Yuck!

U1A is awfully slow (~500Hz?), U1D input is being overdriven, Q1 has weak
drive, yet a huge "speedup" capacitor, and U1D's output is pretty much
short circuited into C3-Q1 and C3-(diode). R3 and R4 hardly do anything.
Transformer polarity is not indicated, so we don't even know if you have
it backwards or not, but it hardly matters because D1 shorts it out anyway
(if not for R2-C2 narrowing the pulse width).

Q1's hFE isn't terribly high, and it's got 2N3904-ish current capacity.
It's no MOSFET. That transformer better have at least 2mH primary
inductance. If not, that would be why you observe constant voltage drop,
it's already in the constant current region.

Offhand, you could use a circuit like this:
http://webpages.charter.net/dawill/t..._Generator.gif
The toroid is probably #43. It should really be gapped, in which case a
much smaller toroid (or more likely, E or U core with coil former) will
work, with many more turns of course.

There are other circuits that would do a better job. A proper flyback
supply is easy to make, using a UC3842 in the appnote circuit, or a
(discrete or IC) BCM flyback. Royer oscillator also comes to mind:
simple, low noise, easy to regulate.

In any case, for a watt or two output, figure a thousand turns or so of
rather fine wire (#40). Something resonant or quasi-resonant will help
mitigate the winding capacitance and leakage inductance inevitable in such
a transformer.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

"Robert Baer" wrote in message
net...
The diode across the E-B junction of Q1 was added, mostly to prevent
zenering and then beta reduction.
D1 is removed for the waveforms seen; it clamps the negative swing
seen at V3 (and the +20V supply current increases with reduced HV
output).
The question is, if the voltage on C5 is changed from about +10V to
+50V, that negative swing DOES NOT change, but the positive swing of Vc
does.
Can anyone explain how and why?