The romanblack circuit has gone under some additional study here. With the component values that he as shown on his web site the I have some doubt that the circuit can be any where near as efficient as he has claimed it to be. Maybe he measures efficiency with a load of a few milliamperes or something.

In the simulation I performed I placed a 10 ohm load on the output to get a 500mA load. I replaced the transistors with some that can handle currents like that and set the input at 50volts. (His web page claims that the deisgn will work for wide range like that).

I also changed the bias resistor for the zener diode to 22K becasue of the increase of the input voltage. The simulation also has a 6.2 volt zener instead of a 5.6 volt zener which makes the output voltage be higher at 5.7V.

The simulation predicts an average input current to the circuit of about 0.2A.

This means that efficiency (i.e. output power / input power) is:

(5.7V * 0.57A)/(50V * 0.2A) = 32.49% efficient.

:: lousey

Michael Karas