My first switcher

The +24 output is easy!

One way to get multiple outputs is to do a forward or flyback converter with a custom transformer with multiple secondaries. The transformer is a nuisance to procure and cross-regulation is mediocre. That makes sense in high volume, maybe.

So, the usual way is a mess of buck switchers and linear regs. +15 might be a linear off +24. For negative outputs, you can

1. Buy little charge pump inverter chips, for modest currents. Cheap, easy, no inductor.

1. Charge pump yourself off the positive switcher

2. Build a proper Cuk converter if you need a lot of negative power.

1. Buy some potted isolated DC/DC converters. For about , you can get a couple of watts of +24 to +-5 volt converter. Not much design challenge.

We have some products where the power supplies occupy half the board.

What JL said, plus -- if your volume is low then you may want to stick with Linear Technology parts. They've never failed me, they're easier to design in than other brands, but they're more expensive than equivalent parts from other brands. As a bonus, LT Spice seems to be pretty damned accurate when you're trying out switching supply circuits with Linear switchers (for some odd reason).

If you're going to be shipping 100000 units, Linear Tech may not be the best choice.

Well geesh you're no fun. I spent some time going over switcher design.. and now I want to apply all those equations I had to derive. I'm not sure the linear are going to cut it... just a tad too much power. (~10V @ 0.5A) I've used those little potted CUI converters as inverters...

I was looking at the LM43600 a bit spendy, but easy to invert.

Grin.. I was thinking we could sell a box with a bunch of Phihong wall warts inside and one power switch.. we'd just have to keep the box nailed shut so no one could look inside.

George H.

Yeah low volume 100's. I like LT stuff too. The LT8471 looks interesting,

~$8 in onesies... not much stock at DK...

George H.

When a potted thing costs a few bucks, it's not sensible to design your own switcher. The parts cost may exceed the price of the brick. Agree, no fun.

We use one 3-t 7805 drop-in replacement blob that's a switcher, cheap. It will also make V+ to -5 of you connect it different.

I spent some time going over switcher design.. and

There are some dual, bipolar warts, and a few 3-out, like +-12 and +5.

There are not many negative-to-negative switcher chips around.

Huh.. I was just looking at those.

How do I get -V from one of those? (I mean there are only three terminal... switching ground and output seems like it will screw up the input.) I was thinking I could use those for +V and a switcher for -V... but if I can just use 4 of those, I'm done. (which is boring.. but it's Friday and I could leave early and buy a beer.)

George H.

Never mind.. I didn't read to the end of my own linked spec sheet. I'm sorry to say I think I'm done... order to digikey. (And I don't get to design a switcher... grumble.)

GH

See example on sheet 4. It works.

Beer can be fun, too.

Do something more exotic. A kilovolt switcher. 250 amps polyphase. 50 MHz GaN.

Grin.. well there is always plenty to do. I've got this digital hairball circuit I've got to get to... I was going to post a question on SEB, maybe this weekend.

George H.

Did you inherit an async logic hairball or something? Throw it away and start fresh.

I did rat's nest logic when I was a kid, but some kind soul took me aside one day and showed me the True Path of clean synchronous state machines.

DEC built zillions of minicomputers, up through early PDP-11s, with hairball logic. They usually worked, but not always. There were famous

I *like* to design the odd asynchronous state machine from time to time. Contrary to what many people think, it's nothing like random hairballs. To design them, one uses state transition tables just like for synchronous logic. The principal difference is that an input transition results in a state transition immediately, and not at the next clock. A clocked flipflop is just a little canned asynchronous state machine, in my view.

Modern logic design strongly discourages the practice, is true.

Sorry, a bit OT, although not as much as some well-known frequent posters here.

Jeroen Belleman

Well, more like an input transition results in a state transition whenever the logic gets around to it -- which may be why people object to such things.

I do the occasional small async design, and sometimes we do async parts of an FPGA, if we can fool the compiler into letting us do it. But it doesn't scale, and a big async design is likely to be trouble.

Async logic for CPUs is a repetitive fad. Wiki calls async design "an active area of research" which it has been for decades.

Use an LM2594 or whatever in that series. They're very easy to use, and list recommended parts values.

You can get negative voltages from a buck by using a multi-winding inductor. This makes +/-5V or 12, or +12/-6V (get -5 using an LDO?) very simple. Preferably use a synchronous converter like TPS54233, or put a minimum load on the main output, otherwise the negative rail won't start up.

The negative winding is phased so that it looks like a flyback supply, but it's not actually freewheeling, it's just the buck's diode-clamping phase. This is why you get a ratio to the main output voltage. (The opposite phase is supply dependent, and has forward converter action, so you'd need a choke input filter anyway.)

Afraid I don't have a drawing or appnote handy, but I think I've seen such examples in LT datasheets. LT parts are fine too, you're just paying more for them...

Tim

Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Hi all, Well I get to do a SMPS. There are boat loads of IC's out there and I'm looking for recommendations. (I better go re read AoE3) I need a number of output voltages. +15V, +5V, -5V, -15V. and +24V. maximum current of 0.5A on any one. My current plan is to run the whole thing from a 24V wall wart or brick. Price and size is not too much of an issue, I guess fewer external parts would be nice... Say with the switching transistor inside the IC.

TIA George H.

You can even regulate the negative output, see...

OldStyleBuckSwitcherWithAddedNegativeOutput.pdf

on the S.E.D/Schematics Page of my website.

That's what I was doing ~1980, using LM339's as the controller chip ;-)

...Jim Thompson

Eww-w-w-ww! I'd rather use an MC34063!

You did say "old" ;-)

Tim

MC34063 didn't exist, and some of us knew how to roll our own controller loops >:-} ...Jim Thompson

John, can you supply a link to those? We've been using a ROHM switcher in a TO-220 but cheap they aren't.