Untried and Abandoned Versions.
Some of my ideas never seemed worth actually building and testing, but may be of interest to someone somewhere, so I have included this page, and will add any old ideas I remember. None have been tried, so there may be problems I have not thought of, and I can give no assurance that any will work well or at all.
A DIRECT COUPLED MOSFET AMPLIFIER
This is just the standard MJR7 circuit modified for a direct coupled output. There is then no speaker protection under fault conditions, and it may not even be a good idea to add fuses in the supply lines. If just one fuse blows all sorts of unpleasant things may happen, for example the output could swing to the other supply, applying 50V to the speaker. If you are lucky the other fuse blows before the speaker is seriously damaged. Personally I would add some sort of speaker protection circuit, and include various components to control what happens after fuse failure.
The output offset is minimised by adjusting the 10k preset. For best offset temperature stability the two 2SC2547E should be fairly well matched for current gain at 0.5mA collector current, and kept in thermal contact. In theory two of the 200k resistors should be increased to ensure both transistors operate at the same collector voltage, but in practice this may not be a significant improvement.
Another problem is input stage clipping, and to limit current through this stage I added a BAV20 diode. The recovery from clipping may be less than perfect, but maybe not too bad. There are probably better ways to do this.
The original aim of the mosfet circuits was to produce the simplest possible class-AB circuit with acceptable distortion levels, so they were named according to the number of transistors, with the MJR6 having 6 and the MJR7 having 7. There was an idea for a MJR5, and here is one version. The loop gain is lower than in the other versions, so distortion would probably be poor. The input transistor looks wrong, but it really can work like that.
A FEEDFORWARD BJT VERSION.
I never made a discrete component version of the original Electronics World design, and this is just one of the ideas I had. It uses the input and driver stages from the MJR7, and really adds nothing new, apart from a 3k3 resistor between Tr7 and Tr8 emitters. This is intended to null any distorted signal fed to the output from Tr7 emitter via one of the 33R resistors.
A FEEDFORWARD BJT VERSION WITH BIAS CONTROL.
The original idea for bias control was not very well thought out. As soon as the amplifier clips positive there will be no bias current detected and the control circuit will try to increase the current. This may be a big problem, so this version adds a few extra components to deactivate the control near clipping. Two LEDs are added to give some limited protection against shorted outputs, the BJTs are probably easier to destroy than mosfets, and the simple fuse protection may not be enough.
MY FIRST MOSFET DESIGN, 1984.
This is not actually untried, just abandoned by me as a design approach. It is almost a direct copy from the Hitachi application notes, but I wanted to make it as a commercial product so there had to be additional features not included in the other amplifier boards already available at that time, for example from Maplin. Any direct coupled amplifier really needs speaker protection, and I decided to add relay protection on the board, together with a heatsink temperature detector and supply fuses. This caused a few problems because if just one supply fuse blows bad things can happen, for example the output could swing to the other supply and stay there until the other fuse blows, which may or may not be before the speaker is seriously damaged. Five of the components are there just to control what happens after fuse failure. None of this is needed if we revert to a capacitor coupled output. Then we get other problems such as poor damping factor and some added distortion, but just including the capacitor inside the feedback loop is then an easy solution.