On Thu, 19 Jul 2007 11:27:55 -0700, John Larkin
wrote:

This circuit is back! I'm getting 92% efficiency converting +12 to -12
at 6.5 watts out, and the regulation loop seems to work. In fact, the
whole thing worked first try. The parts are cheap... the fet is 23
cents and the LM5112 is 50 cents at 1K.

The series gate resistor can be tweaked to essentially snub the small
switching rings, at some cost in efficiency, but then snubbing always
costs efficiency.

Now I've got to explore the control space, stability and load
transients at various loads. It will go discontinuous at light loads,
so the dynamics will change. Control-wise, it behaves sort of like a
hysteretic buck converter.

John



OK, I have the standard dilemma, common with most LDOs, negative
regulators, and many switchers:

It's unstable with ceramic output caps; ESR is too low. Stabilizing
the loop wrecks load transient response.

It's fine with aluminum caps, but aluminums go to hell below 0 degrees
C.

Polymer alums are rare at 12 volts, and have super-low ESRs, so may
oscillate anyhow.

It's great with tantalums, between explosions.

It looks like the compromise is to reduce loop gain about 2:1 (bigger
feedback resistor), and make the output cap from a 47 uF, 25 volt
tantalum paralleled with a 10 uF ceramic, and hope the tantalum
explodes infrequently.

John