PowerElecHub

Output Capacitor Calculator for Load Transients: Buck Converter Load Step

Estimate the output capacitance and ESR a buck converter needs to keep a load step within a voltage limit. Compares the loop-bandwidth limit ΔI/(2π·f_c·C) with the inductor slew limits for load application and release, adds the ESR step and ESL spike, and shows how much a faster loop can save.

Converter

Load step

Output capacitor bank

Enter the whole bank: parallel ceramics divide ESR and ESL. Use the capacitance left at the output voltage after DC-bias loss, not the nominal value.

Capacitance needed

765 µF

set by the loop bandwidth

Maximum bank ESR

3.6 mΩ

ΔV/ΔI, with no capacitance budget left

Estimated undershoot

34.9 mV

with 800 µF

Estimated overshoot

34.9 mV

with 800 µF

Capacitance by each limit

Loop bandwidth ΔI/(2π·f_c·ΔV)
765 µF
Inductor slew, load application
83.7 µF
Inductor slew, load release
753 µF
Budget after the ESR step
26.0 mV

Deviation parts with C entered

ESR step ΔI·ESR
10.0 mV
Loop term ΔI/(2π·f_c·C)
24.9 mV
Slew term, application
2.7 mV
Slew term, release
24.5 mV
ESL spike ESL·di/dt
1.0 mV
Inductor slew time up / down
435 ns / 3.92 µs

Capacitance needed against loop crossover

Raising the crossover lowers the loop limit in proportion, until it meets the inductor slew limit. Past that point a faster loop no longer saves capacitance; only a smaller inductance does.

Buck and buck-derived converters. The loop term is a first-order estimate; in a simulated current-mode loop the real peak was 70 to 99 % of it for phase margins from 30 to 85°. The slew terms assume an instantaneous load step and a duty cycle that saturates immediately, the best any controller can do. Boost-type converters with a right-half-plane zero are not covered.

How it works

When the load current of a buck converter steps, the inductor current cannot follow instantly, so the output capacitor supplies or absorbs the difference until the control loop catches up. How far the output moves depends on three things: how fast the loop responds, how fast the inductor current is physically able to change, and the capacitor's own ESR and ESL. The calculator treats each one separately and takes the worst.

The loop term comes from output impedance. A loop with crossover frequency f_c holds the output impedance near the capacitor's impedance at f_c, 1/(2π·f_c·C), which gives the familiar estimate ΔV ≈ ΔI/(2π·f_c·C). It is an approximation, so we checked it: for an idealized current-mode loop with phase margins between 30 and 85°, simulated with a load step, the actual peak deviation came out at 70 to 99 % of this estimate. It is slightly pessimistic, which is the right direction for a sizing rule.

The slew terms do not depend on the controller at all. Even if the duty cycle jumps to its limit at once, the inductor current can only rise at (V_in − V_out)/L and fall at V_out/L. The charge the capacitor has to supply or absorb during that ramp is L·ΔI²/(2·V_L), so the deviation is that charge divided by C. For a low output voltage the release case is the hard one, because only V_out is available to ramp the current down. A 1.2 V rail from 12 V takes nine times longer to shed current than to pick it up.

Finally, the ESR produces an immediate step ΔI·ESR and the ESL a spike ESL·di/dt during the load edge. More capacitance does not reduce either; more capacitors in parallel and a tighter layout do. The calculator subtracts the ESR step from the allowed deviation before sizing the capacitance, so if ESR alone uses the budget, no capacitance is enough. The chart shows the capacitance needed for every crossover from 1 kHz to 1 MHz, which tells you whether faster compensation or more capacitors is the cheaper fix.

Design procedure

  1. 1Define the worst load step, its slew rate, and the allowed deviation, including the regulator's static tolerance in the overall budget.
  2. 2Enter the converter's V_in, V_out, inductance and planned crossover frequency.
  3. 3Check which limit sets the capacitance. If the loop limit dominates, faster compensation helps; if a slew limit dominates, only a smaller inductance or more capacitance does.
  4. 4Choose a bank whose effective capacitance at V_out meets the requirement and whose total ESR leaves enough of the budget.
  5. 5Place capacitors close to the load to keep ESL low, then confirm with a load-step measurement at the real slew rate.

Formulas

Loop-bandwidth estimate
ΔV_loop ≈ ΔI/(2π·f_c·C)
Inductor slew, load application
ΔV_up = L·ΔI²/(2·C·(V_in − V_out))
Inductor slew, load release
ΔV_down = L·ΔI²/(2·C·V_out)
ESR and ESL
ΔV_ESR = ΔI·ESR, ΔV_ESL = ESL·di/dt
Capacitance needed
C ≥ max(ΔI/(2π·f_c·ΔV′), L·ΔI²/(2·ΔV′·(V_in − V_out)), L·ΔI²/(2·ΔV′·V_out)), ΔV′ = ΔV − ΔI·ESR

Worked example

The defaults describe a 12 V to 1.2 V point-of-load buck with a 0.47 µH inductor and an 80 kHz crossover, facing a 10 A load step at 10 A/µs with a ±36 mV (3 %) limit. A ceramic bank with 1 mΩ total ESR takes 10 mV of that as a step, leaving 26 mV. The loop limit then asks for 765 µF, the load-release limit 753 µF and the load-application limit only 84 µF, because 10.8 V ramps the inductor up in 0.44 µs while 1.2 V needs 3.9 µs to ramp it down. With 800 µF of effective capacitance the estimated undershoot and overshoot are both about 35 mV. The ESL of 0.1 nH adds a 1 mV spike on the edge. Raising the crossover to 200 kHz cuts the loop limit to 306 µF, but the release limit stays at 753 µF, so the saving is lost; the two limits meet at about 81 kHz. The maximum bank ESR for this step is 3.6 mΩ, at which point no capacitance would be enough. The slew-limited values were checked by integrating the inductor ramp and capacitor charge numerically in Python.

Common mistakes

  • Using the nominal capacitance of ceramics that lose much of it under DC bias.
  • Checking only load application when the output voltage is low and release sets the overshoot.
  • Raising the crossover frequency to save capacitance after the inductor slew limit already dominates.
  • Forgetting the ESR step, which no amount of capacitance removes.
  • Measuring a load step with a slow electronic load, which hides the ESL spike and understates the deviation.

Frequently asked questions

Why is load release often worse than load application at low output voltages?

When the load drops, the controller turns the high-side switch off and the inductor current can only fall at V_out/L. When the load rises, it rises at (V_in − V_out)/L. For a 1.2 V output from 12 V the release ramp is nine times slower, so the capacitor has to absorb nine times more charge. That is why point-of-load converters for processors are often sized by overshoot, and why some controllers add nonlinear responses aimed at load release.

How accurate is ΔV ≈ ΔI/(2π·f_c·C)?

It is a first-order estimate. We simulated an idealized current-mode loop, a PI controller with an extra pole and the output capacitor, and adjusted the phase margin from 30 to 85°. The actual peak deviation was 70 to 99 % of the estimate, closer to it at low phase margin. A voltage-mode loop with a poorly damped LC resonance can ring more, so treat the estimate as a guide and confirm with a load-step measurement.

Should I enter the nominal capacitance?

No, the effective capacitance at the output voltage and temperature. Small ceramic capacitors can lose a large fraction of their capacitance under DC bias, even at a few volts for low-voltage parts. Polymer and electrolytic capacitors hold their capacitance but have higher ESR, which the ESR input covers.

Does this apply to boost converters?

Not directly. A boost converter has a right-half-plane zero that limits the crossover frequency and makes the initial response go the wrong way, and the inductor current only reaches the output during the off-time. The capacitor-sizing ideas carry over, but the formulas here are for buck and buck-derived converters.

How were the numbers checked?

A separate Python program integrates the inductor ramp and capacitor charge after an instantaneous load step to get the slew-limited deviations, which match the closed forms. The loop estimate was compared with time-domain simulations of an idealized current-mode loop over a range of phase margins, as described above.

References

  1. R. W. Erickson and D. Maksimović, "Fundamentals of Power Electronics", chapters on closed-loop converter design (output impedance under feedback and its relation to crossover frequency).
  2. Infineon (International Rectifier), application note AN-1162, "Compensator Design Procedure for Buck Converter with Voltage-Mode Error-Amplifier" (crossover chosen at 1/10 to 1/5 of the switching frequency and its effect on load-transient speed; ESR zero).
  3. The slew-limited deviations were recomputed by numerical integration in Python, and the loop estimate was compared with simulated load steps of an idealized current-mode loop.

Results are engineering estimates. Verify your design with datasheets, simulation and measurement.

Related calculators