This is very close, but you have made some errors in signs
somewhere along the way.
As an aid to those who actually do all the algebra, if you set
the value of all the components and the amplifier gain to unity, the
roots of the denominator are: -2.61803, -1 and -.381966
Sure enough... inspection of the circuit reveals that the gain at infinite
frequency (short the caps) should have been -k. I did indeed miss a sign,
resulting in several sign flips by the end. Try this:
Vo -ks^3c1c2c3
-- = -------------------------------------------------
Vi sc3(sc3+kg2)(sc1+sc2+g1) + s^2c2^2(sc3+g3) -
(sc3+g3)(sc2+sc3+g2)(sc1+sc2+g1)
I approached this in almost the same way Jim Thompson did (except that I
used G instead of R), and getting the equation in the form I posted took
around 15 minutes. Checking the answer (which I didn't do until after you
pointed out the error) took a couple minutes, and finding the sign error
took another five minutes or so. The analysis took ten equations, showing
all the steps, and would have fit on one page if I hadn't drawn the
schematic at the top of the page. Putting the denominator in standard
polynomial form would have taken an extra couple of equations.
It doesn't seem particularly ugly, especially since it's easy to check the
resulting transfer function at zero and infinity.
-- Mike --