So as far as I understand, the basic idea of a linear regulator is to use a series-pass element along with an opamp with negative feedback to stabilize the output. This is what I usually see as an example:
simulate this circuit – Schematic created using CircuitLab
Now, to use a MOSFET in place of the BJT, can I just do a direct replacement like this:
Or should the gate of the MOSFET be driven by another BJT like below?
Or does it not make any difference? Is one more stable than the other?
Also I remember reading somewhere on the internet that in order for the opamp to regulate correctly, a small current should flow out of the output through the negative feedback loop, but a MOSFET gate doesn't allow any current in, which is also confusing.
Best Answer
It is quite possible that none will work effectively without a compensation network applied directly to the op-amp. Whether it's an op-amp supplying a load directly or via a buffer transistor, adding bulk capacitance to the voltage output will shift the phase margin towards the point of instability: -
The object of the above modification is to reduce the gain of the op-amp at higher frequencies so that the excessive phase shift brought about by the bulk output capacitance won't act on a part of the spectrum where the loop-gain is greater than unity. You might have noticed that some regulators do state that they have a maximum capacitance that can be added to their output - this is the same issue - they have internal compensation that counters "so-much" bulk capacitance but no-more.
Regards driving the MOSFET, because it is a source follower, it's gate-source capacitance won't be fully seen at the op-amp output and it might be OK stability wise but the bulk output capacitance is still present and is still a problem (see above). There's no real reason to use circuit 3 given what I've said above and, if you did you would need to add an emitter resistor to ground.
Simulators are a good friend for fixing and uncovering the problems mentioned above.