Is this the reason there's so few 2 Transistor Fwd Sync Rect controllers?

This highlights an issue with the LT1681/LTC1698 Two Transistor Forward Synchronous Rectifier controller....


I have the LTspice sim working now as attached.

However, at 1.7301ms, the simulation shows the great danger of using this kind of synchronous rectifier controller...

The LTC1698 has no fast inductor current reversal detection and also, it allows the sync fets to stay on for more than one switching period. Due to this, when the 2 transistor forward is suddenly switched from full load to no load, the inductor current subsequently reverses as the “power” sync fet switches on. The “power” sync fet is shown to stay on for 3.5 switching periods….this should never happen, but LTC1698 allows it to happen. When the power sync fet turns off at 1.7301ms, an enormous voltage spike erupts across the VDS of the “power” sync fet. This is because following the switch off of the “power” sync fet in which reversed inductor current is flowing, the inductor current has nowhere to flow….hence the massive overvoltage.

This is why these type of Synchronous fet controllers should have a severe health warning on their datasheet.

At the very least…these type of sync fet controllers should always be implemented with suitable TVS’s across the sync fets……however, the TVS capacitance is not welcome as it will ring with the secondary referred leakage inductance.

Really, these type of sync fet controllers should only be implemented in conjunction with load current detection circuitry. That is, when the load current suddenly falls below say 25% (or whatever level would lead to DCM in the equivalent non sync fet design), then the sync fet controller should be immediately disabled. With LTC1698, this would mean driving the sync fets from external gate drive ICs and then pulling their resistor fed inputs to ground in order to disable them. This could affect primary/secondary switching interleaving timings…so this also means that a similar external gate driver IC should be used to delay the primary side fet switchings.
Another way would be to have a dedicated microcontroller to "watch" the sync fet gate drives, and disable them if they remain on for more than one switching period.
And i believe...also have a backdrop TVS across the syn fet just to be sure.

The only topology that’s cast-iron safe with synchronous rectifiers is the Buck converter…because if the buck sync fet is turned off whilst reverse inductor current is flowing in it…then the inductor current simply starts to harmlessly flow through the diode across the Buck power fet, and into the input capacitor.

These sync fet controllers are absolute death traps for the uninitiated.

Who else makes sync rect controller chip set pairs for two transistor forward?
Why do so few IC co's make them?...is it because of these issues highlighted here?

I am not knocking them….as discussed, if the above mentioned additions are made to the circuit, then they are do-able.

Actrually, adding RCD clamps across the sync FETs does solve the problem with the LT1681/LTC1698 combo. (As attached LTspice sim and pdf.)
In fact, the lack of the usual fast current sense (with the LTC1698) now makes me think this solution may be better than others which have fast sync rect current sense….since with the low sense thresholds on offer, they look like noise disasters waiting to happen.
I think I would add a TVS across the clamp capacitor too.
Any ideas why these seemingly essential additions aren’t advised on the LTC1698 datasheet?

treez said:

Any ideas why these seemingly essential additions aren’t advised on the LTC1698 datasheet?

Click to expand...

It seems to me that in olden times, the datasheets were
also where companies would put applications information
(National especially was good in that respect). But now
I think they prefer to make the datasheets only that, and
throw app notes when deemed worthwhile - finding them,
or knowing they even exist, is your problem. I see many
datasheets without a whit of useful applications info,
just an URL for the "product resources page" if you're lucky.

As far as why not sync rect, maybe the benefit of losing
the catch-diode Vf is still just not worth the complexity /
BOM cost for high VIN operation - catch-diode inefficiency
term is something like Vf/VIN*(1-duty), right? So if that
rolls up like 0.8/48*0.5, you're beating back a 0.8% loss
term (like dropping from 85% to 84% efficiency number,
or so) at what cost? Just moving conduction loss to
the switching loss column, by adding components and
sensitivities / risk?

Now lower VIN makes this trade more worthwhile, sure.
At what point "more" becomes "sufficiently", I can't say
offhand.
But maybe you have better fish to fry, than chasing
that particular loss-term.

New chips generally require a lot of development time - as the companies supplying them simply do not test them in the way a real application does ...