Buck & Boost Converters · Guide · glidepathscience.org
GuideBuck & Boost Converters
Contents
This Guide covers the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round. Read Part 1 before the Intro track and Part 2 before the Advanced track, about ten minutes each. Come back whenever a question sends you here.
Part 1 · The essentials
The big idea
A switching power supply turns one DC voltage into another by chopping it, not by burning the difference. A switch connects the input for a fraction D of each cycle, many thousands of times a second; a coil and a capacitor smooth the result; a diode keeps the coil’s current flowing while the switch is off. Three ideas explain it all: a coil keeps its current flowing ($V = L\,dI/dt$), energy is conserved, and in steady state the coil’s volt-seconds balance.
A coil and its kick
A coil (an inductor) makes a voltage that opposes any change in its current:
$V = L\,dI/dt$
L, in henries, is its self-inductance. Close a switch on a coil in series with a battery and, at the first instant, no current flows yet, so the whole emf is across the coil and the current starts to rise at $dI/dt = ε/L$. As it grows, more of the emf drops across the wire’s resistance R, and the current levels off at ε/R with the time constant $τ = L/R$.
Now open the switch. A coil cannot stop its current at once: that would take an infinite voltage. Instead its voltage leaps to whatever keeps the current flowing for a moment. With nothing else to flow through, the current charges the switch’s tiny capacitance until the air in the gap breaks down: a spark, the inductor’s kick. A transistor switch is driven into breakdown (avalanche) the same way.
Give the current a path: the energy in a coil
A diode across the coil, pointing so the battery holds it reversed while the switch is closed, gives the current a path when the switch opens: the coil’s voltage reverses, the diode conducts, and the current circles through coil and diode until it dies away. This “flyback” diode protects every switched relay coil and motor. The energy the coil hands over is what it stored in its magnetic field:
$U = ½LI^2$
Without the diode it goes into the spark; a converter hands it to a load instead, every cycle.
The buck: a fast switch and a filter
In a buck the switch connects the input to the switch node for the on-time DT of each period T, and a diode from ground carries the coil’s current during the off-time. The switch node is a square wave between $V_{in}$ and about 0 V; the coil and capacitor, a low-pass filter, pass its average with a small ripple.
The coil’s current is a triangle: during the on-time it has $V_{in} - V_{out}$ across it and ramps up; during the off-time, $-V_{out}$, and ramps down. In steady state it ends each cycle where it began, so its average voltage is zero: its volt-seconds balance.
$(V_{in} - V_{out})D = V_{out}(1 - D)$ $V_{out} = D\,V_{in}$
This holds whatever the load, as long as the current never reaches zero (continuous conduction).
The boost: the kick added to the source
Rearrange the parts: the coil from the input to the switch node, the switch from there to ground, and the diode on to the output. While the switch is on, the source ramps the coil’s current up with $V_{in}$ across it. While it is off, the coil’s kick adds to the source and pushes its current through the diode into the output, with $V_{out} - V_{in}$ across it the other way. Balancing the volt-seconds,
$V_{in}D = (V_{out} - V_{in})(1 - D)$ $V_{out} = V_{in}/(1 - D)$
In theory the output rises without limit as D approaches 1; real parts’ losses cap it. With the switch off the diode still joins input to output, so a boost cannot cut its output off: it sits near $V_{in}$ at $D = 0$.
The energy books
A converter cannot make energy. The input pays for the load, the losses and any change in the energy U stored in the coil and capacitor:
$P_{in} = P_{out} + P_{loss} + dU/dt$ $E_{in} = E_{out} + E_{loss} + ΔU$
In steady state U ends each cycle where it began (at start-up most of the input is being stored), so averaged over whole cycles $P_{in} = P_{out} + P_{loss}$, the efficiency is $η = P_{out}/P_{in}$, and a converter that raises the voltage must draw more current than it delivers: $I_{in} = P_{out}/(ηV_{in})$. The engine’s books count every joule and balance to rounding.
Linear against switching
A linear regulator is a transistor in series with the load, turned partly on so that it drops exactly the difference. It carries the whole load current with $V_{in} - V_{out}$ across it, so
heat $= (V_{in} - V_{out})I + V_{in}I_q$
where $I_q$ is the regulator’s own small current; its efficiency can never beat $V_{out}/V_{in}$. A switch is never half on: closed, it has only $IR_{on}$ across it; open, it carries nothing. So a switching supply runs at 85–95 % near its rated load, less at light load. A linear regulator wins where its heat is small: it is simpler, cheaper and quiet. It can only lower a voltage, with some headroom (its dropout).
Part 2 · Going further
Ripple: the coil and the capacitor
From the on-time ramp, the coil’s peak-to-peak ripple is
$ΔI_L = (V_{in} - V_{out})D/(Lf)$
(with ideal parts the off-time ramp, $V_{out}(1 - D)/(Lf)$, gives the same in steady state; a diode’s drop adds $V_f$ to the off-time voltage). The capacitor takes the ripple current. It charges for half of each triangle, which carries $ΔI_L T/8$, so with negligible ESR
$ΔV = ΔI_L/(8Cf)$
The ripple current also flows through the capacitor’s series resistance (ESR): when that dominates, $ΔV ≈ ΔI_L R_{ESR}$, triangular rather than curved. Hence low-ESR ceramics.
Discontinuous conduction
The ripple does not depend on the load. Below half the ripple the triangle’s valley would fall below zero; a diode cannot carry current backward, so the current stops at zero until the switch turns on: discontinuous conduction mode (DCM). The boundary, where the valley just touches zero, is at
$I_{crit} = ΔI_L/2$
in the coil’s mean current: a buck’s load current, but a boost’s load is then only $(1 - D)ΔI_L/2$. A buck stays in continuous conduction while $K = 2L/(RT)$ exceeds $1 - D$. In DCM the volt-second balance has a third, idle part, and the output depends on the load:
$V_{out}/V_{in} = 2/(1 + \sqrt{1 + 4K/D^2})$
A synchronous rectifier in forced PWM lets the current go negative instead, so it never enters DCM.
Losses and efficiency
Each part loses its share while it conducts: the main switch ($P_{sw}$) for the fraction D of each cycle, the diode ($P_D$) or a synchronous switch ($P_{sr}$) for 1 − D. The diode drops $V_f$, the switches and coil lose $I^2R$, the capacitor $I_{rms}^2R_{ESR}$, and each transition about ½VI times the switching time, plus the gate drive. In continuous conduction, with $I_L$ the coil’s mean current ($I_{out}$ in a buck, $I_{out}/(1 - D)$ in a boost):
${P_{sw} ≈ I_L^2R_{on}D}$ ${P_D ≈ V_f I_L(1 - D)}$ ${P_{sr} ≈ I_L^2R_{sr}(1 - D)}$
At a low output voltage a diode’s fixed drop is a large share; a synchronous rectifier, a second switch driven opposite the first, replaces it with a small resistance. Efficiency against load rises, peaks and falls: fixed losses dominate at light load, $I^2R$ losses at heavy load (the 24 V to 5.0 V buck of the ripple example peaks near 1 A, at 90 %).
Designing to a specification
A design starts with the frequency. Each ramp lasts about 1/f, so for a given ripple L and C shrink as f rises ($ΔI_L ∝ 1/(Lf)$, $ΔV ∝ 1/(Cf)$), but every transition costs energy, so switching loss grows with f. Choose L for a ripple of about 20–40 % of the load,
$L = (V_{in} - V_{out})D/(ΔI_L f)$
then a part whose saturation current is above the peak, $I_{out} + ΔI_L/2$, with margin: above $I_{sat}$ the inductance collapses and the peak shoots up. Then C for the ripple, the switch and rectifier for their ratings and losses, and the price and area.
Closing the loop
A real supply measures its output every cycle and adjusts D. Proportional control changes D in proportion to the error, so holding a new duty (a heavier load needs a slightly larger D) leaves a lasting error, a droop. An integrator keeps adjusting while any error remains and removes it:
$D = D_{ff} + K_p e + K_i Σ e\,T$
with $e = V_{set} - V_{out}$ and a feed-forward $D_{ff} = V_{set}/V_{in}$ for input changes. Too much gain makes the loop ring or oscillate. In a boost, raising D first shortens the time the diode feeds the output, so the output dips before it rises (the right-half-plane zero): boost loops are kept slow.
The inverting buck-boost, and beyond
With the coil to ground and the diode from the output, the switch charges the coil from the source; when it opens, the coil pulls its current up out of the output, so the output goes negative. Balancing $V_{in}DT$ against $-V_{out}(1 - D)T$,
$V_{out} = -D\,V_{in}/(1 - D)$
stepping down below D = 0.5 and up above it; its switch and diode see $V_{in} + |V_{out}|$. Mains supplies add isolation: a flyback stores the energy in a transformer and releases it from the other winding, so the two sides share no wire. This lab stays with low-voltage DC.
Reference
Common mistakes
- “The coil’s current stops when the switch opens.” It cannot stop at once; its voltage leaps to keep it flowing.
- Losing the ½ in $½LI^2$, or forgetting to square the current.
- Using $V_{in}$ for the coil’s on-time voltage. In a buck it is $V_{in} - V_{out}$.
- Mixing up the buck and the boost. $D\,V_{in}$ lowers; $V_{in}/(1 - D)$ raises.
- “A boost makes energy.” It draws more current than it delivers.
- Linear heat as the output power. It is $(V_{in} - V_{out})I$.
- Losing the 8 in $ΔI_L/(8Cf)$, or forgetting the ESR.
- Using $D\,V_{in}$ in DCM. There the output depends on the load.
- The diode conducting all the time. Its loss carries $(1 - D)$.
- Dropping the buck-boost’s minus sign.
Equation sheet
| Part 1 | |
|---|---|
| a coil | $V = L\,dI/dt$ |
| at switch-on | $dI/dt = ε/L$; $τ = L/R$ |
| stored | $U = ½LI^2$ |
| buck | $V_{out} = D\,V_{in}$ |
| boost | $V_{out} = V_{in}/(1 - D)$ |
| the books | $P_{in} = P_{out} + P_{loss} + dU/dt$ |
| steady state | $P_{in} = P_{out} + P_{loss}$; $I_{in} = P_{out}/(ηV_{in})$ |
| linear | heat $= (V_{in} - V_{out})I + V_{in}I_q$ |
| Part 2 | |
|---|---|
| coil ripple (buck) | $ΔI_L = (V_{in} - V_{out})D/(Lf)$ |
| output ripple | $ΔV = ΔI_L/(8Cf)$ |
| ESR ripple | $ΔV ≈ ΔI_L R_{ESR}$ |
| boundary | $I_{crit} = ΔI_L/2$ (coil mean) |
| DCM buck | $V_{out}/V_{in} = 2/(1 + \sqrt{1 + 4K/D^2})$, $K = 2L/(RT)$ |
| losses (CCM) | ${P_{sw} ≈ I_L^2R_{on}D}$; ${P_D ≈ V_f I_L(1 - D)}$; ${P_{sr} ≈ I_L^2R_{sr}(1 - D)}$ |
| buck-boost | $V_{out} = -D\,V_{in}/(1 - D)$ |
Symbols: L inductance (H) · I current (A) · ε, $V_{in}$ input voltage (V) · $V_{out}$ output (V) · D duty cycle · f switching frequency (Hz), T = 1/f · C capacitance (F) · $R_{ESR}$ the capacitor’s series resistance · $ΔI_L$, ΔV peak-to-peak ripples · η efficiency · $V_f$ a diode’s drop · $R_{on}$, $R_{sr}$ the main and synchronous switches’ resistances · $I_L$ the coil’s mean current · $I_q$ a regulator’s own current · K = 2L/(RT) · $K_p$, $K_i$ the loop’s gains.
Further reading
Samuel J. Ling, Jeff Sanny and William Moebs, University Physics Volume 2 (OpenStax, 2016), free online under CC BY-NC-SA 4.0: Chapter 14, Inductance (self-inductance, energy in a magnetic field and RL circuits), which the Intro track’s first steps follow. Robert W. Erickson and Dragan Maksimović, Fundamentals of Power Electronics, 3rd ed. (Springer, 2020), for volt-second balance, discontinuous conduction and control. And this lab’s Field Manual (H): how the model works, and what it leaves out.
Buck & Boost Converters v1.4 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org