Programmable tracking pulse supply.

Looks good, though I'm nobody's iidea of a switcher expert.

I usually use the old 150-kHz Simple Switcher bucks for inverters, with tor oids. They're less efficient (especially generating 50V from 5V!) but their nice squishy edges make them pretty quiet for SMPSes, which is usually mor e important to me. They can easily share a board with an ultrasensitive fro nt end.

Cheers

Phil Hobbs

NSC / TI promotes the trick of tying the switcher IC's ground pin to the negative output of the inverting converter. But with that scheme it's harder to have ground-based programming voltage. I prefer keeping the IC's ground pin at ground.

If the average current is very low it should work but one of the voltages (in your case the positive one) won't be accurate.

IME they do not match that well even with a K0.99+ inductor. The diode is the problem, making it soft. Then there (possibly) is the issue of uneven loads. But how can they even get a trademark on this? That kind of architecture has been around since at least the MC34063. The days the Beatles were popular.

If you can tolerate a huge mismatch (north of 10%) at the low end of the output voltage range the architecture should be ok. Otherwise I'd prefer an inverter for negative and a SEPIC for positive, and control them together. That would result in much better accuracy but, of course, also more complexity and cost.

Den tirsdag den 19. maj 2015 kl. 15.13.08 UTC+2 skrev Winfield Hill:

LT8471 ?

-Lasse

It's a trademark registered for a specific name, not a configuration. I agree, it could use a name (I'm not aware of one, are you?), but "fly-buck" is so bad, to my mind, I'm not sure I'd start using it, even if they hadn't trademarked it.

The loads are quite well-matched. I'm hoping for 5% voltage matching, but a little experimenting will tell a lot.

Nice topology. It's sort of a reduced dual-output flyback.

To get a bit of negative voltage, I've either

1. Added a voltage doubler thing to the switch node of a buck converter. Some switchers choke (peak-current shutdown) (no pun intended) if you do that; a smallish series resistor helps

1. Add a second winding to a buck switcher inductor, and rectify that. Works but lacks voltage symmetry, and can also upset the switcher chip. That's good to make a bit of auxiliary negative voltage.

Here's an inverting switcher. I could have used a dual inductor and made it into your configuration, but I had the +12 that I needed, and only wanted to make -12.

Certain Parties around here trash this circuit as noisy (which it is) without fully appreciating its weirdness. CP prefers a more conventional Cuk.

I guess one could do a combined Cuk+Sepic. Have to think about that.

Absolutely ancient (like ~40 years ago :-), free-running, noisy, but works...

...Jim Thompson

Den tirsdag den 19. maj 2015 kl. 17.35.59 UTC+2 skrev John Larkin:

-Lasse

I'm designing a programmable tracking pulse supply. This is an SMPS, with dual tracking positive and negative outputs. It must run from a 5 to 24V input, and deliver +/-5 to +/-50 volt outputs. A 0 to 5V input signal programs the output voltage. It's a "pulse supply," because it must deliver up to 3A, using a pair of large output capacitors.

I've chosen the inverting switching mode because it can deliver output voltages from less-than to more-than the input. It's like a buck-boost, but without the multiple-switch complexity. My hope is to get a reasonable tracking output from a 2nd winding on the inductor (I've found a transformer with very low leakage inductance).

TI recently named this configuration "fly-buck", and trademarked the name: Ugly, times two. They promote the LM5017 for this application. It's a synchronous converter (normally good), but using it means there's an extra diode in the positive output path. If I use a non-synchronous part, both outputs will be matched at one diode drop.

I'm attracted to buck-style converter ICs with external p-channel MOSFET switches, allowing for arbitrarily-high negative output voltages. Table 9.6, page 658 in AoE-III, has a handy table of candidate parts (see free copy here).

Parts with 1P in the table's Ext switch column use p-channel switches. There are four such parts, plus LTC1772 and LTC3801, for which the ADP1864 is a second-source. All of these parts have adjustable input-current limits, which is good, because the huge output caps will be a severe load until they're charged. Note: Even tho the buck controller IC may be able to operate at 100% duty cycle, it must have a reliable shutoff scheme for each cycle.

I'm leaning toward ON Semi's NCV8852, with a Vishay LPT-4545-101 inductor / transformer.

Here's my opening sketch.

IC, don't feel bad, normal rookie mistake.

We need to buy Win a good book on electronic design, something to get him started.

Thanks for your good suggestions.

Actually, I have thought about it and have a plan, but simplified my drawing not to show any of that. A resistor from the opamp to the FB pin, plus the required compensation components should work fine.

I thought about using the pos output for feedback control, but it felt awkward, being removed two steps from the action. What's really important, it'd force a negative supply for my programming. Either scheme requires an opamp (sot-23), but eliminating a negative supply is good.

Ah, nice part, I didn't know about that one.

Thanks for your good suggestions.

Actually, I have thought about it and have a plan, but simplified my drawing not to show any of that. A resistor from the opamp to the FB pin, plus the required compensation components should work fine.

I thought about using the pos output for feedback control, but it felt awkward, being removed two steps from the action. What's really important, it'd force a negative supply for my programming. Either scheme requires an opamp (sot-23), but eliminating a negative supply is good.

That's an interesting app note, I hadn't seen it before. Thanks very much!

That's an interesting app note, I hadn't seen it before. Thanks very much!

-- Thanks, - Win

That app note is nice but as it points out the positive (SEPIC) output current is discontinuous and requires added filtering. For extra credit Win, let see you add current steering to that output and circulate the ripple current to ground.

I could not resist, Harry D.

Harry, will you show us what you mean if Win doesn't? (It was a nice app note, I like how it gets the most out of one switch.)

George H.

Yeah, ahem, "add current steering to that output and circulate the ripple current to ground", that sounds too much like work. As for me, I'm having a Bushmills.

Indeed, a single switch and a single controller feedback path, from the positive output. Doesn't that raise the same objection that was made to my version: the other side is not tightly controlled?

Arrgh! Leessee, the Inverse SEPIC, or zeta converter topology accomplishes that goal, moving the ripple from the output capacitor to the input capacitor. But it also requires moving the switch, which breaks the cute single-switch scheme of the SEPIC + Cuk design. You have something else in mind, I imagine.