MOSFET Capacitor Selection: Gate, Bootstrap, and Snubber Rules

When designing or repairing a switching circuit, the 'MOSFET capacitor' you need depends entirely on its physical location in the schematic. For bootstrap circuits, use a 100nF to 1µF X7R ceramic. For gate filtering and soft-start, use 1nF to 10nF C0G/NP0 ceramics. For drain-source snubbers, use high-voltage C0G/NP0 ceramics or polypropylene film. Never use Y5V or high-K Class II ceramics for snubbers or high-frequency gate drives; their capacitance collapses under DC bias and temperature spikes, leading to catastrophic FET failure.

The Three MOSFET Capacitor Jobs (And Which Type Wins)

A MOSFET isn't just a switch; it's a voltage-controlled device burdened by internal parasitic capacitances ($C_{gs}$, $C_{gd}$, and $C_{ds}$). To control these parasitics and protect the silicon, we add external passive capacitors. Each job demands a specific dielectric material.

Decoding the Markings: What's Actually on the Bench?

If you pull a surface-mount capacitor from a tape reel or a salvaged board, you need to read the EIA code to know what you're holding. Let's break down a typical marking: 104K X7R 50V.

• 104 (Capacitance Code): The first two digits are the significant figures (10). The third digit is the multiplier in picofarads (4 = $10^4$). So, $10 \times 10,000 = 100,000$ pF, which is 100nF (or 0.1µF).
• K (Tolerance): J = ±5%, K = ±10%, M = ±20%.
• X7R (Dielectric): Tells you the temperature coefficient. X = -55°C, 7 = +125°C, R = ±15% capacitance change over that range.
• 50V (Voltage Rating): The maximum DC voltage the part can withstand before dielectric breakdown.

The DC Bias Trap: Here is where beginners get burned. A 10µF X7R ceramic in a tiny 0805 package rated for 25V will not give you 10µF when you apply 24V to it. Due to the physics of ferroelectric domains, Class II ceramics suffer severe DC bias derating. At rated voltage, that 10µF part might drop to 2µF. Always consult the manufacturer's DC bias curve, and physically oversize the package (e.g., use a 1206 or 1210 instead of an 0805) to maintain actual capacitance under load.

Worked Scenario: The 48V Motor Controller That Kept Blowing High-Side FETs

The Setup: A bench prototype for a 48V BLDC motor controller running at 40kHz PWM. The gate driver was a classic IR2110 half-bridge, driving IRLB3034 N-channel MOSFETs. The schematic called for a 100nF bootstrap capacitor.

The Numbers: The builder used a standard 0805 X7R 100nF (marked 104) rated at 50V. The IR2110 requires a minimum of ~10V above the source pin to fully enhance the high-side FET. The bootstrap diode was a fast-recovery UF4007.

The Outcome: Under no load, the controller worked perfectly. Under a 15A mechanical load, the high-side MOSFETs began to overheat and eventually failed short-circuit (drain-to-source) after about 10 minutes of operation.

What Went Wrong: Hooking up an oscilloscope to the high-side gate (referenced to its source) revealed the gate voltage was sagging to 7V during the on-time—well below the 10V+ needed to keep the FET in the low-$R_{DS(on)}$ ohmic region. The FET was spending too much time in the Miller plateau (linear region), dissipating massive heat.

The root cause was DC bias derating. The bootstrap capacitor sits between the switch node (which swings from 0V to 48V) and the high-side drive rail (which pumps up to ~60V). The 50V-rated 0805 X7R capacitor was seeing nearly 50V across its terminals. At that bias, its actual capacitance had collapsed from 100nF to roughly 18nF. It couldn't hold enough charge to keep the gate driven through the 25µs on-time.

The Fix: We swapped the 0805 104 for a 1µF 1206 X7R (marked 105) rated at 100V. The larger physical volume and higher voltage rating meant it retained ~850nF of actual capacitance at 50V bias. The gate drive held rock-solid at 12V, and the FETs ran cool.

Failure Modes and Visual Forensics

When a MOSFET circuit fails, the capacitor is often the victim, but sometimes it's the murderer. Here is how to read the physical evidence on the PCB.

1. Flex Cracking (The Silent Short)

Visual Symptom: A microscopic hairline fracture running diagonally from the PCB pad up into the ceramic body. Often invisible without a 10x loupe or microscope. Sometimes accompanied by a faint scorch mark near the terminal.

The Physics: Multilayer ceramic capacitors (MLCCs) are brittle. If the PCB flexes during depaneling, connector insertion, or thermal cycling, the board bends but the ceramic doesn't. The crack breaches the internal electrode layers, creating a dead short. Prevention: Use soft-termination MLCCs for large packages, or place caps away from board edges and mounting holes.

2. Dielectric Breakdown (The Explosion)

Visual Symptom: The capacitor is physically missing, leaving a black crater on the PCB, or a film capacitor has a blown-out hole in its epoxy coating with resin splattered nearby.

The Physics: Voltage transients (like inductive kickback from a motor) exceeded the capacitor's dielectric strength. In snubber circuits, if the dV/dt exceeds the component's rating, internal arcing vaporizes the metallization.

3. Piezoelectric 'Singing' (Acoustic Noise)

Visual Symptom: None. Audible Symptom: A high-pitched whine or buzz coming from the PCB at the PWM frequency (e.g., 16kHz to 20kHz).

The Physics: Class II ceramics (X7R, X5R) are piezoelectric. The AC ripple voltage physically deforms the dielectric, vibrating the PCB. It's not a failure, but it's annoying. Fix: Switch to C0G/NP0 ceramics or film capacitors for that specific node.

The Substitution Matrix: When the Exact Part is Missing

You're at the bench, the BOM calls for a specific part, and your drawer is empty. Here is the decision framework for safe substitution without redesigning the board.

By respecting the dielectric physics and reading the DC bias curves rather than just the nominal printed value, you ensure your MOSFETs switch cleanly, run cool, and survive the brutal realities of inductive loads.