The standard definition of class B operation is that the conduction angle is 180 degrees or that the plate of the tube in question conducts for 1/2 the input cycle. Two tubes in a push pull configuration are needed for audio, so that the other tube can complete the output for the other half of the input cycle while the first tube is cutoff.
When tubes are to be grid driven, then the positive going half of the input cycle turns on the plate conduction. Driving the grid more positive, causes the plate current to increase and the plate voltage to decrease, until the plate voltage has been reduced to zero. This point is known as saturation (a more positive driven grid, will not cause an increase in plate current, because the plate voltage has been reduced to zero).
Class B circuits typically have a lower plate load resistance than Class A. This allows a much larger swing in plate current before the plate voltage hits zero (saturation). Which of course means it can accept more drive, reaching saturation at a higher plate current which yeilds more output power.
The grid is often driven far above the zero voltage point (with respect to the grounded cathode) so that the tube draws grid current. This causes an extra load on the driving stage, but only while the grid draws current. This sudden increase of load on the driving stage will cause distortion of the applied signal voltage. That is to say, that the voltage will not go as high as it would have, had the load not increased.
The other problem is what to do with the current that is drawn from the grid. If the circuit uses zero bias tubes and the plate supply for the tubes is kept low enough to use the zero bias technique, then we can simply send the grid current directly to ground. Generally this is done by grounding the center tap of the driver transformer. But if the tube requires a negative bias voltage to hold the plate current back to near zero, then the grid current must be disposed of without letting it cause the bias voltage to go more negative. Should this occur, then the bias would not be right for Class B, instead it would be more like Class C, (very distorted for audio).
Driving Class B control grids poses two problems.
This is done using a shunt (parallel) regulator from the bias point to ground.
The bias power supply does not have to supply current to anything,
it is just there to provide a static voltage to the grids of the tubes,
but the regulator must be there to get rid of the current that is produced
by the grids input signal.
Another way to help this condition is to have a driving stage that is capable of much more driving current than is ever drawn by the grids of the modulator tubes. In this way, the grid current is only a small percentage of the capability of the driving stage. Therefore the distortion produced by grid current load change is also very small.
There is no perfect solution, but the distortion caused by the above
mentioned problems can be reduced to levels undetectable or certainly exceptable.
The following diagram shows my solutions in more detail.