Capacitor Code Calculator — Numeric, SMD & Colour

Why three digits and picofarads

A resistor has room on its body for four coloured rings. A 3 mm ceramic disc has room for three characters, and that is the constraint the entire code exists to satisfy. Two significant figures and a power of ten, read in picofarads:

104  →  10 × 10⁴ pF  =  100 000 pF

The trouble is that nobody works in picofarads. The schematic says 100 nF, the supplier lists 0.1 µF, the part is marked 104, and all three are the same capacitor. That is three units and a factor of a million between them, which is why this page reports every value in all three at once rather than making you pick.

The 8 and 9 exception

A third digit of 8 or 9 does not mean ×10⁸ or ×10⁹. Those would be values no ceramic capacitor reaches. Instead:

• 8 means ×0.01
• 9 means ×0.1

They exist so that values below 10 pF can still be written in three digits — 479 is 4.7 pF, not 47 gigafarads. The calculator says which rule it applied, so a misread announces itself.

R for the decimal point

Small values are also written with an R where the decimal point belongs: 4R7 is 4.7 pF, 2R2 is 2.2 pF. Exactly the same convention as 4k7 on a resistor, and for the same reason — a full stop is the first thing to vanish from a smudged silkscreen or a faxed drawing. A marking of one or two digits with no multiplier at all is simply read as picofarads.

Tolerance letters

A trailing letter gives the tolerance, and the codes are settled: F is ±1 %, G ±2 %, J ±5 %, K ±10 %, M ±20 %. K on the end of 104 means ±10 %, not kilo — a genuine trap, and one that turns 100 nF into 100 000 nF in somebody's notes.

B, C and D are absolute figures in picofarads rather than percentages: ±0.1 pF, ±0.25 pF, ±0.5 pF. They only appear on very small values, where a percentage of a few picofarads would be meaningless. Z is the odd one: +80 % / −20 %, which is not sloppy manufacturing but an honest admission about the high-permittivity ceramics used for bulk decoupling.

What the code does not tell you

The value, and nothing else. Two capacitors both marked 104 can behave completely differently:

• Voltage rating. Not in the code at all. A 104 may be a 16 V part or a 1 kV part, and fitting the first where the second belongs fails immediately.
• Dielectric. C0G/NP0 holds its value across temperature and voltage. X7R drifts with both. Y5V can lose most of its capacitance at rated voltage and temperature — the same 104 marking, a quarter of the capacitance in circuit. For timing or filtering this is the specification that matters, and it is printed as a separate code or not at all.
• DC bias derating. A modern MLCC can lose half its capacitance simply by having its rated voltage across it. Nothing on the body warns you.

On the colour bands

Older ceramic, mica and tantalum parts carry colour bands instead, and the digit and multiplier colours are the same ones the resistor code uses — which is the reason that code is worth memorising at all. This decoder reads those two roles and stops there.

It stops there deliberately. Capacitor tolerance and voltage bands were assigned differently by different manufacturers in different decades, and there is no single authority to appeal to. Charts on the internet present one table as definitive; they disagree with each other. Read the tolerance and voltage from the datasheet or the part number, and treat a confident-looking voltage band chart as a guess someone has typeset.

For the resistor version of the same code — where the tolerance and temperature-coefficient bands are standardised, in IEC 60062 — see the resistor colour code calculator.

Voltage rating, and the capacitance you quietly lose

The code gives you a capacitance and, sometimes, a tolerance. It says nothing about working voltage, and on a multilayer ceramic that omission costs more than it appears to.

Class II ceramics — the X7R, X5R and Y5V parts that make up most of the small decoupling capacitors in circulation — lose capacitance under DC bias. Not slightly. A 0603 X7R marked 104 and rated 16 V can be down to half its marked value at 10 V applied, and a Y5V part can lose 80 per cent. The capacitor is not faulty and a meter will measure it correctly at zero volts. It simply is not a 100 nF capacitor in the circuit you put it in.

The mitigation is to buy voltage headroom you do not need for breakdown: a part rated at two to three times the working voltage keeps most of its capacitance, and a physically larger package at the same rating loses less than a small one. This is the reason a rail that measures a clean 3.3 V on the bench develops ripple in production after a package change that looked like a pure cost saving.

Why two capacitors marked 104 behave differently

The three-letter dielectric code is doing more work than the capacitance code. C0G, also written NP0, holds its value to within 0.3 per cent across its whole temperature range and does not age. X7R shifts by up to 15 per cent from minus 55 to plus 125 degrees. Y5V is permitted to fall by 82 per cent at its limits, which is a specification most people would read as a failure if they met it in the field.

Class II parts also age. Capacitance falls logarithmically with time after the part is soldered — typically 1 to 3 per cent per decade hour, so a capacitor is measurably smaller a year in than it was on the reel. Heating it above its Curie point resets the clock, which is why a measurement taken shortly after reflow reads high.

For a decoupling capacitor none of this matters much. For anything that sets a frequency, a time constant or a filter corner, it decides whether the circuit works — and it is why oscillator and timing positions specify C0G by name rather than a value alone.