This is an attempt to list all significantly different and viable permutations of 2 or 3 transistor amplifiers or output stages. Most of these circuits need an input stage and feedback to set the operating voltages, only one or two are in any sense complete amplifier circuits. Trivial permutations such as common-emitter followed by common-base followed by common-collector are not included, if they were that alone would add up to 27 possible permutations.
The first two are shown to illustrate what I mean by 'significantly different', the only difference here being that the first has voltage gain from a common-emitter input stage and the other has slightly less than unity gain using an emitter-follower input. However, the complementary push-pull emitter follower output stage is the important defining feature, so these are really just 'trivial variations' as are use of darlington pairs or cfps in place of the single transistors, or adding cascodes, or a current source in place of a resistor. For the remainder of the permutations only a representative example will be given unless there is some more interesting variation. A third diagram is included which differs from the first by just one resistor, Rx, but the principal of operation is then significantly different. A constant current source instead of Rx would be even better. This addition takes current in one direction from the output stage and keeps it in class-A for signals up to a certain peak negative output level. Assuming the quiescent current is optimised for minimum crossover distortion prior to adding Rx we then get the advantage of high bias class-AB with no crossover at low levels, but without the 'gm doubling' effect with its consequent distortion. The static dissipation, i.e. with no input, will be somewhere between optimum biased 'class-B' and full class-A.
This idea appears to originate with attempts to improve the performance of early opamps, an example is in a 'letter to the editor' in Wireless World, March 1973, page 119, from M.L.G. Oldfield, under the heading: 'Modular i.c. audio mixer'. This suggested ways to improve the audio performance of a 741 op-amp, including the addition of a resistor from the output to the negative supply rail to ensure the output stage stays in class-A over the intended operating range.
The above circuit is less widely known, it is a common-emitter single-ended output stage but with a current source collector load which is modulated by the current in the amplifying stage Tr1. This example is 'class-B' in the sense that the top half switches off during negative going output. This is actually a simplified copy of the output stage of an integrated circuit power amplifier, the Sinclair IC10 (or Plessey SL403) from 1968. This has a range of variants, but in its class-B form it is not very linear, so most examples are class-A.
An early class-A commercial implementation was the Sugden A21 from 1968, the output stage being shown next (ok, it's 5 transistors, but the Sinclair circuit also had more transistors than shown, Tr2 and Tr3 were darlington pairs, so in the same way we could regard Tr5,6 and Tr 7,8 as single darlington devices.), anyway it's sufficiently different to deserve inclusion:
Next below is a version using bipolar transistors and operating in 'class-B', related to the 'White cathode follower' patented by Eric White in 1944. This could perhaps be called a 'White emitter follower', this circuit was published in Wireless World, June 1973, pp 301-302, in an article by P.L.Taylor. The original White version used capacitor coupling from the current detection resistor to the lower modulated current source, but here a common-base stage is used as a non-inverting level shifter. I have an old page here with a few variations of my own incorporating feedforward error correction.
A small-signal class-A version was published earlier in an article 'A Wideband Oscilloscope probe' by L. Nelson-Jones, Wireless World, August 1968, page 276, Figs 3a and 3b.
Next is the circuit used by John Linsley-Hood in his 10 watt class-A amplifier published in Wireless World, April 1969. I only show the output and driver stage to stay in my 3 transistor limit, but the whole amplifier used only 4. The second diagram is a unity gain version.
The next circuit is usually a CMOS IC logic inverter, e.g. CD4069, but these have occasionally been used as linear amplifiers. The complementary enhancement mode mosfets make possible a very simple push-pull output stage. A discrete component version is shown also, it's not very good, the quiescent current will be affected too much by the supply voltage, but it's the simplest version I know.
The next circuit is a jfet common source stage with current source load and a source-follower output stage. The only slightly unusual feature is the way the current source determines the drain voltage of the input stage. This is a useful feature because the input stage current is not well defined, with the 100R source resistor the current through Tr1 could vary between something like 2mA and 6mA for different samples of a typical jfet, and here the current source adjusts its DC current to match the drain current, while acting as a high impedance load for AC. I used this sort of circuit as a headphone amplifier, but with an additional source follower driver stage, otherwise the highly nonlinear input capacitance of the output mosfet adds lots of high frequency distortion. I first saw this idea in a 1970 National Semiconductor application note AN-32 'FET circuit applications', page 5.
The next circuit was patented in 1962 by L.W.Erath, I think it will only work well for a limited range of load impedance. There is some similarity to the Linsley-Hood class-A shown earlier on this page. I saw this on the JLH class-A thread on diyAudio.
One last example for now, there may be later additions.
This uses transformers at input and output, I included something similar on an old page about Symmetry but with a current mirror output. As a class-A output stage if the jfets are accurately matched and ideal square-law devices, then the second harmonics cancel to give theoretically zero distortion, one of the various circuits where symmetry can cancel distortion. In reality jfets are not perfect square-law devices, so distortion will not be zero. A drawback with this common-source circuit is that the output impedance will be fairly high. It could be made lower using source-followers instead to drive the output transformer, but then, again in theory, we lose the purely square-law responses and get lots of higher order harmonics, an effect of the local feedback.
