
The two inverters on the left generate a square wave and the other four inverters are connected in parallel to increase the current drive to the diodes. The diode on top clamps the voltage on the top capacitor at about 4.5 volts when the inverters go low. When the inverters go high, their output voltage is added to the 4.5 volts to give about 9.5 volts. The second diode rectifies this voltage to give a little over 9 volts on the output. The bottom two diodes work in the same way only the voltage on the first capacitor is clamped to about 0.5 volts on the positive swing and then goes down to about -4.5 volts on the negative swing of the inverters, giving about -4 volts out.
The prototype is operating at only 500 Hz to allow for the use of some old-fashioned germanium rectifiers that I have in large numbers. If more modern schottky rectifiers are used the frequency may be set higher by lowering the .001uF capacitor or the 1 megohm resistors. The 4, 330 uF capacitors are larger than necessary and a few uF will suffice if the frequency is raised to, say, 5 kHz (try 100k resistors or a 100pF capacitor). Yep, I have a lot of those 330 uF capacitors, too. In fact, I have a few thousand of the CD4049, if you would like a few. (charles@wenzel.com)

This little circuit is going into a sub-picoampere leakage meter for characterizing JFETs and other components for extremely high impedance circuits.