Opamp with transistor output linearity

[vossie]

Below is a diagram showing an opamp and two output transistors (Q1 and Q2) connected so that they are driven at their gates from the current-changes over R1 and R3 respectively. Depending on the opamp used a certain level of quiescent current is drawing to internally bias the opamp. This current flows through R1 and R3 from the supply rails and causes a voltage over R1 and R3. the voltage left over R2 and R4 therefore sets the current that will flow through the emitter-collector terminals of the Q1 and Q2. The idea is that when a signal is amplified the current drawn through R1 and R3 changes as a function of the gain/amplification of the signal. This causes a voltage change over R1 and R3 which is used to drive Q1 and Q2. This is a nice way to give an opamp more output current but I'm unsure of how linear this configuration is.

Here is my thinking... With manufacturers omitting more and more of the complete circuit in the data sheets of their opamps it becomes difficult to judge what other internal circuitry (besides the amplification part) of the opamp also draws current during operation. In the case of no other fancy internal circuitry of the opamp used I suspect linearity would be good and the current drawn from the supply rails is a very good/close function of the signal being amplified. However, with additional internal circuitry that could possibly require more current under different operating conditions the changing current drawn from the supply rails during amplification would not be a pure function of the signal being amplified.

This thought bugs me and makes me question how much it would affect the linearity of those two output transistors.

Any thoughts on this?

[richie00boy]

Unless you want to run the op-amp at it's absolute max voltage rating, I would just add a conventional emitter follower buffer and bias network driven from the output. With the configuration you have, the op-amp will see a modulating supply voltage as more current is drawn. OK, PSRR is very high, but it's better to try and avoid problems in the first place. Also, as you point out, it relies on the supply current being influenced linearly by the signal under dynamic conditions, which is not always the case.

As for the transistor linearity, there will be a kink in the curve as the op-amp output is supplemented with the transistors. You may be able to bias the transistors enough to set the kink below normal operating conditions, but that probably means larger op-amp supply resistors.

The method you draw has been used by a number of designers trying to run chip amps right at their max voltage.

[Nelson Pass]

It's a very old technique. It was taught to me by Charlie
Button (I'll bet JC remembers him) and was my transition
from op amps to discrete. I have not looked back.

The circuit works fine, but I blew a couple of them up before
I got the hang of it.

[vossie]

richie00boy - thanks for the reply.

Mr Pass, you actually managed to blow a couple! I've played around with these - I used the LM3875T and always had 2 of them mounted on a 2 ft industrial heatsink (just in case they'd get hot ). Well, numerous times they got very hot, sometimes for minutes on end they had about 25V constant over the output (in bridged mode with 10 Ohm dummy load, +/- 38V supply). At one stage I thought I blew two of them as the new design idea just did not want to work. At this stage I changed them but I've used them again and there is nothing wrong with them!