Coupled Inductors--how coupled is coupled?

Of course, hence the quotation marks. Distributors seem to restrict the noble term "transformer" to things that work at 50/60 Hz, or at most audio frequencies.

Cheers

Phil Hobbs

On Mon, 25 Jan 2016 20:01:03 -0500, Phil Hobbs Gave us:

snip

So do RF guys with antenna baluns.

The sync buck half bridge looks like a winner. I should be able to use a couple of those 27-cent Bourns coupled inductors and a $1 Richtek RT7272 to make two isolated 5V outputs. As a bonus, putting capacitors in series with the primaries roughly halves the effective V_in, which lets the buck's duty cycle be longer, which improves efficiency.

The cross-regulation isn't great unless the coupling coefficient is at least 0.995, though, so I'll need some scheme for feeding back from whichever output sags the most. The usual scheme for one output is to use a TL431 with the LED of an optocoupler in series with its anode, with feedback taken from the phototransistor. I could use two of those and diode-OR them into the feedback, I suppose, but depending on cost, it's probably better to use some of those toroids of John's to get better cross-regulation.

Thanks, Lasse, for prodding me that direction.

Cheers

Phil Hobbs

The Versa-Pac multi-winding inductors are cool, but not cheap, $3 range. We're lately paying around $3 for the 1:1:2:2 ISDN transformer; maybe ISDN is going out of fashion. Maybe one could take the feedback off an AC winding and get pretty good regulation.

Coilcraft makes some really cool planar-winding transformers.

I guess one might build a transformer into a PCB, with holes for the legs of a core. Cheap, but lots of engineering.

You could buy a CUI dual-out surface-mount isolated converter for around $4.

you are welcome

this might give a few hints,

-Lasse

Newark has those on closeout sale right now; that particular value is $0.19 each.

Yeah, but they're ratiometric, which won't fly in this case because of the power supply wiring resistance.

Bummer that the toroids are getting so expensive. We need this gizmo to have a long production life, so something more generic is good.

I'll play with the diode-OR trick. I'm not too worried about positive excursions because there'll be an LDO on each.

Cheers

Phil Hobbs

I looked at that feedback scheme, but it requires a pretty good transformer (k > 0.995 or so, < 1 ohm Rs) in order to get decent load regulation. It falls apart with k=0.985 to 0.99--the outputs sag but the voltage on the cap stays still.

I ordered a bunch of inductors, so we'll see. (Newark has them on super sale at the moment.)

Cheers

Phil Hobbs

I saw that, and ordered 50 of them, which in my line of work is probably a career's worth. I've recently done the same with a lot of TO92 and DIP packaged parts. It's probably a bit under $1k in extra inventory, but I can do dead bug protos for the foreseeable future.

Cheers

Phil Hobbs

Sounds like you could use a flyback, maybe regulate off the primary side. But that would almost certainly involve a custom transformer, which is always a nuisance.

Does closeout = EOL?

Get yourself an inductor selection kit, most suppliers will then top it up for free as you use parts.

That's where I started on this thread. The sync-buck "flybuck" converter is basically a half-bridge version of a flyback, with the voltseconds zeroed out by a series cap (which would be the output filter cap on a normal buck). (You could also think of it as an AC-coupled forward converter.)

The thing I was initially missing is that the sync buck allows the isolated supply to be more heavily loaded than the main buck (which needn't be loaded at all, in fact).

Normal switch/diode bucks don't work like that, because there's no way to sink current from the output cap, so the isolated supplies are really only good for light-duty auxiliary power. I've done photodiode bias supplies that way, for instance, piggybacking off the free AC from an LM2594.

Cheers

Phil Hobbs

Thanks. I got one from one of Mike Engelhardt's seminars awhile back. They aren't the cheap kind, though.

Cheers

Phil Hobbs

No sign of that on the Bourns website, and DK has lots.

Cheers

Phil Hobbs

I think it means Farnell (England) is dropping that product line, that's all.

Made some measurements with my hp 4192 LCR meter. DCR 5.00 ohms both sides (spec 5.64 ohms) L (100kHz) 480uH [each or parallel] (spec 470uH) L_series (100kHz) 1.97mH (spec 1.88mH) Lell (any freq) 6.6uH (1.4%, very good) SRF 550kHz (poor, but OK for 100kHz use)

These will do well, even with unmatched loads.

Thanks, Win. So k~= 0.986 then.

Looks like I can mostly get rid of the line regulation sag by putting a resistor from V_in to the feedback point, which makes it practical to use the flybuck topology with an LDO per output. At that level it does need a lightweight snubber to protect the Schottky rectifier, but it doesn't cost much power.

Probably when I've done this a few times I can make it simpler, but this seems pretty good.

Cheers

Phil Hobbs

Are you asking for a way to measure mutual inductance, i.e., how to measure the coupling coefficient ?