Try the LM298 half bridge drivers from SGS and for 3.5 A you will need to operate 2 of them in parallel.

Alternately if you convert your final output stages to Mosfets then the circuit will become less complicated than yuor 16 transistor stuff and also cut down on the fanatastic coolers that you will now be using. Mosfet + PWM is the way to go when driving from 8052.

Raghu