Chapter 2 · Foundations

Semiconductor Switches

Every converter in this tutorial is built from a handful of semiconductor switches turning on and off. The switch you choose sets the converter's voltage rating, its efficiency, how fast it can switch and how hot it runs. This chapter explains how each family of devices conducts and blocks, how to model its losses, and how to pick one from a datasheet.

The ideal switch, and why no real one is ideal

An ideal switch has four properties. Real devices trade them against each other, and that trade-off is what separates one device family from another:

The device families

Diodes: the uncontrolled switch

A diode conducts when its anode is more positive than its cathode and blocks otherwise. You can't turn it on or off; the circuit does. Converters use diodes as freewheeling paths: when a transistor turns off, the inductor current has to keep flowing somewhere, and the diode gives it a path. Three types matter:

• Silicon PN fast-recovery diodes block high voltages, but when they turn off they first conduct a burst of reverse current while stored charge is removed. That reverse-recovery charge Q_{rr} causes loss in the diode and in the transistor that turns on against it.
• Silicon Schottky diodes have a lower forward drop and almost no stored charge, but are practical only at low voltages, typically up to a few hundred volts.
• Silicon carbide (SiC) Schottky diodes combine high blocking voltage with almost no reverse recovery. They are a common upgrade in PV boost converters because they cut the turn-on loss of the main switch.

MOSFETs: voltage-controlled and fast

A MOSFET turns on when its gate-source voltage is raised above a threshold, typically by driving it to 10–15 V for silicon parts. Only majority carriers conduct, so there is no stored charge to remove and it switches very quickly. When on, it behaves like a resistor, R_{DS(on)}, and it can conduct in both directions. It also contains a body diode from source to drain, which conducts when the channel is off and current flows in reverse.

The catch is voltage. In silicon, the on-resistance of a MOSFET rises steeply with its voltage rating, so low-voltage silicon MOSFETs are excellent and high-voltage ones get expensive. Superjunction designs push silicon MOSFETs to around 600–900 V, which covers many single-phase inverters and boost stages.

IGBTs: MOSFET control, bipolar conduction

An insulated-gate bipolar transistor is driven like a MOSFET, through an insulated gate, but conducts like a bipolar transistor. Minority carriers flood the device and lower its resistance, so at high voltage and high current it conducts with a small, almost constant drop. The price is speed: at turn-off the stored carriers take time to disappear, producing a tail current that adds switching loss. IGBTs are available from 600 V up to several kilovolts and are the traditional choice for three-phase string inverters, central inverters and wind converters, usually at switching frequencies below about 20 kHz.

An IGBT conducts in one direction only, so it is almost always paired with an anti-parallel diode, often in the same package or module.

Wide-bandgap devices: SiC and GaN

Silicon carbide and gallium nitride are semiconductors with a much wider bandgap than silicon. A wider bandgap means a much higher critical electric field, so a device can block the same voltage with a thinner, more heavily doped drift region, and therefore with far less on-resistance and less charge to move when it switches.

• SiC MOSFETs, commonly rated 650 V to 1,700 V (1,200 V is the most common), switch much faster than IGBTs with no tail current, and tolerate high temperatures. They suit 800 V DC links and 1,000–1,500 V PV systems, where they let designers raise the switching frequency and shrink the magnetics.
• GaN HEMTs, commonly rated 100 V to 650 V, are the fastest switches available. They have no body diode and so no reverse recovery, though they still conduct in reverse with a higher voltage drop. They suit compact, high-frequency converters such as microinverters and panel-level DC-DC converters. Their speed makes the gate driver and PCB layout critical.

Summary of the families

Thyristors, the original power switch, still appear in very high-power line-commutated converters, but they can only be turned on by their gate, not off, so they play little part in the switch-mode converters of this tutorial.

Modelling conduction loss

When a switch is on, the power it loses depends on how its voltage drop varies with current. Two models cover every device in the table:

Because a MOSFET's drop grows in proportion to current while an IGBT's starts at a knee and then grows slowly, each one wins in a different current range:

Worked example: which device loses less?

Take the two devices in Figure 2.1, each conducting a steady DC current (so I_{avg} = I_{rms} = I):

Setting the two losses equal gives the crossover: I^2 R = V_0 I + r I^2, so I = V_0 / (R - r) = 1.0 / 0.025 = 40 A. This is why high-current converters at high voltage have traditionally used IGBTs, and why paralleling several MOSFETs, which divides R_{DS(on)}, is a common way to move the crossover up. Remember that the MOSFET's resistance rises when it gets hot, which moves the crossover down.

Choosing a voltage rating

A switch must block the highest voltage it will ever see, including the spikes caused by stray inductance when it turns off fast. Designers therefore leave a generous margin over the DC-link or input voltage:

Reading a datasheet

A switch datasheet holds dozens of numbers. These are the ones that decide whether a device suits a converter: