Buck-Boost Converter

Why buck and boost aren't enough

Consider a lithium-ion-powered device that needs a regulated 3.3 V rail. A fresh cell sits at 4.2 V, and a depleted cell sags to 3.0 V. The supply voltage swings across the rail in both directions during the discharge cycle. A simple buck (step-down only) gives up at 3.3 V because it cannot pull above its input minus a dropout volt or two. A simple boost (step-up only) gives up at 4.2 V because it cannot push below its input. Neither can hold 3.3 V across the full battery range.

A buck-boost is the topology that can. It can step the input voltage up or down, depending on the chosen duty cycle, with the same hardware. The classic version inverts polarity as a side effect of the topology — useful when you specifically want a negative rail; an obstacle when you just want a positive rail and have to add level-shifting circuitry. Variants like the SEPIC and the modern 4-switch buck-boost solve the polarity problem at the cost of more components.

The four-component inverting buck-boost

         SW
   V_in ──/──┬───────────┐
              │          ▲ D
              UUUU      / \
              UUUU L  (diode pointing up)
              UUUU      \ /
              │          ▼ cathode
              │          │
              ├──────────┤
              │          │
             GND      V_out ── load
                                │
                              ── C
                                │
                               GND

   Switch closed: V_in → L → GND. Current builds in L.
   Switch open:   L → D → V_out (negative w.r.t. GND).
                  The inductor's "GND-side" terminal becomes
                  positive; its "V_in-side" becomes negative
                  relative to ground after the switch opens.

   V_out = −V_in × D / (1 − D)

The topology uses one switch (to V_in), one inductor (between the switch node and ground), one diode (from the switch node out to V_out, blocking until needed), and one output capacitor.

The two phases

Switch closed (on-time, fraction D). The MOSFET pulls the inductor's V_in side up to V_in. Current ramps up through L at dI/dt = V_in / L, storing energy in the magnetic field. The diode is reverse-biased (its cathode sits at the lower-voltage output side); no current flows to the load. The capacitor C sustains the load alone during this phase.

Switch open (off-time, fraction 1−D). The MOSFET turns off. Inductor current cannot stop instantaneously, so it forces a new path. It pulls its V_in side down, which forward-biases the diode and lets current flow through it to the output capacitor — but with the inductor's original V_in side now negative w.r.t. ground. Polarity inverted. Energy stored in the inductor's field now drives current through D into the load, while charging C up to the new (negative) output voltage.

Where V_out = −V_in·D/(1−D) comes from

Steady state requires zero average voltage across the inductor. The inductor sees V_in during on-time and V_out during off-time (note: V_out is negative). Equating:

V_L_avg = V_in × D + V_out × (1 − D) = 0
       → V_out = −V_in × D / (1 − D)

D = 0.25 → V_out = −V_in × 0.25/0.75 = −V_in/3
D = 0.50 → V_out = −V_in × 0.50/0.50 = −V_in
D = 0.75 → V_out = −V_in × 0.75/0.25 = −3·V_in
D = 0.90 → V_out = −V_in × 0.90/0.10 = −9·V_in

The duty cycle continuously spans bucking (|V_out| < V_in) at D < 0.5 and boosting (|V_out| > V_in) at D > 0.5. At D = 0.5 the magnitude crosses through V_in exactly.

Worked example: −12 V from 5 V USB

Generate a −12 V analog supply for an op-amp from a USB 5 V rail, at I_load = 50 mA.

Duty cycle. −12 = −5 · D/(1−D) → D = 12/17 = 0.706. Switch on 70.6% of each period.

Switching frequency. f_sw = 500 kHz → T = 2 μs, t_on = 1.41 μs.

Inductor. Choose ΔI_L = 30% of I_in_avg. Since I_in_avg = I_load · |V_out|/V_in / efficiency ≈ 50 · 12/5 / 0.85 ≈ 141 mA, ΔI_L = 42 mA. L = V_in · t_on / ΔI_L = 5 · 1.41 μs / 0.042 = 168 μH. Round up to 220 μH shielded inductor rated for at least I_peak = I_in_avg + ΔI_L/2 ≈ 162 mA — pick 500 mA or 1 A part for margin.

Output capacitor. 22 μF tantalum or low-ESR electrolytic, voltage-rated for at least 16 V (1.3× |V_out|). Pay attention to polarity — the cap's negative terminal connects to ground; its positive terminal sits at the V_out node (which is negative w.r.t. ground in inverting topology). Some designers use bipolar ceramics to avoid the polarity confusion.

Switch and diode. Both must withstand V_in + |V_out| = 17 V. Pick 30 V-rated low-side MOSFET and 30 V Schottky. The MOSFET is referenced to ground at source, gate-driven from V_in side — a low-side N-channel works directly. The Schottky drops 0.4 V at 50 mA, contributing about 3% to inefficiency.

Efficiency. Expect 80–87% from this topology — about five points below a buck or boost at the same power. Acceptable for a low-power analog supply; the alternative (charge pump) has worse regulation and ripple.

Buck-boost family

Topology	Polarity	Transfer	Pros	Cons
Inverting buck-boost	Reversed	V_out = −V_in·D/(1−D)	Single switch, simple, cheap	Negative output complicates load referencing
SEPIC	Same as V_in	V_out = V_in·D/(1−D)	Shared ground, non-inverting	Two inductors (or coupled), more parts
Ćuk	Reversed	V_out = −V_in·D/(1−D)	Continuous I_in and I_out (low ripple)	Inverting, AC-coupled output node
4-switch non-inverting	Same as V_in	V_out = V_in·D_buck/(1−D_boost)	Best efficiency 95%+, smooth V_out polarity	4 MOSFETs, complex control IC
Flyback (isolated)	Isolated, either polarity	V_out = V_in·N·D/(1−D)	Isolation, multiple outputs from one transformer	Transformer adds cost/size; snubber required
Charge-pump inverter	Reversed	V_out ≈ −V_in	No inductor; tiny	Limited current (mA), no regulation under load

The modern four-switch non-inverting buck-boost

Putting one inductor between a buck half-bridge and a boost half-bridge gives a circuit that can operate as: (a) pure buck when V_in > V_out (only the buck switches operate, boost switches park as a wire), (b) pure boost when V_in < V_out (only the boost switches operate, buck switches park as a wire), or (c) full 4-switch buck-boost in the transition region near V_in ≈ V_out. The output stays positive, efficiency stays in the 92–96% range, and the control IC handles the mode-shifting transparently.

      V_in ──[HS1]──┬──UUUU──┬──[HS2]── V_out
                     │        │
                    [LS1]   [LS2]
                     │        │
                    GND      GND

   Buck mode: HS1+LS1 PWM; HS2 fully on, LS2 fully off
   Boost mode: HS2+LS2 PWM; HS1 fully on, LS1 fully off
   Buck-boost mode: all four switch in coordinated phases

Modern USB Power Delivery chargers and power banks use exactly this topology to deliver any negotiated voltage (5 V, 9 V, 12 V, 15 V, 20 V, 28 V) from a battery that swings 2.7–4.2 V per cell. The LTC3119, LTC3789, TPS63810, and TPS61253 are all variations on the 4-switch buck-boost theme.

Control loop and the right-half-plane zero

Like all boost-derived converters, the inverting buck-boost has a right-half-plane (RHP) zero in its small-signal control-to-output transfer function. Physically: increasing duty cycle to demand more output transiently decreases output (because the increased on-time steals time away from the energy-delivery off-time), before the new equilibrium pushes it up. This non-minimum-phase behaviour limits achievable loop bandwidth to a fraction (typically 1/5 to 1/10) of the RHP zero frequency, which sits at:

f_RHPZ ≈ V_out² × (1−D) / (2π × L × P_load × D)

For V_in = 5 V, V_out = −12 V, D = 0.71, L = 220 μH, P_load = 0.6 W:
   f_RHPZ ≈ 144 × 0.29 / (2π × 220e−6 × 0.6 × 0.71)
          ≈ 7 kHz

   Maximum useful loop bandwidth ≈ 700 Hz – 1.5 kHz.

This is why buck-boost and boost converters have slower load-step response than buck converters — they're physically slower to respond to demand changes. Compensation networks must accommodate the RHP zero, typically with low loop bandwidth and aggressive low-pass filtering. Engineers sometimes use feed-forward of input voltage to improve transient performance, since input-voltage changes don't suffer the RHP zero limitation.

Where buck-boost converters live

• Battery-powered portables. Lithium cell 2.7–4.2 V, regulated rail 3.3 V — buck-boost spans the full range. Almost every modern Bluetooth-low-energy sensor uses one.
• USB-PD chargers and power banks. 4-switch buck-boost regulates programmable outputs from a battery whose voltage is far from the output.
• Automotive 12 V rails. The car battery sits at 10–16 V depending on alternator and load state; many in-car electronics need a precisely 12 V output across that range, so a 4-switch buck-boost (or a synchronous boost cascaded with a buck) does the regulation.
• Negative supply generation. Op-amp ±15 V supplies from a single +15 V rail; LCD bias rails; GaAs RF circuits needing −5 V from a +5 V system bus.
• P-channel high-side FET drives. A high-side P-FET needs its gate pulled below source to turn on; an inverting buck-boost generates the gate-drive negative rail.
• LED drivers spanning input range. An LED string with V_f = 12–15 V driven from a wide-input adapter (9–24 V) uses a buck-boost for constant current across the full V_in range.
• Solar / energy-harvesting. Source voltage from a small PV panel or thermal harvester varies hugely with conditions; a buck-boost extracts maximum power across a wide V_in range into a stable battery-charging rail.

Common pitfalls

• Underestimating switch voltage stress. Switch sees V_in + |V_out|, not just V_in. A 5 V → −12 V buck-boost needs a 30 V+ MOSFET, not a 12 V part.
• Forgetting capacitor polarity in inverting topology. The output cap's positive terminal sits at the V_out node, which is below ground. Many engineers wire it the wrong way; on power-up the cap pops.
• Slow load-step response from RHP zero. Designing for a 100 kHz bandwidth in feedback when physics caps you at 10 kHz produces an unstable loop that oscillates on transients.
• Inductor saturation at peak. Peak inductor current is higher than I_load (it must be: the inductor stores all the energy during on-time). Use peak-current rating, not average, for I_sat sizing.
• Reference grounding errors. Inverting buck-boost outputs are referenced to ground from the cap negative; the load sees a negative supply. Loads expecting a positive rail need either a level-shifter or the topology choice should be SEPIC instead.
• Continuous vs discontinuous mode confusion. The clean V_out = −V_in·D/(1−D) only holds in CCM. At light load the converter slides into DCM and the equation becomes load-dependent — easy to misregulate if the controller can't sense mode.

The topology that has always existed

The inverting buck-boost is sometimes credited to R.D. Middlebrook's 1976 paper "A general unified approach to modelling switching-converter power stages", which formalised the analysis of buck, boost, and buck-boost as a unified family. In practice the topology existed in vacuum-tube DC-DC supplies of the 1950s and was reinvented multiple times. The Ćuk converter (Slobodan Ćuk, 1976) and the SEPIC (1977) were direct evolutions that solved the inverting-polarity problem of the basic buck-boost. The 4-switch non-inverting variant emerged in the 2000s as battery-powered USB devices needed both buck and boost behaviour from the same hardware with a positive output.

Today every major analog-IC vendor — TI, Analog Devices, Maxim, Microchip, ROHM — offers dozens of integrated buck-boost controllers and complete-module products spanning sub-mA medical-implant supplies up to 100 W laptop chargers. The four-component textbook circuit at the top of this page hasn't changed in seventy years; the silicon and the controllers wrapped around it have.