Read the DC-bias curve before the datasheet front page

The number printed on an MLCC datasheet is measured at zero DC bias and room temperature. Your circuit applies neither. Murata shows the gap plainly: a 100 uF, 6.3 V, X5R capacitor biased at only 1.8 V keeps about 90 uF, while a Y5V part under the same conditions keeps only about 60 uF. This DC-bias loss is peculiar to high-dielectric-constant Class 2 ceramics built on barium titanate; temperature-compensating Class 1 parts such as C0G barely move. The buyer rule is simple: effective capacitance at your rail voltage must clear the circuit minimum with margin, and the number that counts comes from the maker DC-bias curve for the exact part number — not the nominal print. A common pitfall: a power board fails a ripple retest because its "10 uF" output capacitor measures barely 4 uF once the rail voltage is applied — the fix is usually a larger nominal value in a bigger case, not a better brand.

  • Always check the DC-bias curve for the exact part number before trusting the nominal capacitance.
  • Loss grows with applied voltage, so a capacitor run near its rated voltage loses the most.
  • If the effective value matters for timing, filtering or loop stability, start with C0G or derate aggressively.

Pick the dielectric for the temperature window

Class 2 MLCCs change capacitance irregularly with temperature, and the characteristic code tells you the guaranteed window. Murata explains that X7R and X5R are specified by maximum capacitance change over their temperature range — X7R is rated for −55 °C to +125 °C, X5R for −55 °C to +85 °C — while Class 1 parts such as C0G are specified by a near-linear temperature coefficient in ppm per degree. Pick X7R when the board sees wide temperature swings, X5R for cost-sensitive consumer goods at moderate temperatures, and C0G for resonant circuits, timing and RF matching where drift directly breaks the design.

  • X7R covers a wider temperature range than X5R; confirm the range against your worst-case enclosure temperature.
  • Class 1 (C0G/NP0) is the right choice when capacitance drift, not bulk value, is the failure mode.
  • Aging also eats Class 2 capacitance over time — leave margin in long-life products.

Derate the voltage, then budget the ripple current

Running a Class 2 MLCC near its rated voltage maximizes DC-bias loss and leaves no headroom for transients, so choose the rated voltage with real margin above the working rail. Ripple current matters too: ESR turns ripple into heat, and TDK states the heating directly — self-heating power equals ESR times the square of the current (P = ESR × I²) — so sustained self-heating on a hot board derates the part further and shortens its life. There is no industry standard for how allowable ripple current is measured. TDK’s FAQ is explicit about its own rule: TDK caps self-heating temperature rise at 20 °C above ambient for all of its MLCCs, calculating the allowable ripple current from the part’s ESR and thermal resistance for that 20 °C rise. Treat 20 °C of self-heating as the ceiling in your own thermal checks, and keep the MLCC’s total temperature — ambient plus self-heating — inside its rated range. Smaller cases lose a larger fraction of capacitance under bias and shed heat through less area, so upsizing the package is often cheaper than upsizing the nominal value.

  • Keep continuous working voltage comfortably below the rated voltage, especially on hot boards.
  • Verify self-heating temperature rise stays within about 20 °C of ambient on switching converter inputs and outputs.
  • Prefer a larger case size over a higher nominal value when DC-bias loss dominates.
Macro view of MLCC chip capacitors soldered on a printed circuit board

Design out flex cracks before they crack in the field

Electrical selection is only half the job; mechanical selection is the other half. KEMET’s flex-crack tech topic reports that, by Pareto analysis of customer complaints, the number-one failure mode of MLCCs is insulation-resistance failure caused by flex cracks — KEMET’s customer-complaint Pareto in that report — not manufacturing defects, which KEMET notes have been driven down to parts-per-billion levels. Flex cracks form when bending stress on the PCB is transmitted through the solder joint into the brittle ceramic body, which is far less elastic than the board. The usual sources are depanelization, connector assembly and use, and box-build stresses — all worst for capacitors placed near board edges or next to connectors. KEMET’s guidance for the mitigation: flexible-termination designs shift the flex stress into the termination area instead of the ceramic body (with up to 5 mm of flex-bend capability in KEMET’s FF-CAP example), while floating-electrode and open-mode constructions change the failure signature so a crack cannot complete a short-circuit path. For buyers, the rule is to keep large MLCCs off board edges and away from connector stress zones, and on 1206 and larger parts on power rails, to ask the supplier to confirm the termination type on the quotation.

  • Flex cracks are the leading MLCC field failure; keep large parts off board edges and away from connectors.
  • Depanelization and connector stresses are the common sources — route them away at layout time, not after qualification.
  • For 1206 and larger power-rail parts, prefer flexible-termination or open-mode designs and confirm the termination type on the quote.

Qualify second sources against curves, not part numbers

Two "identical" 10 uF 0805 parts from different makers can differ materially in effective capacitance at your rail voltage — the dielectric formulation, layer count and case geometry behind each maker’s DC-bias curve are not the same. So a second source is not qualified by matching capacitance, voltage, dielectric and case size on paper; it is qualified when its DC-bias curve, at your working voltage and temperature, clears the same effective-capacitance bar as the original, and the approval is recorded while the original is still the reference. This is also the discipline that pays off when allocation forces a substitution. When the exact ordering code cannot be had, our MLCC shortage 2026 buyer’s guide walks through the five-step substitution tree — discontinued versus allocated, stepping up a package size, renegotiating voltage rating, swapping X5R for X7R, paralleling smaller parts — each step re-qualified against the candidate’s datasheet. Run the same curve check there; a substitution approved from memory is how boards come back from the field.

  • Match the four numbers first — capacitance, voltage, dielectric, case size — then re-qualify against each candidate’s DC-bias curve.
  • Never let the effective capacitance at rail voltage fall below the circuit minimum on any approved alternate.
  • Record each approval with the curve and date; six months later nobody remembers which differences were accepted and why.

Selection snapshot: one workhorse value

The Sourceics catalog includes GRM21BR61E106KA73L, a Murata 10 uF, 25 V, X5R MLCC in 0805 — a workhorse value for decoupling and bulk capacitance on 12 V rails after derating. On 24 V systems, a 25 V rating leaves little transient headroom, so select a higher rated voltage or verify the effective capacitance against the exact DC-bias curve and the rail transient profile. Before ordering, confirm the full ordering code including the packaging suffix, and pull the DC-bias curve for this exact code from the maker so the effective capacitance at your rail voltage is on record.

  • 10 uF nominal in 0805: budget for bias loss and verify the effective value at the working voltage.
  • 25 V rating suits 12 V rails with margin; recheck for 24 V systems with transients.
  • Keep the same footprint qualified with a second maker for supply resilience.

Buy like the market is tight even when it isn’t

MLCC pricing and availability move with capacity cycles, so procurement discipline matters as much as engineering. Always RFQ the complete ordering code, state the acceptable date-code range, and ask the supplier to confirm the capacitance characteristic code on the reel label. When a substitute is proposed, demand its DC-bias curve — nominal values are not a qualification. Our guide to writing an RFQ that gets a useful answer covers the wording that prevents the common back-and-forth on tolerances and packaging. Ready to source? Send us an RFQ with the part number, manufacturer, package, date code, target price, quantity, and your project notes — we confirm the full ordering code, including the packaging suffix and the capacitance characteristic, against the maker datasheet before quoting.

  • RFQ the exact ordering code; a different suffix can mean different packaging or tolerance.
  • Require the characteristic code (X5R, X7R, C0G) on the quotation, not just capacitance and voltage.
  • For second sources, compare DC-bias curves, not nominal values.
  • Flag any reel older than your date-code policy before it ships.
Tape-and-reel component packaging on a warehouse shelf

Frequently asked questions

Quick answers to the MLCC questions buyers ask most.

  • Q: Why does my 10 uF MLCC measure only 4 uF in circuit? A: DC bias. Class 2 ceramics lose a large share of capacitance under working voltage, and smaller cases typically lose more — check the maker DC-bias curve for the exact part number and case size.
  • Q: X5R or X7R — which should I choose? A: X7R for wide temperature ranges such as automotive or outdoor gear (−55 °C to +125 °C); X5R is fine for benign indoor temperatures (−55 °C to +85 °C) and usually cheaper.
  • Q: How do flex cracks happen, and how do I avoid them? A: Board bending stress travels through the solder joint into the brittle ceramic body — KEMET identifies flex cracks as the top failure mode in its customer-complaint Pareto. Keep large parts off board edges and away from connectors, or specify flexible-termination or open-mode designs.
  • Q: Can I swap an equivalent MLCC from a different maker during a shortage? A: Only after its DC-bias curve clears your rail-voltage bar — nominally identical parts can differ materially in effective capacitance. Record the approval; our MLCC shortage 2026 buyer’s guide covers the full substitution sequence.

Manufacturer references

Use the latest guidance for the exact part and manufacturer when deciding acceptance criteria.