A field return came back from a DC-DC converter prototype with a note: “output ripple high, suspect capacitor.” The board used a 22 µF, 25 V, X5R MLCC in the output filter. The part passed incoming inspection, and a handheld LCR meter measured 21.6 µF at 1 kHz. Replacing it with another 22 µF MLCC from the same reel produced the same excessive ripple. The root cause was not a defective part but a measurement condition that omitted the converter’s actual operating voltage.
Why DC Bias Reduces Capacitance in Class 2 Dielectrics
Class 2 dielectrics such as X5R and X7R rely on ferroelectric barium titanate grains. Under a DC electric field, the domain walls become partially pinned, so the permittivity drops. The effect is nonlinear and becomes pronounced above roughly 30–50% of the rated voltage. A 22 µF, 25 V X5R MLCC in a 1210 or 1206 case can measure 12–14 µF at 12 V DC bias, and in smaller cases the same nominal value can drop below 8 µF. This is not a manufacturing defect; it is a material property documented in most manufacturer datasheets as a “DC bias characteristic” curve.
The practical consequence is that a replacement part selected solely by nominal capacitance and voltage rating may deliver far less effective capacitance in circuit. For output filtering, that shift directly raises ripple voltage and can push the converter outside its control loop stability margin. For bulk decoupling, the reduced charge storage may cause supply droop under transient loads.
Layered Root-Cause Checks for Suspected DC Bias Issues
When a circuit behaves as if the capacitor is missing, work through these checks in order.
- Confirm the nominal part: Verify marking, case size, capacitance, voltage rating, and dielectric code against the BOM. A 25 V vs. 16 V rating changes the bias curve for the same nominal capacitance.
- Check the dielectric: X7R retains a higher percentage of capacitance under bias than X5R at the same field strength. If the design originally specified X7R and an X5R substitute was used, the effective capacitance difference can exceed 20% at the same DC voltage.
- Measure without bias first: Use an LCR meter at 1 kHz and 0.5 Vrms or lower to confirm the part is not open, shorted, or below tolerance. This step isolates gross failures from bias-related loss.
- Measure with bias: Use a DC bias fixture or a component analyzer with an internal bias source. Apply the exact DC voltage present across the part in the intended circuit, not the rated voltage.
- Evaluate at operating temperature: If the board runs hot, measure at that temperature. X5R capacitance also drifts with temperature, and the combined derating can be larger than either effect alone.
Measuring Effective Capacitance Under DC Bias
A standard LCR meter without bias will read the small-signal capacitance at zero volts. To capture the real value, you need a bias-capable instrument or a bias tee. The measurement setup should include the following parameters.
| Parameter | Recommended Setting | Notes |
|---|---|---|
| Measurement frequency | 1 kHz | Matches most MLCC datasheet conditions; some curves use 120 Hz or 10 kHz. Use the same frequency as the datasheet curve you compare against. |
| AC test level | 0.5 Vrms or lower | Higher AC levels can induce additional capacitance change in Class 2 dielectrics. |
| DC bias level | From 0 V to 80% of rated voltage, in 20% steps | Plot the curve and record the value at the circuit’s operating point. |
| Acceptance threshold | Effective capacitance at operating voltage must be at least 70% of the design target | Typical design margin for output filters; some power designs require 80% or higher for stability. |
| Temperature | 25 °C nominal, plus the expected operating temperature | For X5R, capacitance can change by ±15% over −55 °C to +85 °C; the bias effect adds on top. |
If you do not have a bias fixture, a simpler check is to place the MLCC in the actual circuit, power the board, and measure the ripple voltage or transient response. Compare against the design target. This functional test does not give a capacitance number, but it confirms whether the part is adequate in that specific application.
Prevention Checklist for Sourcing and Replacement
DC bias capacitance loss should be addressed at the selection stage, not discovered during debug. Use this checklist when sourcing substitutes or qualifying alternative brands.
- Request the DC bias curve: Ask the supplier for the capacitance vs. DC voltage plot at the relevant temperature. If the datasheet does not include one, treat the part as unverified for bias-sensitive designs.
- Derate by voltage, not by multiplier: For a 12 V rail, a 25 V rated X5R is a common choice, but verify the effective capacitance at 12 V. If it falls below the design target, move to a higher voltage rating, a larger case size, or an X7R dielectric.
- Check case size effects: For the same nominal capacitance and voltage, a larger case generally retains more capacitance under bias because the dielectric layer is thicker. A 1210 part may outperform a 0805 part even when both are rated 25 V.
- Compare across brands: Bias behavior varies between manufacturers and even between production lots. When qualifying a second source, measure the effective capacitance of samples at the operating voltage, not just the nominal value.
- Document the acceptance value: Define the minimum effective capacitance at the operating point in the incoming inspection specification. This prevents good parts from being rejected and ineffective parts from being accepted.
For designs that cannot tolerate capacitance loss, C0G/NP0 dielectrics show virtually no DC bias effect, but they are limited to lower capacitance values. When higher capacitance is required, consider a parallel arrangement of multiple smaller MLCCs to reduce the voltage stress per part and improve the overall retained capacitance.
DC bias capacitance loss is a predictable, measurable characteristic. Including it in the sourcing and qualification process turns a common field failure into a preventable design consideration.