|
|
If these circuits are so poorly conceived, why do they persist? The answer requires only one word: feedback. Feedback irons out the distortion caused by the unbalanced drive: the greater the feedback ratio, the less the distortion.
Interestingly enough, under one special circumstance, both circuits will fall into a balanced drive state: when the output is shorted to ground. In this case, the top and bottom output devices will see a balanced drive signal. But for any impedance other than zero ohms, the circuits will be see an unbalanced drive for the output stage. For example, if the output device has 1 amp/volt transconductance figure and the load impedance is 8 ohms and the output stage in run in Class-B, then a +10 volt pulse is applied to the top device's input, the output will swing 8.888 volts positively. On the other hand, if a +10 volt pulse is applied to the bottom device's input, the output will swing 17.777 volts negatively. Why the asymmetry? When the top device saw the +10 volt pulse, the bottom device was completely turned off, which means that the current flowing through the load and the top device must match, as these two define only one current path. The current flowing through the load is given by:
Iload = Vo / Rload.
Where Vo is the output voltage swing. The current through the top device is given by:
Itop = (Vin - Vo)Gm.
Where Vin is the drive voltage. The current flowing through the bottom device is given by:
Ibot = Iload + Itop.
The positive output voltage swing is given by:
Vo = RloadVinGm / (1 + RloadGm).
Where Vin equals the drive voltage. The negative output voltage swing is given by:
Vo = 2RloadVinGm / (1 + RloadGm).
Obviously, these two formulas differ by a factor of 2.
Restoring a balanced drive would seem easy enough: just provide twice the signal voltage to the top device as is provided to the bottom device. And in a pure Class-B amplifier this ratio works; but in a Class-AB amplifier, the bias current will skew the ratio. Furthermore, even in the Class-B amplifier, this fixed ratio is still suboptimal. For example, as the load approaches zero ohms, the top device will needlessly dissipate twice the wattage as the bottom device: a +2 volt pulse to the top device's input results in 2 amps of current conduction, but a +1 volt pulse results in only 1 amp of current conduction.
|
|
|
|
In a pure Class-A amplifier this fixed ratio leads to a strange state. Normally, a push-pull Class-A amplifier's idle current equals one half of its peak output current swing. And when the load equals infinity, the idle current remains constant. But when we force a fixed 2:1 ratio, the amplifier will cease to draw any idle current when input signal swings negatively enough to cutoff the bottom device. Conversely, when input signal swings positively the bottom device's increased current conduction forces the amplifier dissipation to at least double (and this happens in the absence of a load).
Feedback saves the day for this poorly designed amplifier. The feedback resistors are themselves a load for the output stage. And the feedback senses the lopsided behavior and attempts to correct it. Break the feedback loop, however, and the distortion becomes apparent. An analogy might be having a car that veers to the right when driven on a flat road. So rather than fix the poor geometry of the front-end, you replace the tires on the right side of the car with a larger pair. Now the car drive straight on flat road, but on a inclined road it veers once again. And on it a winding road it takes all your concentration and strength not to crash your car. This will not do. The only solution (other than gobs of feedback) is to dynamically create an equalizing drive voltage for the top and bottom devices while they are working into any load impedance in any class of operation. Julius Futterman understood this requirement and all of his OTL amplifiers include dynamic drive equalization. True, the Futterman amplifiers also used a huge amount of feedback. But in his amplifiers the feedback wasn't used up equalizing the drive signals. Instead, it was used to lower the noise, the distortion, and the output impedance of the amplifiers.
Most importantly, dynamically equalizing the drive voltage for the top and bottom devices allows us to build low and even feedback free designs, designs that do not need feedback's heavy hand to steer the amplifier out of distortion.
Of the possible six push-pull nonsymmetrical output stage topologies only half will work with a feedback free design. These topologies arrange the both output devices to function in a follower mode by returning 100% of the output to the output device's inputs. In other words, these topologies realize 100% degenerative feedback just as a cathode follower or source follower or emitter follower does.
|
|
