Voltage Regulators in Parallel

They shouldn't fight. If mismatched, one will carry all the current upto its inbuilt limit, then the 2nd will kick in. Prepare for that thermally.

They won't fight the way op amps do, because they can't pull down, so the highest-voltage one will provide most of the current. (Their loop gain isn't super super high, so it may droop far enough for others to contribute a bit.

When the hot one hits thermal limit, though, it'll abruptly shut off, transferring the load to the next few in line. That's not a nice way to treat the poor regulator, and depending on the quality of the heat spreader, it may result in complete shutdown of the regulators.

Some power resistors in series before connecting them together would be the ticket, if you have them laying around.

Cheers

Phil Hobbs

Here's an old chestnut:

I have concerns about how this one behaves at temperature extremes, though. Looks fine in the sim on a temp sweep, that is until you pick two different type of series silicon diodes with similar ratings and then it all goes side ways

That's mega-lame. The 7812s will probably have different output voltages (some are spec'd at +-4%), and the diodes aren't good current equalizers.

Why does he parallel two 78L12s when he could use one 7812?

Why would pass transistors need "critical power resistors" at their collectors?

So many bypass caps that are not bypassing much.

At least use LM317s or 1117s to minimize the output voltage differential, and resistors as the equalizers. 1% of 1.25 volts is a lot less than 4% of 12 volts.

There is a simple way to boost a 3T reg with a PNP. Or an NPN! Or make a good reg with an opamp and a giant mosfet.

I have concerns abut it

I'm not convinced it works great at operating temperature unless all the devices are on a common heat sink, maybe not even then.

LM3xx seem easier to parallel, they're basically op amps with an offset and we know how to parallel op amps.

Well, it wasn't the question..

This is pleasingly goofy:

Well, yeah, but... the regulators would take a heatsink to really deliver current.

LM317 and PNP pass transistor can get you from 1A to 5A pretty easily, as long as you need the heatsink anyhow.

is the circuit I'd prefer, other variants (setting R4 and R5 to zero) don't really benefit from the current limit of the LM317.

This is a little better maybe:

While we're being "pleasingly goofy":

Load regulation isn't great, but at leas it doesn't go crazy when you use two different kinds of power diodes and do a temp sweep like the EDN one does..

oops, it works better when you use the right number of diodes:

Naw... "Goofy" is transformerless power supplies when you use a handful of diodes, resistors and caps to take 240VAC and turn it into

5VDC stabilised. Okay, the final "5VDC stabilised" comes courtesy of a linear IC, but the rest of it is discretes. Transformerless keeps costs way down, but you need to know the load parameters within *very* tight tolerances as the source resistance of these suckers is quite high. The lack of galvanic isolation provided can also be concerning. Best to troubleshoot with a low VA mains isolation transformer. ;->

OK, that wins the goofy prize. Why use diodes to merge the two reg outputs?

Looks like someone was trying to maximize the current imbalance.

Paralleling fixed regulators is annoying, since you don't have access to the feedback terminal, and can't connect it to the "real" output after the ballast resistors. So the only thing that can bring a regulator's output into line is current against the ballast resistors. But if you use ballast resistors large enough to ensure tight current sharing you end up with a large voltage drop at max output.

I think the diodes were a hack basically to try to limit the max voltage drop at max output while ensuring there's always _some_ amount of current sharing. vs if you just connect them directly together they're guaranteed to hog.

And in the sim they do share very nicely, at 25 degrees C, with ideal

7812s and ideal diodes, but just start swapping diode types around and it goes sideways over temperature. My goofy idea isn't really much better. The LM7812 is +/- 5% over temperature and the temp sweep doesn't take that into account, you'd have to monte carlo that.

Any solution for fixed regulators has to take into account the output of one regulator could be sitting at 12.5 and the other at 11.5 and be within spec, and none of the simple ideas I've played with can do much with that. You could actively monitor the input current to each and do something with that maybe, but that seems like a waste of time.

The diodes probably make the sharing worse than resistors. Diode behavior is complex, NTC at low currents and PTC at high currents.

Sims share nicely with no sharing parts at all! Just slap the regs in parallel.

That's one reason to use LM317s or 1117s. They will share to millivolts.

Or add opamps.

Those spherical cows don't know how good they've got it, for sure.

Yeah, at some point you might as well build the regulator you actually want. ;)

One op amp, one Darlington or MOS pass transistor, a voltage reference, and a small BJT plus a few resistors to make a foldback current limiter to stay within the SOA and thermal limits.

Cheers

Phil Hobbs

You can parallel 78xxs with a few op amps and transistors, but you have to play the leap frog game:

I've done super low dropout regs with a big n-fet and an opamp, which is great if you have a higher opamp voltage supply somewhere to overdrive the fet gate.

I've seen a few new low-dropout current limiter chips lately, which could go upstream. Or use one of the shunt sensors, like INA281 or similar, to sense current if you really need to.

Yikes. Why not fiddle with the ground pins?