buck converter for negative voltages

I can't recall ever seeing a real negative-to-negative switcher! People sometimes hack a positive controller.

Hey, here's one:

Ajmal Godil suggests using a positive boost converter, with a level-shifting transistor feedback trick. Cute.

I think Ajmal is at Broadcom now.

Yep, good find. Whew, that IC is a bit complicated. LTC also suggests the positive boost-converter hack, in design note DN1022.

if you can figure out a way to flip the feedback or live with the output being relative to input, wouldn't any synchronous buck work?

The feedback is inverted in a *boost* converter, one just needs to translate it down to the negative rail.

To make negative voltages at a watt or two, I often buy an isolated potted dc/dc converter.

They are also handy for making, say, +-5 from a +12 or +24 wart.

For + to -, a Cuk is nice.

What sorts of i/o voltages and power do you need?

afaict it is the same issue, you need to make the feedback relative to ground instead of to the negative rail

Den onsdag den 1. februar 2017 kl. 23.38.27 UTC+1 skrev John Larkin:

if you have +5V this look interesting:

I recall doing push-pull once. I used this Jim Williams AN-70 app note.

The hard part was sourcing the transformer.

LT3580 is a fun chip. It's a flyback that will accept either a negative or a positive feedback voltage.

Yes it will, nice. Too bad it's in such a small package.

Actually, besides level-shifting the feedback to the IC's FB pin, which is referenced to -Vin, it's necessary to invert the sign of the feedback. That becomes clear when comparing buck-converter duty-cycle formula Vout/Vin = D, with a boost converter, Vout/Vin = 1/(1-D). Also, it's obvious that when -Vout increases (more negative), FB must increase (more positive).

The inversion occurs in the level-shift transistor, Q1.

. gnd ----+---------+--------+--------, . ___|____ | _|_ | . | | _|_ --- R1 . | boost | /_\ | | . | | | | | . | SW |----+-#####--+---+------- -Vout . | | B | . | FB |----+----------C E--' . |________| | Q1 pnp . | R2 . | | . -Vin ---+---------+-----

I sometimes put cap multipliers inside the feedback loop of a positive-supply switcher, with an RC highpass to make the loop stable (the way you do with a slow buffer or a big C load). (It helps the load regulation without wasting power on maintaining worst-case headroom all the time.) You can't do that when using a buck in inverting mode, for exactly this reason.

There are shunt references that work upside down, e.g. the LM385 and LM4041-ADJ, but no switchers that I know of.

Using one of those references instead of your PNP (with an RC highpass as above) would get rid of the V_BE error for fairly cheap.

Cheers

Phil Hobbs