  • ✓ Goal: < 1mVp-p at full load
  • ✗ Problem: > 10mVp-p (missing/far ceramic caps)
  1. Load transient: Step load from 0% to 100%
  • ✓ Goal: < 100mV deviation, < 50µs recovery
  • ✗ Problem: > 500mV deviation (missing bulk caps)
  1. Oscillation check: Probe output with 100MHz scope, no load
  • ✓ Goal: Clean DC, no oscillation
  • ✗ Problem: MHz oscillation (ceramic too far or missing)

Summary: Quick Reference

ParameterInput SideOutput SideReason
Ceramic value470nF100nFInput needs more switching noise filtering
Ceramic typeX7R/X5RX7R/X5RStable across temperature
Ceramic placementRIGHT NEXT to pinRIGHT NEXT to pinMinimize trace inductance
Electrolytic value470µF470µFBulk storage and load transients
Electrolytic placement10-50mm from pin OK10-50mm from pin OKLess critical for low frequencies
Electrolytic polarity+ to voltage, - to GND+ to voltage, - to GND(Reversed for negative rails!)

Key Takeaways

  1. Always use BOTH ceramic and electrolytic - they work together across different frequencies
  2. Ceramic placement is CRITICAL - must be right next to IC pins (< 5mm)
  3. Different values for input/output - input handles more noise (470nF), output is cleaner (100nF)
  4. Electrolytic provides bulk storage - placement less critical (10-50mm OK)
  5. Negative rail polarity - don't forget to reverse electrolytic polarity!
  6. Use quality parts - X7R/X5R ceramics, low-ESR electrolytics
  7. PCB layout matters - short, wide traces for ceramics, good ground plane

Further Reading

  • LM7812 Datasheet: Section "Application Information" for recommended capacitor values
  • LM7805 Datasheet: See "Output Capacitor" section for stability requirements
  • LM7912 Datasheet: Note reversed polarity requirements for negative regulators
  • Decoupling Capacitor Guide: Understanding ESR, ESL, and frequency response
  • PCB Layout Guide: High-frequency design techniques for power supplies