In the system with negative feedback, what is described by G(s) actually? is it G(s)=A(s)/Y(s) ?
If I want to derive Y(s)/X(s), is it a correct way to derive G(s)=A(s)/Y(s) and then I can plug it in to the equation (8.2)?
With simple resistor as feedback, I usually derive the transfer function directly.
But right now, I have a quite complicated feedback network with some frequency-dependent components.
That's why I want to derive G(s) first.
Yes, you treat G(s) as a stand alone network with an input
and output. But dont use Y(s) as input because that would make
it a f( G(s), X(s), Y(s) ). So make input Z(s), output A(s), and then
solve for is T(s) = A(s) / Z(s) and plug that into the general T(s)
for the overall problem.
In this case, G(s) will be described by a rational function with numerator and denominator coefficients (zeros and poles). Consider that the "detached" feedback network must be loaded similarly to the final circuit.
As additional remark, I'd suggest an extended feedback scheme with separate forward transfer function F(s). It gives more options to design physical plausible feedback amplifier models.