In Fig. 26.41 it says that this is NOT the way to implement gain enhancement (GE)
in an op-amp, however, I've seen papers that use this topology (are they wrong?).
Would you comment?
As indicated in the text (page 889) describing Fig. 26.41 this is not the way to
implement GE unless the added GE amplifiers, with differential inputs/outputs,
employ common-mode feedback (CMFB) to balance their outputs.
Note that no deep thought is needed...when you have fully-differential outputs you
must employ a CMFB circuit (don't let anyone tell you otherwise...yes in SPICE
where the devices are perfectly matched things may work fine without CMFB, see
discussion in the beginning of Ch. 26).
If the papers you've seen employ CMFB for both the added GE amplifiers and the main
amplifier then I don't see a problem.
I prefer to avoid the use of fully differential GE amplifiers because they add two
more CMFB paths (and the associated problems with stability and settling) and don't
provide any additional benefits over the topology using single-ended output GE
amplifiers seen in Fig. 26.42.
Note that power isn't an issue since the added GE amplifiers can be slow and thus
low power amplifier topologies (see Sec. 24.4 starting on page 808 for a discussion
of the requirements of the GE amplifiers).