Lm3875 + Tip35c/tip36c

[djk]

Driving the booster transistors from the power supply pins is very similar to the gainclone afterburner thingo.

The IC drives the load directly until the 0.65V threshold of the outputs are reached, this part is basically the same.

What is different is output efficency.

At full output the gainclone afterburner thingo wll only get within 4V~5V or so of the power supply rail.

Driving from the power supply pins gives you rail-to-rail output, less the saturation drop of the booster transistors used (usually less than 1V).

One built with 2N3055/2955 powered on ±28V got within 0.3V of the rail with 7A peak collector current.

30% more power was available, although that is only 1.2dB more, but it ran much cooler too. I considered this a big plus, but the powers that be just decided to make the heatsink smaller and pocket the savings.

[Circlotron]

Originally posted by djk
One built with 2N3055/2955 powered on ±28V got within 0.3V of the rail with 7A peak collector current.

Wow!
With the transistors driven from the power supply pins the IC output transistors and the external transistors together become a pair of complementary feedback pairs. Some people say CFP's can be inclined to oscillate on peaks but I think the benefits are well worth the risk.

Maybe you could make a combination of both types...?