Why two transformers in "SERIES" in an isolated DC-DC switching power supply ?

[BravoV]

From Dave's video #589 Voltech PM300 Power Analyzer Teardown, there is an interesting section of the board that has an unique power supply section that uses two transformers in series as above picture.

Understand the need for an isolated power for other circuit sections, but why two ? Is it because of this configuration gives "much-much better" isolation than one ? Is that it ? or there are other reasons ?

Btw, Dave also wondered why and he didn't give any explanation either. 

Also from the picture, the primary winding of 1st transformer, looks like has the same winding count as the secondary at the 2nd transformer (red wires). While secondary at the 1st has only 2 turns and primary at 2nd has single turn or even half (yellow wires), cmiiw.

First transformer does step-down, and then step up again at 2nd transformer, question is why only so little turns at the intermediary section (yellow wires) ?

Also any further comments on this technique are welcome, I'm all ears.

[T3sl4co1l]

Seems like a resonant converter, step up, probably full wave doubler on the output there for +/-15V or whatever, at bad regulation.  The load is probably pretty constant so that's not a big deal, and is probably cleaned up with some regulators or zeners or something anyway.

Guess it's possible they wanted the extra isolation (and reduced capacitance) with more leakage inductance (makes it resonate) and without having to pile more turns on a single core (which would ruin the creepage...depending on just what kind of wire that is).

These days, you'd use triple insulated wire in a divided coil former for the utmost isolation while meeting UL.  I'm fond of these: http://www.digikey.com/product-detail/en/PH9185.011NLT/553-2052-1-ND/3503429 though pricey, they come with UL "Reinforced" ratings.  The windings are interleaved so the coupling and bandwidth are *very* good.

Tim

[BravoV]

Tim, thanks for the reply, that makes senses on the low turn counts, to increase the creepage factor.

Btw, is this kind of circuit can be easily build using the jelly bean MC34063 chip ? I mean nothing spectacular, only low volt DC like 5 or 3.3 Volt with say < 50 ma "constant" load.

Thinking using this technique for powering the measurement circuit which needs an isolated power from the main circuit it measures.

[Rufus]

Understand the need for an isolated power for other circuit sections, but why two ? Is it because of this configuration gives "much-much better" isolation than one ? Is that it ? or there are other reasons ?

Yes the isolation is improved but wire with a lot more insulation could have been used. They probably use two transformers mostly because it then requires more than a single point failure to become unsafe.

[BravoV]

Yes the isolation is improved but wire with a lot more insulation could have been used. They probably use two transformers mostly because it then requires more than a single point failure to become unsafe.

This is better explanation and makes more sense now, thanks Rufus. 

Learned something new here about safety circuit practice, as usual the devil is always in the details isn't it ? 

[T3sl4co1l]

MC34063  ugly thing...

You could just as well run a 555 at constant frequency, or a couple-transistor multivibrator.  You'll want to adjust it (and the cap on the other side of the transformer(s)) so it's more or less on resonance, and probably shunt regulate the output.  Don't expect it to be too fancy or anything...

On the plus side, you could make a discrete oscillator like they use in CFLs.  It's current mode and tracks resonance; of course, you need two transformers (one for current feedback, the other for the actual work).

You can make a proper resonant controlled thing with variable frequency and all, but I don't see why that would be worthwhile (aside from playing around!) until you're up into 100W+, say.

Tim

[BravoV]

I see, so for this simple no fancy primary driver circuit, it doesn't even need feedback loop from the output isn't it ? I mean for a "constant" low current and voltage load.

Interesting, any hint or simple examples for such circuit will be appreciated.

[T3sl4co1l]

This thing is sort of resonant, I posted it a... few weeks? ago.  Just ignore the fact that it has a regulator anyway.

http://seventransistorlabs.com/Images/HVPower1.png

Ideally, you'd have a cap tuning the primary, which is somewhat loosely coupled to the secondary, which is tuned as well.  The secondary tuning consists of a parallel cap (which sets the unloaded resonant frequency) and a series cap to the load (which, plus the parallel cap, sets the frequency under short circuit conditions).  Kinda-sorta the idea is to make a Tesla coil, but not with a gratuitous voltage ratio; the coupling is tighter than that, which makes it easier to transfer power in a given size.

As it turns out, my transformer didn't have a lot of leakage inductance, so I ended up with a sort of multivibrator with quasi-resonant switching edges, and some squigglies.  I didn't end up using a series capacitor on the output for this reason.  Maybe I still should.  Since this is a high voltage example, the parallel capacitance arises from simple winding capacitance, so I don't have to add anything explicit across the secondary.

In a double tuned circuit like this, you can tweak the parameters (driving frequency, primary resonance, secondary resonance loaded and unloaded, and coupling) to get better regulation, possibly even negative resistance for some range (i.e., a heavier load pulls resonance closer to the operating frequency, delivering more power).  It's not completely helpless, but there's a lot of stuff to tweak (and a lot of potential for drift, and line and load sensitivity).

You could also couple to an over-driven transformer core and do ferroresonant regulation: normally, the winding resonates to an excessively high voltage, causing saturation on the peaks; which reduces the effective inductance, raising the resonant frequency and regulating voltage.

Tim

[BravoV]

Thanks for the circuit Tim, really appreciate it, its just the component counts is abit too high.

But if I'm going to do it, maybe using a custom chip, randomly searched and found something like this TI chip SN6501 looks simple enough, SOT-23 body.