You have probably already seen the thread at
http://www.designers-guide.org/Forum/YaBB.pl?num=1155668476;start=all. The article mentioned in reply #28 describes the method used by Spectre and is one of the best treatments of loop gain I have seen so far. In reply #12, there is a link to an implementation of Spectre's method in another Spice simulator.
It is generally not a good idea to break the loop because of the difficulties in getting the operating points and the impedances at the breaking point right. However, for a simplified analysis, you can also use the loop gain definitions Gv and Gi according to
http://www.spectrum-soft.com/news/spring97/loopgain.shtm. They will always tell you correctly if a circuit is stable, although you will get different values for phase margin and gain margin than with Spectre's method and the loop gain will also vary depending on the position of the probe, unlike with Spectre's method.
Gv and Gi can be used because they fulfill equation (3) of the article mentioned above. You can find a derivation of this equation in section 4.4 of Bode's book, which is reference 1 of the article. For the circuit of figure 7 of the article, the determinant Δ is
k1+
k3+
Ye+
Yf. For the loop gain defined in the article, we have
x=
k1+
k3. If you do the math, you will find that for Gv, you get
x=
k1+
Ye and for Gi, you get
x=
k1+
Yf. So, Gv gives results close to Spectre's method if
k3 and
Ye are small with respect to
k1 and
Yf, and Gi gives results close to Spectre's method if
k3 and
Yf are small with respect to
k1 and
Ye. In other words, you will get best results if there is no significant backward transmission and the impedance looking forward is much larger than the impedance looking backward (use Gv in this case) or vice versa (use Gi in that case).
For multiple loops, first look if you can find a wire that breaks all loops and perform a stability analysis at this point if it exists. If such a wire does not exist, the rigorous method calls for disabling all loops and then activating them one by one, plotting a Nyquist diagram in each case. This is described in section 8.8 of Bode's book. In Bode's words: "If a circuit is stable when all its tubes have their normal gains, the total number of clockwise and counterclockwise encirclements of the critical point must be equal to each other in the series of Nyquist diagrams for the individual tubes obtained by beginning with all tubes dead and restoring the tubes successively in any order to their normal gains." In practice, if the loops are not dependent on each other, you can often get away with examining each loop separately.