Why Do ICs Need Decoupling Capacitors? Design, Placement, and Selection

A familiar recommendation is “put 100nF near each power pin.” That is a useful starting point for many devices, but not a universal design rule. Some pins need a different value, a separate network, or capacitors that satisfy a regulator’s stability requirements.

How Does a Decoupling Capacitor Supply Transient Current?

Think of the power distribution network, or PDN, as the complete route from the regulator to the silicon and back. It includes connectors, copper planes, traces, vias, package connections, and capacitance. None of those connections is ideal.

When internal logic switches, outputs charge loads, or an ADC starts a conversion, the IC can need a brief burst of current. Resistance causes a voltage drop. Inductance opposes a rapid change in current. The resulting disturbance appears between the IC’s own power and ground terminals—even if a meter at the regulator looks steady.

1. BEFORE THE EVENTThe capacitor is chargedThe supply establishes the rail voltage and stores charge in local capacitance.
2. DURING THE EDGEA short loop supplies currentThe capacitor provides part of the transient current close to the IC.
3. AFTER THE EVENTThe network replenishes chargeThe regulator and upstream network restore charge as the load and control loop respond.

Current must complete a loop. Describing decoupling as “sending noise into ground” misses half the problem: ground is a return conductor with impedance, not a place where current disappears. A long return path can undermine a capacitor placed close to VDD.

A battery does not remove this need. Batteries, connectors, planes, traces, and package connections all have impedance, while an IC can still demand fast current pulses locally.

Murata’s digital-IC supply simulations show why wiring inductance produces supply ripple and why adding local capacitance changes that response.

How Do Decoupling, Bypass, and Coupling Capacitors Differ?

On a supply rail, decoupling capacitor and bypass capacitor often refer to the same component. “Decoupling” emphasizes reducing interaction through a shared supply; “bypass” emphasizes providing a lower-impedance path for high-frequency current. A coupling capacitor is different: it is commonly placed in series with a signal to block DC while passing AC.

Is a 100nF Decoupling Capacitor Always Enough?

No. A 100nF capacitor is not a certificate of correct power integrity. The required network depends on the device, package, rail, operating mode, and PCB implementation. These two manufacturer examples show why a copied generic circuit is not enough.

0.1µF and 100nF are the same nominal capacitance. That unit conversion does not establish voltage rating, tolerance, dielectric, effective capacitance under bias, or installed impedance.

Sources: TI ADS111x datasheet, §10.2; ST AN4488, §2.2. Check the current datasheet for your orderable device and any family-specific exceptions.

The practical review sequence is pin requirements → reference layout → capacitor characteristics → board validation. A first-order calculation helps explain the design, but does not overrule a documented minimum or a required stability network.

How Do You Estimate the Required Effective Capacitance?

For a simplified rectangular current pulse, estimate how much charge the capacitor must provide before the upstream network contributes the missing current:

C_effective ≥ ΔI × Δt / ΔV_charge

Here, ΔI is the current deficit supplied by the capacitor, Δt is its duration, and ΔV_charge is the voltage-drop budget allocated to charge depletion. For a changing current waveform, use the integral of that current deficit instead of multiplying one current by one duration.

A longer event changes the answer sharply. In the second illustrative example, 80mA for 20µs with a 100mV charge budget requires 16µF effective. A regulator may supply part of a real burst, reducing the charge deficit; an upstream current limit may make it worse.

If 50mV is the entire rail tolerance, allocating all 50mV to capacitor discharge is optimistic. Leave room for DC accuracy, resistive loss, inductive effects, ripple, and other disturbances. This calculator is not a power-integrity or regulator-stability sign-off.

How Do ESR, ESL, and Resonance Change the Result?

A real capacitor includes equivalent series resistance (ESR) and equivalent series inductance (ESL). The pads, traces, vias, and return path add more impedance. Increasing nominal capacitance cannot remove all of those effects.

Do not add these example peaks to the charge-drop result and call it a measured waveform. Their timing depends on how the real network shares current. The examples isolate mechanisms; they do not simulate the complete PDN.

What Happens at Self-Resonance?

Below its series-resonant frequency, a simple capacitor model has falling impedance as frequency rises. Near resonance, ESR limits the minimum impedance. Above resonance, inductance dominates. Mounting inductance changes the installed behavior, so a component-only curve is not the whole board response. See Analog Devices MT-101.

Several capacitor values in parallel can help across frequency, but they can also create antiresonance: a frequency band where the network impedance rises. Murata’s installation and parallel-capacitor examples illustrate this risk. A 10µF + 100nF + 1nF stack is not automatically better than the network recommended by the IC vendor.

For a more demanding PDN, ΔV/ΔI provides a starting impedance target. A 50mV allowance for a 100mA step corresponds to 0.5Ω. The useful target must cover the relevant frequency range and account for the actual current spectrum; that single ratio does not establish stability.

How Much Capacitance Remains Under DC Bias?

Many high-capacitance ceramic parts use Class II dielectrics such as X5R or X7R. Their capacitance can decrease when DC voltage is applied. Temperature, tolerance, and aging also matter.

X7R describes a temperature characteristic; it does not guarantee a particular DC-bias curve. C0G behaves differently, but is not a drop-in solution for every high-capacitance requirement. Murata explains these distinctions in its DC-bias FAQ and dielectric comparison.

For a purchase decision, request the candidate’s capacitance-versus-bias data at the intended rail voltage. Then include tolerance, operating temperature, and aging in the review. Do not apply a fixed “50% derating” to every MLCC, or assume a higher voltage rating guarantees more effective capacitance.

If the curve is missing, the alternative is not yet electrically demonstrated. That is a request for more evidence—not proof the part is bad, and not permission to approve it by appearance.

Where Should Decoupling Capacitors Go on the PCB?

Place the local capacitor so that the power connection and ground return form a short, low-inductance loop with the IC. “Close to the pin” is shorthand for that electrical objective, not a universal millimeter limit.

1. Follow the recommended pin-to-capacitor arrangement.Place local parts before less sensitive components. Route short, direct connections rather than long narrow branches.
2. Give the return an equally short route.Use the intended ground plane and appropriate nearby vias. Avoid forcing transient current across a plane split or through a long shared return.
3. Review vias as part of the loop.Layer transitions add inductance, but vias are not automatically forbidden. The stackup, via arrangement, and package escape routing matter.
4. Check every rail and package option.A BGA may benefit from capacitors beneath the package with a suitable via structure. A leaded package may use adjacent top-side parts. Copy the vendor’s relevant layout, not an unrelated board photograph.
5. Keep verification points accessible.Plan a safe way to probe local VDD relative to local ground. A convenient test point far upstream may miss the disturbance that reaches the IC.

For device-specific context, compare the PCB against ADS111x §10.2–11 or the applicable IC’s own layout guidance.

How Do Local, Bulk, Reference, and Regulator Capacitors Differ?

A large bulk capacitor near a connector does not automatically replace local parts at an IC. Conversely, scattering small ceramics around a board does not guarantee enough stored charge for a long load burst.

When Can a Larger Capacitor Destabilize the Regulator?

An output-capacitor change can alter a regulator’s control-loop behavior. The allowed capacitance and ESR depend on the regulator; “lower ESR is always better” is not a universal rule. Review the approved operating range, startup load, and load-step response. Analog Devices discusses this interaction in its LDO capacitor-selection guidance.

When Does a Ferrite Bead Help—and When Can It Resonate?

A bead and capacitor can form a useful supply filter, but they can also resonate. Check the bead’s impedance across frequency and under actual DC current, its DC resistance, and the need for damping. A catalog impedance specified at 100MHz and zero bias does not prove filtering at your operating point. ADI AN-1368 explains those tradeoffs.

Similarly, do not transfer an external-reference circuit from one ADC to another. Supply decoupling, reference-drive requirements, and internal-regulator capacitors are different design questions.

How Do You Measure and Troubleshoot Decoupling Performance?

Measure voltage between the IC’s local supply and local ground during the relevant operating event. A handheld meter can miss a short dip, and a quiet regulator output does not prove a quiet IC supply.

How Can the Probe Create False Ringing?

A long probe ground lead adds inductance and can create ringing or pick up noise. On suitable low-voltage circuits, use a compatible ground spring or another short, low-inductance connection, with adequate probe and oscilloscope bandwidth.

Tektronix’s ABCs of Probes demonstrates how ground-lead length changes observed waveforms. Record the probing method with the result.

Probe only within the instrument’s voltage and common-mode ratings. A standard oscilloscope ground clip is commonly earth-referenced. Never lift protective earth to measure a floating or hazardous node; use an appropriately rated differential or isolated measurement system and qualified procedures.

What Must Buyers Verify Before Approving a Replacement Capacitor?

A line item reading “100nF, 16V, 0603” is useful for sorting candidates. It is incomplete for demonstrating equivalence. Two candidate parts may differ in their capacitance under bias, impedance, termination construction, or qualification status.

1. Identify the exact approved part.Record manufacturer, full MPN, dielectric, tolerance, dimensions, termination, and any application-specific qualification requirements.
2. Record where and how it operates.Include the IC, rail voltage, temperature range, capacitor role, and relevant schematic location. A VCAP or regulator-output part deserves its own review.
3. Compare effective capacitance and impedance.Request the candidate’s DC-bias data and relevant frequency information. A matching nominal value or higher voltage rating alone is insufficient.
4. Keep provenance separate from electrical approval.Packaging, labels, traceability records, and agreed inspection address sourcing questions. They do not replace design validation.
5. Obtain engineering approval before substitution.Document the accepted alternative and any required board tests. Do not silently consolidate several local capacitors into one “equivalent total” value.

For a shortage-driven purchase, separate two questions: Can the required component be sourced with the agreed documentation? and Has the proposed alternative been approved for this circuit? Availability does not answer the second question.

Related YURUNOX information: quality assurance and the purchasing process.