PowerElecHub

Bulk Capacitor Calculator for Rectifiers: Ripple Voltage, Hold-Up Time and Ripple Current

Size the smoothing capacitor after a diode bridge or half-wave rectifier. The page simulates the line cycle with a constant-power load and source resistance, then gives the valley voltage, ripple, capacitor RMS ripple current, peak and RMS line current, conduction angle, power factor, and the capacitance needed for a hold-up time.

Line and rectifier

R_s is everything the charging current flows through: line and wiring impedance, an NTC thermistor when hot, the diodes' dynamic resistance and the capacitor's ESR. It sets the peak current and cannot be zero here.

Load and capacitor

The load is the converter after the capacitor, modeled as constant power P_out/η. V_min is the lowest bus voltage at which that converter still regulates.

Valley voltage

91.1 V

peak 124.3 V

Ripple, peak to peak

33.1 V

27 % of the peak

Capacitor ripple current

1.25 A

RMS, all frequencies

Hold-up from the valley

3.5 ms

down to 70 V

  • Hold-up from the valley is 3.5 ms, short of the required 10 ms.

Line and diode currents

Peak charging current
3.94 A
Line current, RMS
1.42 A
Average current per diode
340 mA
Conduction time per pulse
2.94 ms (53°)
Power factor
0.61
Loss in R_s
3.03 W

Sizing

Power taken from C (P_out/η)
73.9 W
Average bus voltage
109.5 V
Hold-up if the line drops at the peak
10.7 ms
C for 10 ms from the valley
251 µF
Simple estimate of the valley
76.8 V

The simple estimate assumes the capacitor discharges for the whole half-cycle, V_pk² − 2P/(C·f_r). It ignores the conduction time and comes out lower than the simulation.

One line cycle in steady state

Current flows only while the rectified line is above the capacitor voltage. The narrower that window, the higher the peak and RMS current for the same power.

Capacitance sweep

Adding capacitance raises the valley but does not reduce the ripple current. The conduction window gets narrower, so the RMS current in the capacitor goes slightly up.

Time-domain simulation of the steady-state line cycle: ideal sinusoidal source, constant-drop diodes, one lumped series resistance, constant-power load. Line inductance, diode reverse recovery, capacitor tolerance and aging are not modeled. Allow for −20 % capacitance tolerance and end-of-life loss when choosing the part.

How it works

A capacitor-input rectifier charges the bulk capacitor only near the top of each half-cycle. For the rest of the time the capacitor alone supplies the load and its voltage sags. The lowest point of that sag, the valley voltage, is what the converter after it has to live with, and it sets both the transformer or inductor design and the input undervoltage threshold.

Most quick formulas pretend the capacitor discharges for the entire half-cycle and that the load is a constant current. Neither is right for an offline power supply. The converter after the capacitor draws roughly constant power, so its current rises as the bus voltage falls, and part of each half-cycle is spent charging rather than discharging. This calculator therefore integrates the circuit through the line cycle: a sinusoidal source, diodes with a fixed forward drop, one series resistance standing for the line, any thermistor, the diodes and the capacitor ESR, and a constant-power load. It repeats cycles until the valley voltage stops changing and reports that steady-state cycle.

The series resistance matters more than it looks. With very little resistance the charging pulse is short and tall, the peak current is several times the average, and the RMS current in the line and in the capacitor rises. That is why a capacitor-input rectifier has a power factor of only about 0.5 to 0.65 even with ideal parts. A few ohms of source resistance widen the pulse, lower the peak and slightly reduce the peak voltage.

The ripple current the capacitor must be rated for is the RMS of the charging pulses minus the load current, which this page computes directly from the simulated waveform. The waveform also has a component at the converter's switching frequency, which is not included here and must be added in quadrature, as on the capacitor ripple-current and lifetime calculator.

Hold-up time follows from energy alone. If the line disappears when the capacitor is at voltage V₀, a constant-power load can keep running until the voltage reaches V_min, which takes C·(V₀² − V_min²)/(2P). The worst case is a dropout right at the valley, so the page uses the valley voltage and searches for the capacitance that meets your hold-up requirement, taking into account that a bigger capacitor also raises the valley.

Design procedure

  1. 1Take the lowest line voltage and the lowest line frequency the product must work at; that combination gives the deepest valley.
  2. 2Enter the output power and the converter efficiency at low line, and the lowest bus voltage at which the converter still regulates.
  3. 3Estimate R_s from the thermistor, diodes, filter and ESR, then choose a capacitance whose valley stays comfortably above V_min.
  4. 4If a hold-up time is specified, check the capacitance the page reports for it and then add margin for −20 % tolerance and end-of-life loss.
  5. 5Choose a capacitor whose ripple-current rating covers the RMS value here, combined with the switching-frequency ripple, at the maximum ambient temperature.
  6. 6Use the peak and RMS line current to size the bridge, fuse and EMI filter, and check the high-line case for the voltage rating.

Formulas

Rectified peak
V_pk = √2·V_ac − n·V_F, n = 2 for a bridge, 1 for half-wave
Circuit that is simulated
C·dv/dt = i_in − P/v, i_in = max(0, (|v_s| − n·V_F − v)/R_s), P = P_out/η
Hold-up time
t_hold = C·(V₀² − V_min²)/(2P), V₀ = valley voltage for the worst case
Capacitance for a hold-up time
C ≥ 2P·t_hold/(V_valley² − V_min²), solved iteratively because V_valley depends on C
Simple valley estimate (pessimistic)
V_min ≈ √(V_pk² − 2P/(C·f_r)), f_r = 2·f_L (bridge) or f_L (half-wave)
Capacitor ripple current and power factor
I_C,rms = rms(i_in − P/v) over one cycle, PF = P_in/(V_ac·I_line,rms)

Worked example

The defaults are a 65 W universal-input adapter at its worst case, 90 V rms and 50 Hz, with an 88 % efficient converter, so the capacitor supplies 73.9 W. With 150 µF and 1.5 Ω of total series resistance the bus peaks at 124.3 V and falls to 91.1 V before the next half-cycle recharges it, a ripple of 33 V. Each charging pulse lasts 2.9 ms. The line current peaks at 3.9 A with an RMS value of 1.42 A, the power factor is 0.61, and the capacitor has to carry 1.25 A rms of ripple at twice the line frequency and its harmonics. The simple whole-half-cycle formula would have predicted a 76.8 V valley, about 14 V too pessimistic. If the converter stops regulating at 70 V, a line dropout at the valley leaves only 3.5 ms of hold-up; a dropout at the peak would give 10.7 ms. Meeting 10 ms from the valley needs about 251 µF. Doubling the capacitor to 300 µF raises the valley to 106.8 V but the ripple current goes up slightly to 1.32 A, because the charging window shrinks to 2.3 ms. At 230 V the same 150 µF gives a 308.6 V valley, 92 ms of hold-up and 0.70 A of ripple current, which is why the low-line case decides the capacitor. Lowering R_s to 0.3 Ω changes the valley by less than 0.5 V but raises the peak current to 4.6 A and lowers the power factor to 0.57. All of these numbers come from the calculator and were checked against a separate Python simulation.

Common mistakes

  • Designing at 230 V or 115 V nominal when the product must also start and run at 85 or 90 V; the low-line valley usually sets the capacitor.
  • Using a constant-current load model for a switching converter, which misses that the current rises in the valley.
  • Forgetting the capacitor tolerance and aging, so a design that just meets hold-up when new fails it after a few years.
  • Assuming more capacitance lowers the ripple current; it lowers ripple voltage but the RMS current goes up.
  • Rating the bridge rectifier only for average current while the peak charging current is several times higher.

Frequently asked questions

Why is the result different from C = I/(2·f·ΔV)?

That formula assumes a constant-current load that discharges the capacitor for the whole half-cycle. An offline converter draws roughly constant power, so its current increases as the voltage sags, and the capacitor is actually being charged for part of each half-cycle. For small ripple both effects are minor. For the 20 to 30 % ripple common in low-cost offline supplies, the simple formula underestimates the valley voltage by several volts to tens of volts, which leads to an oversized capacitor.

What value should I enter for R_s?

The sum of what the charging current sees: the source impedance of the line, an NTC inrush limiter in its hot state, the forward resistance of the two conducting diodes, any EMI filter series resistance and the capacitor's ESR. One to a few ohms is typical for small adapters; a stiff line with no thermistor can be well below 1 Ω. The valley voltage hardly depends on it, but the peak current, line RMS current and power factor do.

Why does the capacitor's ripple current not fall when I add capacitance?

Because the charge delivered to the load every half-cycle is fixed by the load, and a larger capacitor holds the voltage higher so the rectified sine catches up with it later. The same charge has to flow in a shorter pulse, which raises the RMS value. The valley voltage improves, but the capacitor's ripple-current rating has to grow with it.

Should hold-up be calculated from the peak or the valley?

From the valley if the requirement must hold no matter when the line drops out, which is how most hold-up specifications are tested. Some designs quote hold-up from the peak, which is longer; the page shows both. Also allow for the capacitor's tolerance, typically −20 %, and for capacitance loss at end of life and at low temperature.

Does this apply to a PFC front end?

No. With active power factor correction the bulk capacitor sits after a boost stage at a regulated voltage around 380 to 400 V, and its ripple is at twice the line frequency with a different shape. Use the PFC boost calculator for that case. This page is for the plain diode rectifier and capacitor found in adapters, auxiliary supplies and small motor drives.

How were the numbers checked?

The calculator integrates the circuit with a semi-implicit step of 4,000 points per line cycle. A separate Python program integrates the same equations with a fourth-order Runge–Kutta method at 100,000 points per cycle; valley voltage, peak and RMS currents and power factor agree within 0.3 %. For a very small series resistance the valley also matches a closed-form solution of the ideal rectifier, and the input power equals the load power plus the R_s and diode losses.

References

  1. Fairchild Semiconductor (now onsemi), application note AN-4137, "Design Guidelines for Off-line Flyback Converters Using Fairchild Power Switch" (bulk capacitor sizing from the minimum DC voltage and a charging duty ratio).
  2. N. Mohan, T. M. Undeland, W. P. Robbins, "Power Electronics: Converters, Applications, and Design", chapter on line-frequency diode rectifiers (single-phase bridge with a DC-side capacitor, effect of source inductance, power factor).
  3. The steady-state waveforms on this page were recomputed independently in Python with a fourth-order Runge–Kutta integration, and the ideal case against a closed-form solution.

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

Related calculators