After building the circuit, I made some modifications. I replaced the current mirror with a j201 wired up as a current limiting diode. This seems to make a decent current source. I've also added a trimmer pot between the gate of the 2n7000 and ground to adjust the bias. When you turn down the gain(which increases negative feedback) the second stage is no longer midpoint biased, but that should be ok because the decrease in gain should prevent any clipping unless you are using something that doesn't really need any boosting to begin with. In simulation, this has a very flat frequency response. I really don't have good equipment to test the real thing, but I'll try to set up a test.