Capacitor code
What the marking on a capacitor means — the three digits, the tolerance letter, the voltage prefix — plus the reactance and stored energy at your frequency and voltage, and a reminder of what X7R and C0G actually promise.
Three digits are picofarads with the last as the exponent, so 104 is 100 nF. A trailing letter is tolerance (K = ±10 %), a leading pair is voltage (2A = 100 V), and IEC forms like 4n7 or 0.1u are read directly.
A frequency to report the reactance at. Ideal 1/(2πfC) only — a real part turns inductive above its self-resonance, which the decoupling calculator covers.
A voltage to report the stored energy at, ½CV². Also a reminder that an MLCC loses much of its capacitance under DC bias, so the part will hold less than this suggests.
- Capacitance
- 100 nF (100,000 pF)
- Reactance at 1.00 kHz
- 1.59 kΩ
- Energy at 3.3 V
- 544 nJ
| Dielectric | Range | ΔC over range | Notes |
|---|---|---|---|
| C0G / NP0 | −55 to +125 °C | ±30 ppm/°C | Class 1. No DC-bias loss, no ageing, no piezo noise. Use for timing, filters, oscillators. |
| X5R | −55 to +85 °C | ±15 % | Class 2. Loses capacitance under DC bias; ages ~2.5 %/decade-hour. |
| X7R | −55 to +125 °C | ±15 % | Class 2. The default for decoupling. Same DC-bias caveats as X5R. |
| X7S | −55 to +125 °C | ±22 % | Class 2. Higher volumetric density than X7R, more bias loss. |
| Y5V | −30 to +85 °C | +22/−82 % | Class 2. Can lose most of its value hot or biased. Avoid in anything that matters. |
| Z5U | +10 to +85 °C | +22/−56 % | Class 2. Obsolete for new designs. |
What it computes
The marking on a capacitor is two significant digits and a multiplier, in picofarads. That is the whole scheme, and it is the reason this site is called 100nF: the most common capacitor on any board is printed 104.
104 → 10 × 10^4 pF = 100 000 pF = 100 nF
221 → 22 × 10^1 pF = 220 pF
479 → 47 × 0.1 pF = 4.7 pF multiplier 9 means ×0.1
158 → 15 × 0.01 pF = 0.15 pF multiplier 8 means ×0.01
47 → 47 pF one or two digits are pF directlyThe decoder also takes the unit-letter forms used on film and larger parts, letter in place of the decimal point: 4n7 is 4.7 nF, 4R7 is 4.7 pF, 0.1u and 100nF are 100 nF. A trailing letter is the IEC 60062 tolerance: B ±0.1 pF, C ±0.25 pF, D ±0.5 pF (absolute, for small class 1 parts), F ±1 %, G ±2 %, H ±3 %, J ±5 %, K ±10 %, M ±20 %, N ±30 %, P +100/−0 %, Z +80/−20 %. A leading digit-letter pair is the EIA-198 voltage: digit is the decade, letter the mantissa, so 1H is 5.0 × 10 = 50 V, 2A is 100 V, 1E is 25 V, 1C is 16 V, 0J is 6.3 V.
Going the other way, the tool gives the three-digit code for a value if one exists (two significant digits, 1 pF to 99 µF), plus reactance at a frequency, stored energy at a voltage, and the dielectric table.
Xc = 1 / (2π · f · C)
E = ½ · C · V²Class 1 ceramics (C0G/NP0) are paraelectric: ±30 ppm/°C, no ageing, no voltage dependence, no piezo effect, but low permittivity, so they top out near 100 nF. Class 2 (X5R, X7R, X7S, Y5V, Z5U) are ferroelectric barium titanate: ten to a hundred times the capacitance per volume, and every fault below follows from that material. The code is EIA-198: X = −55 °C, 5 = +85 °C, 7 = +125 °C, R = ±15 %, S = ±22 %, V = +22/−82 %.
Worked example
A tantalum or large ceramic marked 2A104K.
2A → A = 1.0, decade 2 → 1.0 × 10^2 = 100 V
104 → 10 × 10^4 pF → 100 nF
K → ±10 %
Xc at 1 kHz = 1 / (2π × 1000 × 100e-9)
= 1 / 6.283e-4 = 1.59 kΩ
E at 3.3 V = 0.5 × 100e-9 × 3.3²
= 0.5 × 100e-9 × 10.89 = 545 nJEncoding back: 100 nF is 100 000 pF, 10 with four zeros, so 104; 4.7 nF is 4700 pF, so 472. A 3-digit code cannot express 4.75 nF and the tool says so.
Where it stops being valid
DC bias. The marked value of a class 2 part is measured at 1 kHz, 1 V RMS, no DC. Under bias the dielectric saturates and the capacitance falls, faster in small packages where the field per volt is higher. A 10 µF 0603 X5R rated 6.3 V holds around 40 % at 5 V; the same value in 0805 rated 10 V holds around 70 %. Do not guess: Murata SimSurfing and KEMET K-SIM publish per-part-number curves, and parts with the same label differ between vendors. The tool reports the number on the body; the number in the circuit is that times the bias factor.
Ageing. Class 2 ceramics lose capacitance logarithmically after their last trip above the Curie point, 1 to 2.5 % per decade-hour for X7R, more for X5R and Y5V. The marking is referenced to 1000 hours and reflow resets the clock, so a board read the day after assembly is higher than the same board a year later.
Piezoelectric noise. The same material makes a class 2 MLCC an actuator. Ripple makes it sing through the board; worse, flex or vibration produces a voltage across a coupling cap in an audio or sensor path. C0G, film or tantalum for signal chains that get shaken.
Tantalum. Moulded tantalums carry the value in µF and the voltage in digits, or the EIA code this tool decodes. The bar marks the positive end, opposite to an electrolytic. Reverse polarity or a spike ignites MnO2 tantalums, so derate to 50 % of rated voltage, 30 % on a battery input. Polymer tantalums fail short without burning and tolerate 80 to 90 %.
Film. Polyester and polypropylene use the unit-letter form (2n2, 100n, 0.1u) plus the voltage in digits. No bias effect, low loss, self-healing, which is why mains X and Y parts are film. Polyester drifts about +5 % over 0 to 85 °C; polypropylene does not.
Unmarked MLCCs. Nothing smaller than 1206 is marked. Value, dielectric and voltage exist on the reel label only; a loose 0402 is a 0402 of unknown value. Keep the reel, or put it on an LCR meter.
Common mistakes
- Reading 104 as 104 pF. Three digits are never literal; two are.
- Taking the marked 10 µF at face value on a 5 V rail. Check the DC-bias curve for that exact part number; the 0603 may be 4 µF in the circuit.
- Reading the tantalum bar as the negative end.
- Substituting Y5V or Z5U for X7R because the package matches. Y5V at 85 °C and rated voltage can be a tenth of its marked value.
- Confusing the tolerance letter with the voltage code. 104K is ±10 %; the voltage is the digit-letter pair in front. M after the digits is ±20 %, not µ.
- Using a class 2 cap for timing or a filter corner: ±15 % tempco, minus bias loss, minus ageing.
Further reading
- Murata SimSurfing — DC-bias, temperature and impedance curves per part number; the number that replaces the marking.
- KEMET K-SIM — the same for KEMET ceramics and tantalums, including ageing.
- KEMET, Introduction to Capacitor Technologies — class 1 vs class 2, tantalum failure modes and derating.
- IEC 60062:2016, Marking codes for resistors and capacitors — the source of the tolerance letters and the fixed-length capacitance codes.