Coupled Inductors--how coupled is coupled?

We've measured that. The short answer is "about 0.99"

Most seem to be bifalar toroids, pretty good coupling.

Here's some notes on a couple of Coiltronics parts:

Notes on leakage inductance:

Part, OCL, SCL, K ( sqrt(1 - SCL/OCL) ) DRQ74-8R2, 8.03uH, 110nH, 0.993 DRQ74-150, 16.91uH, 270nH, 0.991

Measured on AADE LC meter. RG 16-Sept-10

DRQ127-151 150 uH 4 uH DRQ127-331 330 uH 5 uH DRQ127-102 1 mH 70 uH

JL Mar 2012

Looks too small to be anything far from unity - even with a small gap between ferrite bits - 0.998?

The series mode inductance is spec'd to be 1880 (470*4)! ;)

Thanks. The Bourns ones I was looking at are the cheesy "shielded" kind with the big gap at the top of the core + cup, so I was hoping it would be around 0.95-0.98.

Cheers

Phil Hobbs

if the "leakage inductance" is specified or can be measured, then you can calculate the coupling.

The DRQs are like that. I may be wrong about the construction, but coupling seems pretty good.

The ISDN transformers are bifalar toroids, very good coupling. We use one 1:1:2:2 part that is very versatile.

It's sort of an interesting case--very cost sensitive, but has to be tough enough to withstand a monsoon thunderstorm in a cotton mill with very iffy grounding and long runs of thin wire. We're going to bus around +24V, panel ground, and a twisted pair for 200 kb/s RS485 data.

So I'm looking at some combination of MOVs, TVS zeners, and depletion MOSFETs on all four lines, plus fully isolated power and comms.

The first thing I tried was a half-bridge driving two 150 uH double-wound inductors (L1A + L2A in series, L1B + L1B in parallel for the output), followed by bridges, filters, and LDOs. That worked well with no ringing or other nonsense, and managed to get to 5V okay, but it was a bit marginal at the low voltage limit. Using MOSFETs for the bridge on the output helped some, but it was just getting too complicated.

I'll probably just put in one of those 27-cent A&O buck chips and then drive a couple of 1:1 transformers. I can use the unregulated input to power the high side gate, which helps some.

The bad news about regulated switchers is that they have negative input resistance, which may become a problem when the wiring runs get too long. (I can see somebody running the power on a spare CAT6 pair, 20 ohms per 1000 feet.)

Cheers

Phil Hobbs

Plus or minus 20%, of course. ;)

Cheers

Phil Hobbs

Yup. It isn't, though--the inductance in series is 2 L1 + 2 M, and in parallel it's (L1+M)/2. They list the series inductance as 4x the parallel, so k is around unity. The specs are +-20%, so k could be as low as 0.6 and it would still pass.

Cheers

Phil Hobbs

We like to start with a polyfuse (radial lead, not surface mount) followed by a unipolar transzorb. Seems to protect against user blunders, short of 120 VAC on a 24 volt bus.

Lightning can be a nightmare in tropical climes and old buildings. Horror stories about gear in Florida. Protect those RS485 chips, too.

That recently confused our test folks. We have some bench supplies that come up slow, and some products with purchased switcher bricks, and the combo wouldn't start with the supplies set for apparently reasonable current limits. We try to include UVLO and/or soft-start when we design switchers.

Even soft start won't help when the wiring resistance gets too high. A

R_in = 1/(dI_in/dV_in) = 1/(2W/24V - 2W/23V) = -276 ohms

but that drops rapidly as the input voltage declines--it's 66 ohms at

Thus when the wiring resistance gets into the tens of ohms, life starts to get interesting, and of course when it gets above 72 ohms, it can't supply 2W into any load whatsoever.

Cheers

Phil Hobbs

flybuck?

-Lasse

Thanks. Thought of that--I've done it before, but it really only works when the main supply is more heavily loaded than the isolated ones. I need two isolated outputs and zero non-isolated ones.

Also I can't be sure that one supply will always be drawing more current than the other--a double terminated RS485 link can draw over 80 mA, whereas it's much less with AC termination. The SBC can draw 160 mA max, but there's no minimum specified. So it looks like a buck followed by two 1:1 transformers, bridges, and LDOs. _Not_ the elegant solution I was hoping for, but there you go.

Cheers

Phil Hobbs

Belay that--with a _sync_ buck, you can balance the voltseconds even with no load on the non-isolated output--it's a half-bridge.

Cheers

Phil Hobbs

I've also put unipolar TVSes in series with the ground pin of a PolyZen. You keep the super-fast switching action, which is great. You do have to be careful not to exceed the standoff voltage rating of the polyfuse part, of course.

Cheers

Phil Hobbs

The radial polyfuses come in higher voltages than the surface-mount parts. They are kinda more predictable on current limiting, too.

Not PolyZens--they're magic. The zener heats up the polyfuse super fast on any sort of large overload, so you don't have the problem of the diode unsoldering itself before the polyfuse switches. They aren't as fast for reverse voltage, but then you don't care as much since the diode isn't dissipating much.

Cheers

Phil Hobbs

I've measured quite a few types, not your p/n but similar ones. Like John says, they tend to be bifilar wound, and therefore with quite low leakage inductance. I've measured 0.3 to 0.5% of the magnetizing inductance.** So coupling = 0.96 to 0.98. One caution: They have more winding-to-winding capacitance than you'd normally expect. Easy to measure.

Do you have a decent LC meter? Just short one winding and measure the other. TDR will work too, to estimate small inductance.

Sometimes I use a function generator and a scope to measure small inductances. That helps resolve resonances, which confuse some meters.

On Mon, 25 Jan 2016 16:07:24 -0800, John Larkin Gave us:

"coupled inductor" There is no such animal. If there is coupling it immediately becomes and is empirically defined as being a transformer.

Turns count ratio and voltage transformation rules apply.