Hi Guys
Where Cordell tended to add BJT pairs with identical Re and idle current through each new pair to keep the overall situation stable, Leach suggested that the optimal idle current could be shared among additional paths. I can't recall what Self said about it if he even did. His way might have followed Leach in this regard.
Everyone seems to concur that there is an optimal idle point for lowest THD in a class-B output stage, with most defining class-B as being idled at the low-THD point. Over or under this point THD is higher. Without a THD meter, you can't set class-B idle precisely and you end up with a class-AB compromise. However, if there is enough feedback and/or openloop bandwidth, then THD should be low anyway, especially up to 1kHz or so.
Overbias to an extreme gives you a broad class-A region where crossover distortion is abated but there are still transition artifacts when the signal goes out of the class-A region. Self made a point that the gm-doubling in class-B amps itself is not so bad an effect as when the transconductance drops when the other side of the circuit stop conducting. Were gm constant, then crossover artifacts and class transition artifacts would be greatly diminished. This suggests that a local loop to maintain high-gm, like the error correction circuit in Cordell's amp, could be of benefit to every type of output stage.
In the example of the poster above, gm for each paralleled output will be reduced by conducting a part of the current. The overall gm will be the same as with a single device. Paralleling devices broadens the gain wobble near crossover and reduces its amplitude, so overall THD will be lower even driving the same load and with shared idle current instead of multiplied idle.
Biasing for a larger class-A region pushes the class transition away from where most listening is done, so this can be a good thing.
The point here is to add the output pairs for THD reduction NOT for extra power or current handling. Self showed that this works as hoped and is a reliable improvement. Cordell showed this as well and Leach concurs in his low-TIM amp evolution.
Have fun
Kevin O'Connor
londonpower.com
Where Cordell tended to add BJT pairs with identical Re and idle current through each new pair to keep the overall situation stable, Leach suggested that the optimal idle current could be shared among additional paths. I can't recall what Self said about it if he even did. His way might have followed Leach in this regard.
Everyone seems to concur that there is an optimal idle point for lowest THD in a class-B output stage, with most defining class-B as being idled at the low-THD point. Over or under this point THD is higher. Without a THD meter, you can't set class-B idle precisely and you end up with a class-AB compromise. However, if there is enough feedback and/or openloop bandwidth, then THD should be low anyway, especially up to 1kHz or so.
Overbias to an extreme gives you a broad class-A region where crossover distortion is abated but there are still transition artifacts when the signal goes out of the class-A region. Self made a point that the gm-doubling in class-B amps itself is not so bad an effect as when the transconductance drops when the other side of the circuit stop conducting. Were gm constant, then crossover artifacts and class transition artifacts would be greatly diminished. This suggests that a local loop to maintain high-gm, like the error correction circuit in Cordell's amp, could be of benefit to every type of output stage.
In the example of the poster above, gm for each paralleled output will be reduced by conducting a part of the current. The overall gm will be the same as with a single device. Paralleling devices broadens the gain wobble near crossover and reduces its amplitude, so overall THD will be lower even driving the same load and with shared idle current instead of multiplied idle.
Biasing for a larger class-A region pushes the class transition away from where most listening is done, so this can be a good thing.
The point here is to add the output pairs for THD reduction NOT for extra power or current handling. Self showed that this works as hoped and is a reliable improvement. Cordell showed this as well and Leach concurs in his low-TIM amp evolution.
Have fun
Kevin O'Connor
londonpower.com
Last edited: