Here is the basic circuit for the amps. Turn the pot in one direction and it acts like a current source. Turn it the other direction and it behaves like a voltage source -- almost, you still have that 0.5 ohm series resistor there, so output impedance will always be at least that.

It's stable on the bench with a fixed R and with real speakers.

The adjustability will allow me to:

1. Tone things down at woofer resonance (you'd have a big peak at Fs if driven totally like a current source)
2. Act somewhat like an adjustable Linkwitz transform for the woofer (around that resonance peak again)
3. Control phase somewhat (if I recall right there is a 90deg phase change between current and voltage driven)
4. Adjust frequency response somewhat, depending on the impedance plot
5. Adjust distortion (less is better but compromises with the above must be made)

I figured I'd do this as a blog rather than a new thread somewhere.

My goals were:
1) Play with the distortion reducing effects of transconductance amps (voltage in - current out.) I think it was likely Nelson (via the F1/F2 papers) that really turned me on to Hawksford's paper that described the benefits of driving voice coils with current rather than voltage.
2) More bass than the 2 ways I built, so 3 way w/ a 10" it will be.
3) Given #1 and #2 building a FirstWatt F1/F2 per driver is a bit much, and given this is an experiment we'll go with chip amps (LM3886/LM4780)
4) I've always wanted to play with a ribbon, despite Zaph's warnings.

I've been working on these slowly for quite a while - so a flurry of posts on the various parts will follow quickly.