John Larkin a écrit :
On Sat, 7 Jul 2007 22:13:26 -0500, "Tim Williams"
wrote:

Why does a UC3842, etc. suck, exactly? Wrong sign? (I don't remember if
the 14-pin version has + and - inputs at the pins.) Not cheap enough?

Yup, wrong feedback and gate drive signs, and the current-limit thing
would be damned awkward with a p-fet switch.

6W is a little much for a, say, cap switcher?...

Probably, and it would lose some voltage, too. It would have to be
done with a *lot* of discretes, since the IC charge pump things
couldn't handle the net 24 volts across the chip.

A Cuk converter would be nice, even with the dual inductors, but I
haven't found one that will flip +12 to -12 at this current. I think
the math gets messy.

An unregulated p-p forward converter would be OK, but the transformer
would have to be 1ct to 1.1ct or something like that, custom
magnetics, always a huge hassle.

There are a bazillion buck regulators on the market, and very few
inverters.



Use a buck IC in an inverter configuration with a common gate pfet to
extend the switch voltage compliance to negative voltages. Then twist
feedback adequately.
I did almost that once, with an LT1374, generating both +20V and -20V at
1A from 15V. In my case the feedback loop naturally came from the +20V
output.
Nice, simple and fully protected.


--
Thanks,
Fred.