AKKN ElectronicsA quality engineer pulls five samples from a received reel of 10 µF, 0805, X7R MLCCs and measures them with the incoming-inspection LCR meter. All five read between 3.2 µF and 4.5 µF, far below the marked value. The lot is quarantined as suspect. Before the consignment is returned, one detail should be checked: the meter applies no DC bias. These parts may be fully functional; the test condition simply does not match the application condition.
The same pattern appears in field service. A board shows ripple or startup timing that does not match the design. Every Class II MLCC on the rail is measured at 1 kHz with a handheld meter, and several read 30% to 50% below nominal. The components are tagged as failed and replacements are ordered. In most of these cases, the diagnosis is a test-condition error rather than a component defect.
Symptom: Capacitance Below the Marked Value
Marked capacitance for a Class II MLCC is defined at a specific reference condition: 25 °C, 1 kHz, 1 Vrms, and 0 V DC offset. Real operating conditions are different. A decoupling capacitor on a 12 V rail is biased at 12 V DC. The capacitance that matters for circuit behavior is the incremental capacitance at that bias point, which is the relevant ΔQ/ΔV at the operating voltage.
X5R, X7R and X8R dielectrics are ferroelectric materials. A DC field reorients the domains in the barium titanate structure, and the incremental permittivity falls as the field increases. The effect is not a defect; removing the DC bias restores the original measurement. Because of this, the same capacitor can legitimately show two very different values on the same day.
Layered Root-Cause Checks
Before rejecting parts or ordering replacements, run these checks in order.
1. Measurement setup. Does the LCR meter have a DC bias function or an external bias tee? If not, the reading is the zero-bias capacitance, which is expected to be higher than the in-circuit value. For a true field reading, measure with a bias source set to the rail voltage.
2. Bias severity and voltage rating. Derating depends on the electric field across the dielectric, which is the applied voltage divided by the dielectric thickness. For the same nominal capacitance and case size, a lower voltage rating means a thinner dielectric, and therefore a deeper capacitance drop at a given operating voltage. Manufacturer-published bias curves for typical 0805 parts show that a 10 µF part rated at 6.3 V may retain roughly 50% to 70% of nominal at 5 V DC, while a 25 V-rated part typically remains above 80% of nominal at 12 V DC. The exact curve is formulation-dependent and must be taken from the specific manufacturer's datasheet.
3. Actual component defects. Verify these before closing the case:
- Case flex cracks: low capacitance can be accompanied by high DF or low insulation resistance. Flex the board gently near the capacitor while monitoring IR.
- Solder defects: voids or insufficient wetting raise the measured impedance. Rework the joints and re-measure at 0 V; if the value returns to within ±20% of nominal, the issue was solder-related.
- Wrong dielectric in the BOM: C0G has virtually no bias dependence, while X5R and X7R do. Confirm the dielectric code on the part and against the BOM.
- Age-related drift: Class II capacitors lose capacitance at a rate of roughly 2% to 5% per decade of hours after the last reflow, and recover when reflowed again above the Curie point. A stable, slightly low reading on a stored board is normal.
Measurement Method and Acceptance Values
Measure the effective capacitance under DC bias using the procedure below.
- Set the LCR meter to 1 kHz and 1 Vrms, matching the standard MLCC reference condition.
- Record the capacitance at 0 V DC as the reference value.
- Apply a DC bias equal to the measured rail voltage in the application.
- Record the effective capacitance and dissipation factor under bias.
- Apply the capacitor's rated voltage and measure insulation resistance after a 60-second charge.
For switching-regulator input and output stages, repeat the measurement at the converter's switching frequency, or at 100 kHz if the datasheet specifies it. Acceptance is based on the circuit requirement, not on the marked value alone. Use the following thresholds as starting points for a 10 µF class, 0805, X7R part:
| Parameter | Condition | Acceptance threshold |
|---|---|---|
| Capacitance, 0 V reference | 1 kHz, 1 Vrms, 0 V DC | Within ±20% of marked value |
| Effective capacitance under bias | 1 kHz, 1 Vrms, rail DC voltage | At or above the design minimum, or within ±20% of the manufacturer's bias curve at that voltage |
| Dissipation factor | 1 kHz, 1 Vrms, 0 V DC | ≤10% for X7R, ≤15% for X5R |
| Insulation resistance | Rated voltage, 60 s charge | Time constant ≥1000 Ω·F for X7R (about 100 MΩ for 10 µF) |
As an example, if the design minimum effective capacitance on a 12 V rail is 4 µF, a part that measures 5 µF at 12 V bias passes. A part measuring 2 µF fails, and the appropriate fix is a different component specification, not a replacement of the same part number. Selecting a higher voltage rating, such as 25 V for a 12 V rail, or a dielectric with a flatter bias curve, such as X8R, is a more effective correction.
Prevention Checklist
- Size the nominal capacitance so that the effective capacitance at the operating bias is at least 1.5 to 2 times the calculated minimum for the required transient response.
- Use a voltage rating with margin: a 25 V-rated part on a 12 V rail operates on a flatter portion of the bias curve than a 16 V-rated part.
- When substituting or sourcing cross-references, compare the capacitance-versus-DC-voltage curves of the original and candidate parts. Identical case size, nominal value, and dielectric code do not guarantee identical bias behavior.
- Include a DC-bias measurement in incoming inspection for capacitors used on rails above 3.3 V, and state the bias voltage, frequency, and AC level in every inspection report.
- Request bias-curve data during datasheet review and verify it with sample measurements before committing a new part to production.
