PV MOSFET driver reliability

On a sunny day (Wed, 18 Feb 2015 13:41:58 +0100) it happened Piotr Wyderski wrote in :

mm, if your application has any microprocessor with FLASH memory, then that is likely to go long before the LEDs in optos.

_constantly_???? How about a relay? That can replace the MOSFET too.

Reaction time? Temperature? Voltage? Budget? all depends.

Can't comment on that but imagine them to be pricey. Why can't you use a high-side driver with integrated charge pump?

This example is for logic level FETs, not standard level.

The price doesn't matter, as there will be at most 3 of them and the device is not intended for mass production. A hobby project, but must be done right.

Because these drivers steal power from the drain and the drain terminal is connected to a rectified, but not filtered voltage, i.e. it would be < U_min for a considerable fraction of the duty cycle. If you power the driver from a separate supply, then the Micrel's 20V source to drain absolute maximum limit applies and I cannot guarantee that there will be no such spikes.

Such drivers are good, but not applicable here. I need a floating gate driver and this can be achieved by photovoltaics or inductive coupling in this form or another.

Best regards, Piotr

SSRs and optocouplers seem to be very reliable, used prudently.

Can you overdesign it from both ends?

Design so that the proper FET state is achived with only 1/20 the rated L ED current. And design so in normal operation the current is 1/2 the rated current and this will still be 10x more than needed. The low operating cur rent will reduce the rate of degredation of the LED and you will have a la rge 10x margin.

Mark

Usually all you need to do is stay well below the LEDs rated current. Some device data sheets have lifetime vs current and temperature curves.

One trick is to overshoot the LED current during transients, but back off steady-state. That gives fast switching but low average current.

Sort of like this:

That's a very interesting idea. OTOH, I know that a low power pulse transformer + two Schottky diodes will last for eternity, as there is no wear-out mechanism. The VOM1271's datasheet says "high reliability", but what exactly is that "high reliability"? 30 years of always on with I_F=5mA? :-/

Best regards, Piotr

You need a transformer oscillator/driver, too.

If I want isolated power, I generally buy a SIP dc/dc converter, instead of making the supply out of parts. There are tons of these in the $3 to $4 range.

May be I missed something: if "opened" means "not conducting" (as I interpret it, similar to relays or switches), the PV coupler only will be powered in that exceptional scenario. If you say "opened" thinking of a tap, pleaze ignore this message.

Whatever, is Piotr aware of depletion-mode MOSFETs? If he needs constant gate drive, removing it for a short interval, then change the MOSFET type and enjoy short drive intervals.

Then it probably boils down to whichever costs less. Your PV driver or this kind which costs around $5 in small quantities:

Sometimes switching speed also needs to be considered.

Yes, I use them sometimes to limit the short-circuit current (IXYS has parts able to go as high as 16A@500V) but:

1. It is next to impossible to buy them in small quantities and they cost an arm and a leg.
2. Even if one manages to get them, their R_DS_ON is much higher than that of the enhancement-mode N MOSes.
3. The property of "start disabled" is highly desired, as the initial state is much better defined.

What I am trying to do is a (nearly) lossless device for switching between two rectified out-of-phase AC power supplies 17V/50+A. 4 1mOhm N-MOSFETs seem to be more reliable than a high current relay and they allow me to add overcurrent protection easily.

Best regards, Piotr

The isolation is just for technical reasons, there are no safety requirements. The four transistors form two pairs of the mundane source-to-source AC switches and the input is 17*abs(sin x). At least for most of the time, wild spikes can happen. And here the floating driver is inherently much safer to the gate oxide.

I was thinking about that, but a tiny toroidal transformer driven at 400kHz + 2 diodes + a resistor is so much simpler...

Best regards, Piotr

PS. I use the "open==conducting" meaning, if it should be the other way around, I'll adapt.

Yes, but there often is one already, so can be (ab)used.

Agreed, but for MOSFET driving purposes several microamps are often enough (that's how much a typical PV coupler can yield), so the driver can be an MCU pin/a logic gate. Then a series capacitor, then the primary winding. This can safely scale upto low milliamps, which is more than needed if the switching can be slow.

It obviously is a half-bridge dc/dc converter, but I don't consider it to be a power supply, since the power is so miserable.

Best regards, Piotr

I've used regular isopower icouplers as a drivers for fets it was before they had gate drivers so two was used with the output in series to get 10V gates drive

it was used as a fast disconnect for a bunch of traics, they couldn't handle a dead short for a full cycle so something faster was needed

-Lasse

If they have to switch while 50+ amps are flowing you better swing the gates around darn fast. I don't know if your PV coupler can do that. Else a transformer plus UVLO (or driver with UVLO in there and input tied high) on the other side may be better. That would require some serious thought.

Yes, I've already used iCouplers in my previous project. Great parts in terms of speed, reliability and isolation, but the 8-pin ADuM5241 eats 70mA when idling, which is insane. The price of its internal simplicity -- no feedback channel.

Yes, and then PVs are not an option. They are sloooow...

Best regards, Piotr

If you mean cross-conduction, then it can't happen. But if you want me to leave the linear region ASAP in order not to fry the FETs, it's a concern. Joerg, is ~1ms long enough to thermally damage a TO247 device?

It can take long (