Nigel Goodwin said:
It's really the inductor and diode that make it a switching supply, and the supply itself is an oscillator - a linear supply shouldn't oscillate.
OK, so we have a design for a voltage regulator that is not working. The OP didn't say whether he wanted it to be a switching or linear regulator. It is missing parts to be a linear regulator, and it is also missing parts (a fewer number, actually) to be a switcher.
So, it is neither, but was arguably closer to a switcher. The original design probably oscillated or was easily put into oscillation because of the high gain for UA2. I'll bet that if one had put a scope on the original design (say at the base to the 2N3904 or Darlington), you would have seen variable period, variable frequency switching by UA2.
I responded to the OP's original question, because his design was very similar to something I built a couple of months ago using a comparator. I intended my design as a switcher with a very large capacitor, but no inductor, and my results were similar to the OP's. That is, it does regulate without the inductor and diode, until you try to draw significant current from it. I even have the burned finger tips to prove what happens when I applied a real load.
My conclusion from that misadventure was that the inductor is necessary. That's not a novel conclusion for sure, but the experience did increase my understanding. I still don't fully appreciate why a large capacitor can't "store" the energy; whereas, it is stated that the inductor does. In my mind, I justified the inductor on the basis that it also acts as a throttle on the current being handled by the driving transistor when it turns on. As soon as I added an inductor and diode, everything cooled off, and I could draw current. It was still variable frequency and period, though, as expected.
Coming back to the OP's original post, why not just add an inductor and diode, quit messing with getting the op-amp just right for operating the 2N3904 and Darlington in their linear regions (NB, there is no current limiter on the output yet), and call it switcher?
Nigel, Thanks especially for the clarification. I spent way too much time this morning trying to figure out what I was missing in addition to the diode/inductor that was so obvious to everyone else. I truly thought he was trying to make a switcher and had not included them, just as I had tried that route.
John