Notice that the circuit consists of the voltage supply, the load to be driven, the transistor, then ground. This is called low side switching, i.e. we are connecting the circuit to ground when the circuit is turned on. Notice that the current flows through two PN junctions to get from the low side of the load to ground. Each of these junctions impose a voltage drop. Depending upon the transistor this drop is typically in the range of 0.3 to 0.6 volts per junction. This means that the low side of the load will actually be significantly above ground. One can avoid this by swapping the places of the load and the transistor so the transistor connects the load to the voltage supply when it is turned on (Figure 1b). This is known as high side switching. The voltage drop still exists, but now it is on the supply voltage being applied to the load. The ground for both the load and the controlling circuit is the same. In some circumstances this arrangement is more desirable, but it comes at a very high price. Take a look at this circuit from the point of view of the driver at the base. It looks like a diode going from the base to the emitter. In order to turn on this diode it must be forward biased. For the NPN transistors that we are using here, this means that the voltage at the base must be greater than the voltage at the emitter by at least the amount of the diode voltage drop. This is not too hard to achieve if we are controlling a 3 volt circuit from our 5 volt I/O pin that is driving the base. But what we'd really like to do is switch larger voltages, say 12 volts, from our processor. If we use the transistor this way, it ends up acting more like an amplifier giving some proportion of the voltage at the collector instead of the switch that we really want, this can waste a lot of power this way. The only way to really achieve what we want is to put some kind of voltage amplifying device at the base, so that the base voltage is larger than the collector voltage. In summary, use low-side switching to control power, unless you really must use high-side.
The bipolar transistors that we have used so far, are current controlled devices. Using a Darlington reduces the current required to turn on the transistor but the current is still the controlling factor. If we switch to a MOSFET power transistor, we can do the equivalent things, but now we are dealing with a voltage controlled device. Because it is voltage controlled, very little current is needed to switch the transistor which is nice for driving from digital output pin. There is a voltage drop between the drain and source (which are functionally equivalent to the collector and emitter), but in MOSFETs it is caused by the resistance of the semiconductor material in the on state. This resistance can be quite small, thus there can be very small voltage drops across the device. The small voltage drop means an efficiency gain in our transistor switch, but in order to achieve it we need to apply enough voltage at the gate (the equivalent of the base) in order to fully turn the transistor on (if the transistor is only partially on then the drain-to-source current path has a relatively large resistance). Trying to achieve the fully on state of the transistor is where we run into a disadvantage with MOSFETs: the typical power MOSFET requires gate voltages around 8 Volts or more. So now our device must have a bipolar transistor switching power to the base of the MOSFET so it can turn on the real power to the contolled circuit. At least, with this arrangement the bipolar transistor does not need to handle much current. The situation with MOSFETs is even worse when used in a high-side configuration. In this case the voltage at the gate must be substantially higher than the voltage supply at the drain (by 8 or 10 volts). As with bipolar transistors high-side switching, without the use of a voltage amplifier on the control signal, is extremely inefficient.
The transistors that we have worked with here have been NPN or, the MOSFET equivalent concept, n-channel. We could have used the alternate form, PNP ( or p-channel for MOSFET). However because of differences at the quantum mechanical level, PNP devices are less efficient particularly at high frequencies. Therefore we will avoid using them unless there is no other choice.
There is a device that acts like a diode with a control line attached, the SCR, (silicon controlled rectifier) that has some useful properties. It can be used like a transistor to switch current on or off. When turned on it behaves like a diode in the circuit. But when it is turned off, an interesting thing happens: the SCR continues to conduct until the source current stops. This makes it useful in applications where you want to turn something on permenantly but only after, say, other parts of the system have been properly intialized. Figure 2 shows a typical application of an SCR.