During a recent tutorial the subject of the different classes of amplifier bias arose. As you probably know the base current of a transistor amplifier, with no signal present, usually depends on the bias voltage applied from a voltage dividing pair of resistors. In Class A, we adjust this voltage to make the collector take about half of the saturation current (the most it can manage) then, when we apply an appropriately sized signal between the base and emitter, the output voltage will swing between almost the rail voltage and down to a very low voltage. Assuming we have not overdone the input voltage swing, the output will be a larger, exact reproduction of the input signal, only upside down. Should we overdo the drive signal, the device will reach maximum current too soon in the input cycle and the
current may be cut off altogether during the other half cycle. This means that the output signal will have flat tops and bottoms, generating distortion which, in an RF circuit, produces lots of nasty signals nobody wants to hear. For SSB, where the amplitude is always varying, Class A is almost a must, the output quality being superb, but only if the bias is set correctly and the voltage swing of the input signal is within limits. The only problem with Class A amplifiers is that the amount of amplification is limited and the efficiency is only 30% or so, meaning that much waste heat is being generated, so the device/s will need lots of cooling.
So could we get a higher output SSB signal and yet get the efficiency up? Well, there is Class B biasing, where the device is only just conducting with no signal present. Ah, you say, doesn’t that mean that half the input waveform is not going to be amplified and therefore lost ? Quite true ! However, we can arrange for two devices to work together, one device working when the other is off and vice-versa. Say the input signal is applied to two devices, one PNP and the other NPN. Then the common output circuit gets half from one device followed by a pulse from the other, resulting in a distortion free signal, if, and only if the devices are equally matched for conductance and the bias is set perfectly. Whilst a device is on, the other is doing nothing and so is cooling off and the efficiency is greater than Class A but less than Class C, maybe 50%.
We could use two ‘push-pull’ NPN or PNP devices, working like a two cylinder horizontally opposed motor-cycle engine. Here the input signal has to be split so that input for each device gets the same signal, but in anti-phase. A centre tapped primary of a transformer is connected between the collectors/drains, the supply going to the centre-tap. The signals from each amplifier are then combined in the transformer output winding. Again, the devices should be a matched pair and the Class B bias set very accurately. In Class C operation, the bias is adjusted until the device/s stop conducting altogether, with no signal present. We now size the input signal such that only the very peaks cause the device to conduct. Then the collector/drain voltage falls rapidly but rises again equally rapidly when the device is cut off again as the input signal falls shortly afterwards. For this small fraction of each cycle the device conducts heavily, but for the remainder of the cycle it can cool off. A tuned circuit in the output will ring like a bell when the voltage falls and rises rapidly. The resulting oscillation would normally die away but, fortunately, the tuned circuit is being struck every cycle by another voltage swing from the o/p of the device, enabling the tuned circuit to produce a sine wave output. It should be obvious that Class C can only be used if the driving signal amplitude is constant, i.e. CW and, using FM, where only the frequency changes. Class C efficiency? 66%. Class C devices make good frequency multipliers, with the o/p tuned circuit being hit with a pulse every other cycle (doubling) every three (tripling) and sometimes every five cycles (quintupling) The Q will need to be high to ensure a good ‘flywheel’ effect. And yes, there is also Class D biasing, which can be up to 90% efficient, but that is another story. Finally, for high power PA circuits, it is vital that the supply voltage is ‘stiff’, in our parlance. It should be self evident that if the supply voltage to amplifiers sags on peaks of current demand, the output signal will not be an exact reproduction of the amplifier input. So, always go for a PSU with a few more amps available than your rig needs. Keep that in mind that when your SSB rig is not transmitting anything at all, i.e. between gaps in your speech, most of the other devices in your rig are still drawing current, sometimes several amps. So allow for this as well as the PA current required.
That’s about it for this time. Keep well, keep happy.