bandgap voltage reference

[JPR]

As to your PS, I would suggest that internal vs external compensation should be selected based upon, among other things, the range (and type!) of capacitance expected at the output, the frequency response of internal nodes that might be used for compensation, and the requirements.

I would tend to agree that internal compensation can provide benefits, particularly since the capacitor at the output is likely to be one of the less robust components in the system. Assurance that things will still somewhat function if the capacitor fails is a good thing! However, the pole associated with the output capacitor MUST be considered when compensating the system, which means that an internally compensated system might be very slow, or very large (due to a large C1, which reduces the frequency of the pole for compensation purposes, and reduces the frequency of the zero to allow for cancellation with the pole at the output), or very complicated (with many poles and zeros, for example, to keep phase change to a minimum in order to maintain stability over a wide range of output capacitance)

In retrospect, I over-reacted to your statement, which seemed to trivialize the compensation of such a circuit, which, in the case of a high gain, low output resistance circuit, can be anything but trivial.

 

[tucura]

An amplifier having 2 high impedance nodes is not suitable for
driving large capacitors. The circuit could be modified by reducing
the high z nodes to one. Try removing R5, diode connect M6,
remove the drain-gate short in M7. Add a Pchannel source
follower (very large W/L) , gate connected to drains of M7/M4,
source to the output and drain to ground. The circuit will
tolerate a much larger capacitor at the output, but there will be
a size limit unless a high enough series resistance is included.

 

[cetc1525]

I think this structure is not very good when VDD changed.