The nominal capacitance of a capacitor printed on the datasheet is a best-case scenario, not a physical guarantee. On the bench, a 100µF ceramic might act like a 20µF part under DC bias, and an aluminum electrolytic might drop to 80µF at -20°C. If you treat the printed number as an absolute truth, your circuit will eventually fail. Think of a capacitor like a water balloon: the more you stretch it (apply DC voltage), the stiffer the rubber gets, and the less additional water (charge) it can hold for a given increase in pressure. This is DC bias derating, and it is just one of the reasons why selecting the right component requires looking past the front-page specs.

Decoding the Markings: What the Code Actually Means

When you pull a ceramic or film capacitor from a bin, the physical real estate is too small for a full spec sheet. Instead, manufacturers use standardized EIA (Electronic Industries Alliance) codes. Here is how to read the cryptic three-digit numbers and letters stamped on the casing.

The 3-Digit EIA Capacitance Code

The first two digits are the significant figures. The third digit is the multiplier (number of zeros to add), and the base unit is always picofarads (pF).

  • 104: 10 × 104 pF = 100,000 pF = 100 nF (0.1 µF)
  • 473: 47 × 103 pF = 47,000 pF = 47 nF
  • 226: 22 × 106 pF = 22,000,000 pF = 22 µF

Following the numeric code, you will typically find a letter indicating tolerance, and sometimes a secondary code for voltage or temperature coefficient.

  • J = ±5% tolerance
  • K = ±10% tolerance (Most common for X7R/X5R ceramics)
  • M = ±20% tolerance (Standard for electrolytics and high-value MLCCs)

If you see a letter-number-letter code like X7R or C0G (also known as NP0), this is the temperature coefficient. C0G means the capacitance changes by 0 ±30 ppm/°C (virtually zero drift). X7R means it will not vary more than ±15% across a temperature range of -55°C to +125°C. Always check SparkFun's capacitor guide if you encounter obscure military or industrial date codes on vintage NOS (New Old Stock) parts.

Dielectric Showdown: Which Type for Which Job

You cannot simply swap a 10µF electrolytic for a 10µF ceramic and expect the same behavior. The dielectric material dictates the equivalent series resistance (ESR), leakage current, and microphonic noise. Use this matrix to select the right chemistry for your specific application.

Chemistry / Dielectric Construction Tolerance & Tempco Typical Use Case DC Bias & Edge Cases
Ceramic (C0G/NP0) Class I paraelectric ceramic ±1% to ±5%, 0 drift RF filters, PLL loops, precision timing Zero DC bias derating. Low max capacitance (usually <10nF).
Ceramic (X7R/X5R) Class II ferroelectric ceramic ±10% to ±20%, ±15% drift Decoupling, buck/boost converters, general bypass Severe DC bias derating. Piezoelectric (sings/microphonic).
Aluminum Electrolytic Etched foil with liquid electrolyte ±20%, high temp drift Bulk storage, audio coupling, low-freq filtering High ESR, polarized. Dries out over time (limited lifespan).
Polymer (Solid) Foil with conductive polymer cathode ±20%, stable CPU Vcore decoupling, high-ripple power supplies Extremely low ESR. No dry-out mechanism. Polarized.
Film (Polypropylene) Metallized plastic film wound/stacked ±1% to ±5%, very stable Snubbers, motor run, high-voltage AC/DC Self-healing on dielectric breakdown. Bulky physical footprint.

Bench War Story: When 100µF Isn't Actually 100µF

Let us look at a real-world scenario where ignoring the physics of capacitance led to a failed prototype.

The Setup: I was designing a 5V, 3A synchronous buck converter for a custom carrier board. The datasheet for the regulator recommended 100µF of output capacitance to keep the transient response tight and maintain phase margin. To save board space, I selected a single 1206-size X5R MLCC (Multi-Layer Ceramic Capacitor) rated at 100µF, 6.3V.

The Numbers: On paper, 100µF > 100µF. The BOM cost was $0.18 per unit. I ordered the boards.

The Outcome: When I powered up the prototype, the switching node was ringing violently, and the output ripple was a massive 140mV peak-to-peak (spec was <30mV). Under a 2A step load, the 5V rail dipped to 4.1V, resetting the downstream microcontroller.

What Went Wrong: I fell victim to MLCC DC bias derating. Ferroelectric ceramics (like X5R and X7R) lose their permittivity as the internal electric field increases. I plugged the exact part number into the Murata SimSurfing tool. The graph showed that at my 5V DC operating point, the actual capacitance of a capacitor rated for 100µF had plummeted by 75% to just 25µF. My control loop was starving for capacitance.

The Fix: I replaced the single 1206 part with two 220µF, 10V 1210 MLCCs in parallel. Because they were rated for 10V, the 5V bias only derated them by about 40% each. Two 130µF effective capacitors in parallel gave me 260µF of real-world capacitance, dropping the ripple to 18mV and stabilizing the loop. Alternatively, a 150µF solid polymer capacitor would have provided the exact 150µF regardless of DC bias, albeit with a slightly larger footprint.

Autopsy of a Failed Cap: Visual Symptoms and Failure Modes

When a capacitor fails, it rarely does so silently. Knowing the visual and electrical symptoms of failure modes will save you hours of troubleshooting with an oscilloscope.

Safety Warning: Never short-circuit a large electrolytic or film capacitor to discharge it. The instantaneous current can weld your screwdriver, vaporize PCB traces, and cause severe burns. Always discharge through a high-wattage power resistor (e.g., 100Ω 10W) on an insulated stick.
  • Aluminum Electrolytic (Venting/Bulging): Visual: The scored cross on the top of the can is popped open, or the rubber bung at the bottom is pushed out. You will see a brown, crusty residue on the PCB. Cause: Overvoltage, reverse polarity, or excessive ripple current boils the liquid electrolyte, generating hydrogen gas. Electrical symptom: Capacitance drops, ESR spikes to tens of ohms.
  • Tantalum (Thermal Runaway): Visual: A charred, blackened crater on the PCB. The component is completely missing or fused into a lump of slag. Cause: Tantalum caps use a manganese dioxide (MnO2) cathode, which contains oxygen. If a voltage spike causes a microscopic short in the dielectric, the heat releases oxygen from the cathode, feeding a literal fire. Rule: Never use tantalum on rails with high inrush currents or unclamped inductive spikes.
  • MLCC Ceramic (Flex Cracking): Visual: Invisible to the naked eye. Requires cross-sectioning or X-ray. Cause: Board flexure during depanelization, screw tightening, or connector insertion bends the PCB. Because the ceramic is brittle, it cracks near the solder joint. Electrical symptom: Intermittent short circuits that appear only when the board is mechanically stressed, or a dead short that blows the upstream fuse.

The Substitution Matrix: Swapping Parts Safely

Supply chain shortages happen. When the exact BOM part has a 40-week lead time, you need to substitute safely. Use this decision path to avoid introducing new bugs into a proven design.

  1. Voltage Rating (Always Go Up): You can always substitute a 50V cap for a 25V cap. Never go down. Remember to factor in AC ripple voltage on top of the DC bias when calculating the peak voltage the dielectric will see.
  2. Temperature Coefficient (Upgrade Only): Swapping an X5R (±15% over -55 to +85°C) for an X7R (±15% over -55 to +125°C) is perfectly safe. Swapping an X7R for a C0G is even better for stability, provided you can find the value in C0G. Never downgrade the tempco class.
  3. Capacitance Value (Proceed with Caution): Going from 10µF to 22µF is usually fine for bulk decoupling. However, in active filter networks (like Sallen-Key topologies) or timing circuits (like a 555 astable), changing the value alters the cutoff frequency or duty cycle. Furthermore, in linear regulators (LDOs), drastically increasing the output capacitance can cause massive inrush currents that trip the internal thermal shutdown during hot-plug events.
  4. ESR Matching (The LDO Trap): This is where most hobbyists and junior engineers fail. Older LDOs (like the classic LM1117) rely on the high ESR of an aluminum electrolytic capacitor to create a zero in the feedback loop that maintains phase stability. If you substitute it with a low-ESR ceramic or polymer cap, the zero disappears, the phase margin collapses, and the LDO will oscillate at high frequencies, destroying your output rail. If you must use low-ESR ceramics on an older LDO, you must add a 1Ω series resistor to artificially simulate the ESR of an electrolytic. For modern LDOs specifically designed for ceramic output caps (often noted as "low-ESR stable" in the datasheet), this is not required. Always consult the component selection guidelines before swapping chemistries.

Ultimately, mastering the capacitance of a capacitor means looking past the marketing numbers on the reel. By understanding DC bias derating, reading EIA codes fluently, and respecting the failure modes of different dielectrics, you will build power supplies and signal chains that survive the transition from the prototyping bench to the real world.