isolated synchronous flyback topology idea

I thought you were about to say, use it to predict and optimize timing, but then it took a strange turn...

An adaptive control would be fine with an average current control method, where the waveform is guaranteed to change at a certain rate; not so much for peak current mode (where the timing may be consistent, but cannot be guaranteed).

Dedicated sync rect controllers exist, probably for less cost.

Simplest is just running the gate from the secondary, which can't be done with every topology but is useful enough to remember.

The SMPS controller itself can do sync rect drive, if transformer coupling (for the primary side switch or secondary side rect, as the case may be) is acceptable (maybe not, down at this cost level).

Tim

Sync rectifier controllers with low external parts count for medium-power flybacks seem a bit thin on the ground; LT makes a couple with LT-style pricing, naturally. They usually don't include isolation for the secondary switch drive either, if you want that you need to transformer couple or optoisolate yourself anyway.

Yes, you'd naturally use one or both uPs to optimize timing somehow algorithmically. I was thinking instead of using analog feedback loops that will require isolation anyway why not just digitize everything you need and communicate across the isolation barrier that way.

Hmm, thought I had a board around here somewhere, with such a controller onboard. Think it was Meanwell or Sola, industrial 100W DIN mount sort of deal. Can't find it though.

Did you check ON Semi? TI? Uh, Rohm maybe? I'm sure there's Chinese ones out there too, though good luck accessing/utilizing that market unless you're a big OEM.

Ah, TEA1761T is one from NXP. Oh that's weird and unhelpful, Mouser lists it as a PFC controller. Digikey lists it as "Secondary-Side Controller", and other chips from all over (Taiwan Semi, Diodes, NXP, etc.) starting at less than a buck in singles.

No isolation or drive needed, just tack on a MOSFET (and whatever support components) and you get a miraculously-well-performing diode. (If you need bidirectional current flow for a regenerative sort of system, you probably need, and maybe can afford, the added expense of a more canonical drivers-and-isolators design with a centralized controller.)

Well, start simple. What do you already know on the secondary side? If you're doing sensing/regulation (with a TL431 or what have you), you already have the setpoint signal handy. Make the opto + primary side controller an accurate transconductance or PWM stage, and that setpoint is reliable enough to use as a feedforward (if not directly usable for nanosecond-accurate timing, though).

Would look kind of odd, putting in extra op-amps and one of those matched photodiode optoisolators, just to make a power supply. But, eh, whatever suits you; maybe there'd be some advantage for a particular niche, like an accurate bench supply, without needing an aux supply somehow, or something.

Tim

The controller IC I ordered to play around with (I'd like to learn more about designing "regular" flybacks not just off-the-wall ideas) is the NCP1230, they're 10 for $10 from HK in a 7 pin DIP:

it's not intrinsically a turnkey synchronous solution though it also needs a companion secondary side chip but hey it is cheap.

One more question while you're here, I have several of these transformers I'd like to use/destroy in my experiments, they appear to be flyback trannies but I can't find any data on 'em. Same appearance and part number as this:

Any idea of ballpark specs? They have the same DCR on both sides about

Hi

Flyback synchronous rectification is difficult for many reasons

You don?t know the length of the discharge cycle and a single upset can destroy the primary switch (emc, noise etc)

Best is to be sure it will be continuous mode, then you can drive the switc h from the primary side, like a forward converter

But the micro controller needs to be blazingly fast, best is to let the mic ro wrap around an analog loop, but fast excursions are problematic

You can drive it from the winding itself, self driven

But if you are down that route then better select a different topology whic h lends itself better to synchronous rectification and gives you better eff iciency and smaller xformer

Cheers

Klaus

Looks like a CMC.

Would guess around 5mH, 5A rating, k ~= 0.98, 300uVs flux?

CMCs are usually good for a few watts in a forward converter configuration, more or less independent of size. This is because LL limits power output at high frequencies, so much that you're forced to frequencies near saturation, also limiting output voltage. A resonant topology could cancel LL, giving normalish power levels.

Rewound so that primary and secondary alternate sections, or cut out the dividers and wind a regular (full layers) winding, they're probably good for more like 50, maybe 100W. Gap the core and you can do flyback and such.

Tim

I've been beaten to market in a sense, though this product appears to use some kind of inductive coupling internal to the IC for is digital data link across the isolation

ADUM series magnetic isolators have been around for a decade-ish, and several others make similar or equivalent products. For example, TI's use capacitive coupling. Not all products have the same reliability (the output state is usually latched, so the output is sensitive to noise, in a very dangerous sort of way -- AD's parts send refresh / keep-alive pulses to help with this), but the immunity is usually very good (>20kV/us at 1kV or such) so you might never notice.

They're not usually very cheap, but it's been a while since I surveyed things. Maybe they're quite cheap now.

Tim

de.

I once put together a scheme for getting a very accurate DC unipolar voltag e across an isolation barrier for most of the time.

It depended on the idea that two identical windings on the same core will s ee the same induced voltage as the flux in the common core changes.

I used a three-winding transformer with two identical windings for voltage sensing, and one high current winding to control the flux.

On the primary side, and op amp steadily increased the current through the high current winding to give the desired voltage across the primary side se nse winding, until the current hit it's upper limit, when the drive was swi tched to forcing current down to a negative current of the same mangitude, after which the cycle resumed.

On the secondary side, the negative edge switched the isolated output from coming from the isolated secondary to coming from a sample and hold circuit , and left it there (for some 3msec) until the current through the driving coil was ramping up again.

Messy, but cheap and remarkably accurate. If I'd used something better than a uA748 on the driving side, whose 1mV offset was the entire error budget, I could have done better.

Ratio transformers are accurate to about one part in 10^7, which would have been 0.5uV in 5V. I was isolating a 1V to 5V signal.

If you threw in a single opto-isolator, you could isolated a -10V to +10V r ange.

If I remember rightly, the ramp time was about 3msec. Lunatic perfectionist s might switch between two cores driven in alternation and throw out the sa mple and hold.

side.

age across an isolation barrier for most of the time.

see the same induced voltage as the flux in the common core changes.

e sensing, and one high current winding to control the flux.

e high current winding to give the desired voltage across the primary side sense winding, until the current hit it's upper limit, when the drive was s witched to forcing current down to a negative current of the same mangitude , after which the cycle resumed.

m coming from the isolated secondary to coming from a sample and hold circu it, and left it there (for some 3msec) until the current through the drivin g coil was ramping up again.

an a uA748 on the driving side, whose 1mV offset was the entire error budge t, I could have done better.

ve been 0.5uV in 5V. I was isolating a 1V to 5V signal.

range.

sts might switch between two cores driven in alternation and throw out the sample and hold.

Come to think of it, what I started off looking at was a three-winding tran sformer with switches to make it a full wave bridge.

There are now analog switches which have low enough on-resistance to work i n the application, and can stand off high enough voltage when off. There we ren't back then (1975), or a least nothing that was cheap enough to be wort h thinking about.

The output does have to hold while the windings are switched around, and th e inter-winding capacitances charged up (which would take at least a few mi croseconds) so you are stuck with two optoisolators to tell the secondary s ide what the primary side is up to.

With two cores and six windings you can cut that hold period down to whatev er your switches can manage (which will be less) and the arrangement for dr iving the high-current winding can be simpler.