The first transistors to become popular—the ones that displaced vacuum tubes—were the bipolar junction transistors (BJTs). Bipolar transistors can be thought of as a way of taking an input current and using that to allow a much bigger, proportional current to flow between collector and emitter. For most transistors, this current gain is around 100x, which means that if you pass 10 mA through the base, the transistor will allow up to 1 A to flow from collector to emitter. If you reduce the input current to 1 mA, only 100 mA will be able to flow through the transistor. This is the sense in which transistors are amplifiers: small changes in a small current can create big changes in a bigger current.

BJTs are made up of a sandwich of silicon layers that have been positively and negatively charged. Positively doped silicon has ions with a net positive charge mixed in, and vice versa for the negative kind. These are then layered together to make sandwiches with either the negative-doped silicon on the outside (NPN) or the positive on the outside (PNP). Either configuration is a “bipolar junction.” When they’re left alone, the middle layer prevents electrical conduction from the collector to the emitter:

BJTs become interesting under the effects of a control current. Although the middle base layer in the silicon sandwich normally insulates the two outside layers from each other, you can run a current from the base to the emitter, following the little arrow in the circuit symbol. Because of the geometry of the silicon and some quantum mechanics, once your control current introduces free electrons into the base layer, a larger proportional current then flows between the collector and emitter (also in the same direction as the arrows).

For the NPN transistor, you provide the control current by raising the base voltage up above that at the emitter. For the PNP transistor, the control current is created by having the base voltage lower than the emitter. Either way, the effect is the same: a small current between base and emitter allows a much larger current to flow between the collector and emitter.

One disadvantage of using BJTs, from the AVR perspective, is that they’re current-driven. The AVR’s digital outputs are only rated for a few tens of milliamps, so if you are using a transistor with a current gain of 100, the maximum current you can run through the transistor is on the order of one amp. If you need to control more current than that, you either need to buy a transistor with higher current gain or figure out another way to get more drive current. One way to do this is to drive your transistor with another transistor. This is not as crazy as it sounds.

Darlington transistors are just a pair of transistors built together into the same chunk of silicon so that the first one supplies drive current to the second. Because the first transistor amplifies the current that is again amplified by the second transistor, instead of having a gain of around 35–100, Darlingtons have a current gain around 1,000 to 10,000. Now the puny 10 mA that the AVR will put out is able to switch 10–100 A on and off. Hooray!

The other downside of BJTs, and this one is unavoidable, is that they have a voltage drop across the collector/emitter pair—that’s wasted from the perspective of your motor, and it heats up the transistor to boot. If you look at the datasheet for a TIP102, one of my favorite power Darlington transistors, you’ll see that it has a collector/emitter saturation voltage of around 2 V. This means that if you’re driving a motor that’s designed for 12 V and 1 A, you’ll need a power supply of 14 V, and you’ll generate 2 V × 1 A = 2 watts of heat in the transistor. It’s gonna get hot and probably need a heatsink.

Finally, in order to control BJT transistors from the AVR, you’ll have to convert the digital I/O voltage level into a current for the transistor input. And to do that, you’ll use a resistor and Ohm’s law.

The metal oxide silicon field-effect transistor (MOSFET for short) is a more recently widespread type of transistor. We’re already using a small-signal MOSFET (the 2N7000) in our basic LED-driving setup, and now we’ll take a second to talk more about its inner workings.

MOSFETs work by passing current through a channel of silicon that’s been positively or negatively doped, just like in the BJT. The difference is that the amount of current that is allowed to pass through the channel depends on an electric field that’s imposed on the slice of silicon by putting a voltage on a metal gate plate. The gate doesn’t actually touch the silicon channel layer. Instead, it’s insulated from the silicon by a thin layer of metal oxide, which explains the mouthful of a name. The electrical symbols for MOSFETs are shown in Figure 14-2.

Figure 14-2. MOSFET Symbols

Applying a voltage to the gate creates an electric field that extends through the insulator and into the channel, which makes a thin “inversion layer” in the channel, which is able to conduct electricity. Here comes the bit you need to remember. With an N-channel MOSFET, a positive voltage from gate to source allows a current to pass from drain to source. With a P-channel MOSFET, a negative voltage from gate to source allows current through, from source to drain. In this sense, the N-channel MOSFET is a bit like the NPN BJT, and vice versa.

Unlike BJTs, MOSFETs are voltage-controlled devices, which means that you don’t have to include a base resistor when hooking them up to an AVR I/O pin—just wire up the AVR pin directly to the gate. Even better, small MOSFETs draw very little current when they’re switching on or off, and almost none when they’re in a steady state, so you don’t have to worry about the AVR’s current sourcing capabilities. (This is because the gate is insulated from the channel.) Finally, MOSFETs have almost no voltage-drop when they’re on: even a small-signal switching MOSFET like our 2N7000 has only a couple of ohms of resistance when switched fully on. This means that less power is wasted heating up the transistor and that the load sees the full driving voltage.

Switching MOSFETs are designed to turn on and off quickly, with gate voltages that are between two and four volts. As such, they’re the perfect “switch” to drive in PWM mode with a microcontroller’s pin. Being small and fast, they can’t pass a whole lot of current or withstand very high voltages, but our 2N7000 is good for a couple hundred milliamps at 5 V or 100 ms bursts of 50 mA at 60 V—both situations that the AVR alone couldn’t handle. In short, switching MOSFETs make the ideal next step when your power or voltage demands are just out of reach of the AVR’s digital output drivers.

If you want to power something really beefy, say an electric bike or a heavy robot, you’ll want a specifically designed power MOSFET (or several). Modern power MOSFETs are just like their smaller switching MOSFET cousins, only larger and with a geometry that’s adapted to deliver more current with less resistance, and thus less wasted heat. The trade-off is that power MOSFETs usually require a higher gate voltage to turn fully on and a little more current as well if you’d like to turn them on and off quickly.

As the MOSFETs get bigger and bigger, the area between the gate and the channel gets bigger. To make the same electric field strength inside the channel of the transistor, more charge is needed. In fact, viewed from the gate and source pins, all MOSFETs behave like a capacitor—a certain charge needs to be pushed onto the gate to create a given electric field between the gate and source, just like a capacitor stores charge in the form of an electric field between its two plates.

With power MOSFETs, this gate capacitance limits how quickly the switch can be turned off and on, which puts limits on how quickly you can PWM the MOSFET switch. Small switching FETs have a small gate capacitance and are easy enough to charge up using just an AVR I/O pin’s internal current source, with a gate capacitance in the tens of picofarads, and switching speeds up to 1 MHz. Larger power MOSFETs can have gate capacitances on the order of thousands of picofarads, which means that you’ve got to charge up effectively a 100 times larger capacitor to get the switch open. This means you’ll need more current sourcing capability and probably more voltage to push it through if you want the power FET to switch on at the same speed. The trade-off for the increased gate size is that you get a MOSFET with only 20 milliohms of resistance capable of switching 12 amps at 80 volts.

If you’re truly pushing the high-voltage, high-current, high-frequency PWM frontier, you’ll want a gate driver chip that’s tailor-made to supply the fast charging that the biggest power MOSFETs need. If you just want to drive a fairly beefy (5 A at 12 V) small-robot motor at moderate switching speeds (<20 kHz), you can probably drive it directly from the AVR, or at worst through a switching MOSFET as a first stage.

A lot of people would start out all this discussion of switches with the relay. After all, a relay is just a switch that’s opened and closed by electromagnet, so they’re easiest to understand. Put a current through a coil of wire, and it pulls a piece of metal into contact with another piece of metal and your switch closes. Done.

The reason relays are near the end of my discussion is that they’re fairly special-purpose these days. They switch on and off comparatively slowly, going click-clack as they physically open and close. This means that PWM is out of the question as a method of control—you can turn a normal relay on and off a couple of times per second, but not much more. And the coils that make the magnetic field take a lot of current. With the exception of precision (costly and sensitive) relays, most require too much current to be directly driven by the AVR’s output pins, so you’ll need something like a MOSFET switch just to run the relay.

There’s one last thing about relays that’s miserable. Being electromagnets on the inside, they build up a magnetic field. When you turn them off, this magnetic field collapses and generates a reverse voltage, which can be quite large. This reverse voltage can, in turn, fry your MOSFET or even the AVR if you don’t give the reverse voltage somewhere to go. So relays also need a diode in parallel with them to allow for this “discharge” current, as shown in Figure 14-3.

Figure 14-3. Relay driving circuit

So relays are a niche switch. Where they really shine is when you can take advantage of the fact that they’re physical switches. For instance, if you want to switch a hairdryer or a floodlight on or off, you need to control a lot of (alternating) current at a high voltage. This is a slightly tricky circuit design for solid state, but it’s easy enough to find a 5 V or 12 V relay that is rated to switch these kinds of household appliances on or off, wire it up to an electrical socket, and you’re done. That’s a job for a relay.

Relays are finnicky, noisy, prone to arcing, not shockproof, and require a bunch of current just to keep them going. You have to build that whole circuit in Figure 14-3 just to drive them. If you really want to get fancy, or if you need to PWM your household appliance, you’re going to want something solid state.

The main contenders here are triacs and solid state relays (SSRs). Triacs are like transistors but used for AC current instead of DC. SSRs are basically triacs with some extra circuitry to help isolate the control side from the AC line voltage. Many (most?) SSRs have an LED on the control side and a light-sensitive triac on the AC side, so that turning on or off a washing machine looks to your AVR like blinking an LED; but because there’s no direct electrical connection between the AVR side and the wall-voltage side, there is no corresponding electrocution hazard, and that’s a darn good thing.

We won’t be using either of these in our projects here, but if you ever need to control a wall-voltage device from the AVR, you should look into an SSR. I used one once, along with a temperature sensor, to regulate the heater in an electric coffee roaster.

If you want to control real power with the AVR, there’s a number of switch-like devices out there that will help you do it. What you need to use will depend on the specifics of the motor or laser or home appliance you want to run.

For providing power to small-robot DC motors, BJTs—particularly Darlington BJT circuits—can work just fine. If you need more current, the advantages of a (more complicated) power MOSFET circuit become more apparent. Keep both of them in mind.

Small switching MOSFETs, like the 2N7000 that we’re using, are great for driving small-to-medium loads, even if you have to use a few in parallel. We’ll be using a 2N7000 to drive the laser in the laser-sundial application, and for driving power MOSFETs.

Finally, when you need to control house-voltage AC currents, you’ve got two main choices. If you’re doing low-frequency switching, it’s hard to beat the simplicity of a relay driven with a switching FET. If you need to turn the device on and off a lot, you’ll want an SSR.

A summary of all of the possible switch choices can be found in Table 14-1.