Flyback vs half-bridge

Do you have a decent Spice model for the ISDN transformer? ...Jim Thompson

(a) ... yes. (b) ... why? What happens? Changing to another architecture while using the same kind of loop usually doesn't do much to improve stability.

Not sure what you mean here, but usually current mode control is the way to avoid asymmetrical runaway.

ISDN transformers don't have much air gap, and you need air gap for a flyback. Plus ISDN is on the way out in many areas so if this has to remain in production until the cows come home I wouldn't.

John Larkin has recently used these tiny flyback transformers. Nothing wrong with hanging two in parallel on the primary and using the 2nd one with the secondary flipped around, for the -45V.

Do you need this isolated? Else you could consider just inverting and boost.

It's 2.2 mH CT : 8.8 mH CT.

On the 9 mH side, with the 2 mH side shorted, my trusty Heathkit HD1250 dip meter shows it resonating at 2.1 MHz with 4.9 nF in parallel, which is 1.2 uH. That makes the coupling coefficient about

Cheers

Phil Hobbs

Close enough for government work >:-} ...Jim Thompson

The ISDNs work to pretty low frequencies, so a high frequency driver and a small cap might fix the saturation problem.

I recently saw a chip with an oscillator, flipflop, and two open-drain outputs. You could drive a center-tapped primary winding with one of those. The ISDN trannies caome in all sorts of ratios, like 1:1:2:2 for example.

Hey, I've got the same one. Since childhood.

Careful. You've got tons of winding capacitance in there which can seriously fool one into believing a much too small number. You have to measure the leakage inductance at frequency-of-interest, using an impedance analyzer or do it in a more pedestrian fashion via generator and meter.

I thought in SPICE it is "k" as in "koupling koefficient".

if it is CT push-pull? ir2153 would work for that too, just skip the bootstrap

auir2085 is simlar to ir2153, max 100 volt but faster, programmable dead time, more drive and has over current protection

does it need to be isolated?

-Lasse

The IRS2153D puts a square wave out of a half-bridge--no feedback, no current limit, nada. If you put that into a transformer via a cap, all is well until you saturate the transformer.

Say that happens on the positive half-cycle. At that moment, the voltage on the cap rapidly goes from V_DD/2 to V_DD. At the next edge, the voltage across the transformer is suddenly not V_DD/2 as expected, but V_DD. The transformer saturates in half the time it took previously, and the voltage on the cap goes from V_DD to 0. Then the cycle repeats. It's really obvious on a scope when this happens, and it's far from pretty.

On each half cycle, roughly 0.5 C_coup*V_DD**2 gets dissipated in one of the FETs, which gradually cook themselves.

So I'm discovering. The actual amount of energy I can store in that toroid isn't very large, so to get any power out of it I have to run it pretty fast. With 2.2 mH of primary inductance, that takes quite a bit of voltage, which makes the whole flyback thing sort of moot.

Doesn't have to be isolated, but it does need to be quiet, hence the toroids.

Thanks

Phil Hobbs

The flyback idea was sort of appealing, because it lends itself to voltage feedback, but as Joerg says, the core won't store enough energy without a gap, which sort of defeats the purpose.

This is for another laser noise canceller design, which needs a lot of reverse bias and ought to work out to at least 5 mA per photodiode. There's a current mirror on each one, so that's 10 mA on the signal channel (+40V) and as much as 20 mA on the comparison channel (-40V). (I may need to water-cool the photodiodes.) ;)

It looks like some Royer-style thing might work better, since that automatically fixes the saturation problem. I don't have a spare winding, but I could probably use a voltage divider off the secondary.

Cheers

Phil Hobbs

LT3439 ?

-Lasse

I sort of doubt that there are nanofarads of distributed capacitance in that tiny transformer, though, because otherwise its open-circuit resonance would be way too low. It works as a transformer up to well over 1 MHz, so at 9 mH, the winding capacitance has to be down in the lowish picofarads.

Ja, kamerad.

;)

Cheers

Phil Hobbs

What about ADSL line transformers? They'll probably be around for a while.

Their inductance is somewhat less than the ISDN types, though.

You'd still need a storage inductor if you go for a forward like topology.

Take a look at SN6501 maybe.

With fiber around the corner? Shudder... I'd rather use the flexible transformers which are sold by Coilcraft and Wurth.

Interesting, thanks. Their coupling coefficient looks quite a bit lower--maybe they have a gapped core?

Cheers

Phil Hobbs

Why doesn't that happen in 60 Hz line transformers? If they get shut off at the wrong time, the core is left partly magnetized. At next turnon, there can be a huge 1/2 cycle surge, enough to make the wires in the wall jump sometimes. But it quickly self-corrects.

The 60-Hz line is DC coupled to ground, so there's no capacitor to remember the misbehaviour between half-cycles. It's the gift that keeps on giving. ;)

Cheers

Phil Hobbs

It gives you an extra pi/2 (1.571) volts, at the price of a second inductor. You may know the circuit from Jim Williams Linear technology application notes AN45, AN49, AN51, AN55, AN61, and AN65.

If you use MOSFETs to do the switching, you are reputed to avoid the evil "squegging" referred to in a footnote in the original paper (at the bottom of page 752).