For any power output over about 20 watts into 8 ohms an i.c. op-amp driver stage becomes inconvenient. One means to increase the output voltage swing of an op-amp above its maximum rated supply voltage is to use 'bootstrapped' supply lines, but the extra complexity and design problems make this unatractive. Another solution described in the next note is to use two amplifiers in bridge mode. There are good reasons for prefering this solution, to do with safe operating area rating of power transistors.
A single high power amplifier must work from a higher supply voltage, and a property of bipolar transistors known as 'second breakdown' becomes important. For example, a high power transistor, the BD313, is rated at 10 amps, 80 volts and 150 watts. For an amplifier rated at close to 100 watts into 8 ohms, we need a +/- 40 volt supply. Unfortunately loudspeaker designers pay little attention to the 8 ohm standard impedance, and impedances down to 4 ohms at least must be accommodated. Even so, the BD313 seems adequate to drive a resistive load of 4 ohms with a +/- 40 volt supply. The peak voltage across the transistor is 80 volts and the peak current 10 amps. With a 4 ohm load the maximum power dissipation in the transistor is 100 watts, so there seems to be a good safety margin. One reason why this is not so is that real loudspeaker loads are not pure resistors. A part reactive load, 4 +/- 4j ohms, gives maximum transistor power disipation of 200 watts. Also, the power rating of the transistors is valid only at a junction temperature of 30 degrees C. Even worse, second breakdown occurs above 30 volts collector-to-emitter voltage, and the power rating is reduced from 150 to 75 watts at 50 volts, and only 16 watts at 80 volts. The power rating, far from being adequate, may need to be FAR greater. Transistors with good second breakdown performance must be selected, and the use of parallel pairs may be necessary, together with complex overload protection circuits. The saving in complexity compared to a bridge circuit may be less than expected.
Be that as it may, provided we are not tempted to aim for output power ratings beyond what the power transistors used are capable of under realistic operating conditions, a single discrete transistor amplifier may be the most commercially atractive option. There are plenty of published designs and standard circuits known to give adequate performance, so I will only include one example here.
The most common class-B transistor amplifier circuit uses a single transistor driver stage to provide most of the voltage gain, and drive the output stage. The output stage invariably has unity gain, so the driver transistor has the full output voltage appearing at its collector, and this can lead to high distortion. The earliest reference I can find to this source of distortion was in an article by A.R.Bailey in Wireless World (Reprinted in 1974 in a book called 'High Fidelity Designs', which neglects to mention the original publication date.) describing a power amplifier design in which distortion was 3 times higher than predicted, and this was believed to be caused by the variation in collector voltage of the driver transistor modulating the stage gain. Distortion also increased faster than expected at high frequency because of modulation of the collector to base feedback capacitance. Different transistors were tested and some found to be much better than others. The best found was the 40362 (RCA), but this now seems to be unavailable. More recent types such as the 2SD756A, designed for driving mosfet output stages, could be tried, but the following circuit is suggested to avoid such problems almost entirely.
The output stage is not shown in detail, and this is just the same as in previous circuits. This example uses a 'cascode' driver stage in which a second transistor operates in common-base mode. The lower transistor then feeds into the low impedance emitter, and therefore has low voltage gain and low voltage variations at its collector. The distortion from this source is then small, but there is a disadvantage, that the common-base stage has a high output impedance, and the non-linear input impedance of the following class-B output stage would add distortion at this point. An intermediate emitter-folower stage will reduce the impedance driving the output stage, and may significantly reduce distortion. A darlington transistor could be used instead to even further reduce the impedance.
The diagram is only intended as an example of the sort of input and driver stage which could be used, and there are plenty of variations, and ways this particular circuit could be improved. The cascode driver stage, however, is something I would strongly recommend. It has major advantages, and the only disadvantages are the high output impedance, which is in any case reduced by the feedback from this point by the compensation capacitor, and also the reduction in voltage swing compared to a single transistor. The following stage actually shifts the voltage in the right direction to partly compensate for this, and no serious reduction in maximum output is produced.
