DaveC writes:

Your drawing is wrong; signal goes to the inverting (-) inputs and the
Vcc/2 reference goes to the non-inverting (+) inputs.
JF

Thanks guys. Fixed:

http://i44.tinypic.com/r1k8qa.jpg

All else looks good?

You still have the signal go to the non-inverting input. The way the
schematic is, U1-U3 will throw their output hard to a rail or oscillate
with positive feedback from R13-R15.

Are cap values reasonable? I added C8 & C9 out of habit of seeing in other
designs. Values for these?

I worry that the RC time constant would have the reference be not at the
1/2 way point while C8 charges on powerup. I don't see a C9. Might do
something not so good to the subwoofer.