Power components
In signal electronics, what matters is gain. In power electronics, what matters is how much heat the component generates and how that heat is removed. Everything else —choosing the device, sizing the heatsink, protecting the switching— follows from that.
01The complete family
Some of them already appeared in special semiconductors. Here they are arranged by what really decides the choice: voltage, current and switching frequency.
| Component | Turns on | Turns off | Frequency | Home turf |
|---|---|---|---|---|
| SCR | Gate pulse | By itself, at the zero crossing | Mains (50 Hz) | Controlled rectifiers, soft starters, high-power phase control |
| Triac | Pulse, in both directions | By itself | Mains | AC phase control up to about 40 A |
| Diac | By voltage (~32 V) | By itself | — | Triggering triacs symmetrically |
| UJT | By emitter voltage | By itself | — | Relaxation oscillator: generates the trigger pulses |
| GTO | Positive pulse | Negative pulse | Up to 1 kHz | Rail traction, high power with controlled turn-off |
| MCT | MOS | MOS | kHz | Thyristor with MOS control. It promised a lot; it was displaced by the IGBT |
| MOSFET | Gate voltage | Voltage | 10 kHz to 1 MHz | Low voltage and high frequency: switching power supplies, automotive |
| IGBT | Gate voltage | Voltage | 2 to 20 kHz | Medium and high voltage with fast switching: motor drives, welders, inverters |
- UJT — unijunction transistor. With an R-C it forms a relaxation oscillator: the capacitor charges up to the trigger voltage and discharges suddenly into a pulse transformer. It is the classic way —from before microcontrollers— of generating the trigger for an SCR with an adjustable angle.
- GTO — gate turn-off thyristor. It is turned off with a large negative current pulse (up to one fifth of the anode current). This solves the big limitation of the SCR —which does not turn off by itself on DC— at the cost of a heavy drive circuit.
02The IGBT: the best of both worlds
It is the central component of modern power electronics. Its input is that of a MOSFET —insulated gate, no continuous drive current— and its output is that of a bipolar, with the low drop of a junction instead of a resistance.
In a MOSFET, RDS(on) grows very fast with the voltage it has to withstand: a 600 V MOSFET has an enormous resistance compared with a 60 V one. The IGBT, on the other hand, drops 1.5 to 2.5 V almost regardless of the current or the blocking voltage.
Rule of thumb: above 300 or 400 V and tens of amperes, the IGBT wins. Below that, and at high frequencies, the MOSFET wins.
When it turns off, the charges stored in the bipolar region take time to recombine, leaving a current tail lasting microseconds. That tail, multiplied by the voltage, is energy lost on every switching event: that is why the IGBT cannot reach MOSFET frequencies.
Switching 10 A at 20 kHz:
- At 48 V with a 10 mΩ MOSFET: conduction loss = 10² × 0.010 = 1 W. Unbeatable.
- At 400 V, an equivalent MOSFET would have on the order of 0.5 Ω: 10² × 0.5 = 50 W. Unworkable.
- At 400 V with an IGBT with VCE(sat) = 2 V: 2 × 10 = 20 W of conduction loss, plus the switching losses.
The crossover between the two technologies is exactly that: comparing I²R against V·I.
03The calculation you cannot skip: the thermal one
A power component is not chosen by its maximum catalog current: it is chosen by verifying that its junction does not exceed the allowed temperature with the heatsink it will be given. The thermal circuit is solved just like an electrical one: power plays the role of current, temperature of voltage and thermal resistances of resistors.
MOSFET switching 15 A with RDS(on) = 17.5 mΩ, in an ambient temperature of 40 °C.
- Conduction power: P = I²R = 15² × 0.0175 = 3.94 W.
- Without a heatsink, with Rja = 62 °C/W (TO-220 package in free air): Tj = 40 + 3.94 × 62 = 284 °C. It is destroyed.
- With a heatsink of 3 °C/W, a 0.5 mica insulator and Rjc = 1.5: Tj = 40 + 3.94 × (1.5 + 0.5 + 3) = 40 + 19.7 = 59.7 °C. Perfect.
- The component's limit is usually 150 °C, but the design aims not to exceed 100 to 110 °C: service life is cut in half for every extra 10 °C.
And the switching losses still have to be added, which at high frequency can exceed the conduction losses:
When the datasheet says “50 A,” it is almost always with the case at 25 °C —a condition that does not exist in reality— and without considering switching. The usable current is typically half or less. The number that counts is the one that comes out of the thermal calculation with the real heatsink and the real ambient temperature, which inside a closed panel in the sun can be 60 °C.
04Protecting the switching
Every inductive load —relay, motor, contactor coil— generates an enormous overvoltage when its current is interrupted. The antiparallel diode gives that current a path. Without it, the transistor breaks down on the first switching event. In AC, the equivalent is the snubber.
100 Ω in series with 100 nF, in parallel with the device. It limits dv/dt —which can trigger a thyristor on its own— and absorbs the voltage spike from switching. It is mandatory with inductive loads on triacs.
A thyristor that turns on too fast concentrates all the current in one spot of the junction and is destroyed locally. A small series inductance limits that rate. Manufacturers specify the maximum di/dt in A/µs.
An ordinary fuse does not protect a semiconductor: the semiconductor burns out first. gR or aR fuses are used, characterized by their I²t, which must be lower than the I²t the device can withstand. It is the only real protection against a short circuit.
All the labs in this topic are done with an isolation transformer or with a low-voltage laboratory power supply. The oscilloscope ground is connected to earth: connecting it to a live point of the circuit causes a dead short. To measure voltages between two floating points you need a differential probe, not two subtracted channels.
05Two worked circuits
An SCR between the 220 V mains and a 44 Ω resistive load, triggered at 60°.
- It conducts only on the positive half-cycle, and only from 60° to 180°.
- Average voltage = (Vp/2π)(1 + cos α) = (311/6.283) × (1 + 0.5) = 74.3 V.
- Average current = 74.3/44 = 1.69 A.
- The SCR must withstand the peak reverse voltage, 311 V: a 600 V one is chosen, with the usual margin of double.
With triggering at 0° the average voltage would be 99 V, and at 180°, zero: that is the whole range of regulation.
An IGBT switching 100 V across a motor, with duty cycle D.
- Average voltage across the load = D × 100 V: with D = 0.6, 60 V.
- A freewheeling diode is needed in parallel with the motor: when the IGBT turns off, the inductance current keeps flowing through it.
- The motor current is not interrupted: it rises and falls around its average value, with a ripple that depends on the frequency and the inductance.
That circuit, with four switches in a bridge, is exactly the variable-speed drive of topic 7 of this subject.
06In the lab
With a MOSFET and an IGBT conducting the same DC current (5 A) from a laboratory power supply, measure the voltage drop of each and calculate the dissipated power. Measure the package temperature with an infrared thermometer after five minutes, with and without a heatsink. Compare with the thermal calculation.
Build the UJT relaxation oscillator with a pulse transformer, and use it to trigger an SCR on a resistive load at low voltage. Vary the potentiometer and watch on the oscilloscope how the firing angle shifts. It is the circuit used for phase control before microcontrollers.
Switch a relay with a transistor and observe on the oscilloscope the collector voltage at turn-off, without the diode (with reduced voltage and an expendable transistor) and with the diode. The spike goes from hundreds of volts to less than one volt above the supply.
With an isolation transformer and reduced voltage, build the controlled rectifier from the example and measure the average voltage with a multimeter and with an oscilloscope for three firing angles. Compare with the formula. Repeat with an inductive load and observe that the current no longer stops at the zero crossing.
07Common mistakes
| Symptom | Usual cause |
|---|---|
| The component burns out even though it “can handle the current” | The thermal calculation was not done: the catalog current assumes ideal conditions. |
| It gets much hotter than calculated | The switching losses were left out, or the gate is not driven with enough voltage and the device operates in the linear region. |
| The triac fires on its own with an inductive load | Excessive dv/dt: the snubber is missing. |
| The transistor is destroyed when turning off a relay | The freewheeling diode is missing. |
| The heatsink is lukewarm and the component is boiling hot | Poor thermal contact: missing grease, the mica insulator is misplaced or the screw is loose. |
| A short circuit burned out the semiconductor and the fuse was left intact | Ordinary fuse instead of an ultrafast one: the fuse's I²t is greater than the device's. |
| A 600 V MOSFET gets extremely hot | At that voltage its RDS(on) is high: an IGBT was the right choice. |
| The SCR never turns off | It is supplied with DC. Forced commutation or a GTO is needed. |
08Self-assessment
What is the essential difference between an SCR and a GTO?
The SCR can only be turned on by the gate: it turns off when the current drops below the holding current. The GTO can also be turned off by the gate, with a considerable negative current pulse.
Why did the IGBT displace the MOSFET at high voltage?
Because the on-resistance of a MOSFET grows very fast with the blocking voltage, while the IGBT keeps a nearly independent drop of 1.5 to 2.5 V. Above 300 or 400 V, the IGBT dissipates much less.
A MOSFET dissipates 4 W with Rjc = 1.5, mica 0.5 and heatsink 3 °C/W, at 40 °C ambient. What is the junction temperature?
Tj = 40 + 4 × (1.5 + 0.5 + 3) = 40 + 20 = 60 °C. Very comfortable against the 150 °C limit.
What happens to the switching losses if the frequency is doubled?
They double: they are proportional to frequency, because each transition costs the same energy and there are twice as many transitions per second. This is what limits the maximum operating frequency.
What is a UJT used for in power electronics?
To build a relaxation oscillator that generates the trigger pulses of an SCR or a triac, with the angle adjustable by an R-C. It was the standard solution before microcontrollers.
An SCR on 220 V triggered at 60°, with a 44 Ω load: what is the average voltage?
Vavg = (Vp/2π)(1 + cos α) = (311/6.283)(1 + 0.5) = 74.3 V, and the average current 1.69 A.
Why does an ordinary fuse not protect a semiconductor?
Because its specific energy I²t is much greater than what the junction can withstand: the semiconductor is destroyed before the fuse acts. Ultrafast fuses with an I²t lower than the device's are needed.
What function does the snubber serve besides protecting?
It limits dv/dt, that is, the rate at which the voltage across the device rises. In a triac, an excessive dv/dt fires it on its own even without a gate pulse.
What is the current tail of an IGBT and why does it matter?
The current that keeps flowing for a few microseconds after the gate voltage is removed, due to the charges stored in the bipolar region. Because it occurs at high voltage, it causes losses on every switching event and limits the operating frequency.
Why design for a 100 °C junction if the limit is 150?
Because service life is reduced by roughly half for every 10 °C of increase, and because margin must be left for the worst case: summer, closed panel, dusty heatsink, stopped fan.