Reverse P-MOS

[throbscottle]

There's quite a neat trick you can use to protect power inputs of things from a reverse supply by connecting a p type mosfet with gate to 0v, drain to +V input and source to whatever you're protecting. It works by the intrinsic diode conducting the first bit of current, whilst the transistor turns fully on and presents a low impedance path "the wrong way" through it's body. Gives a lower voltage drop than a diode since you get the saturation voltage of the mosfet instead.

So I have been playing with my PSU design some more and realised it needs better protection against over-voltage at the outputs. A diode across b-e on the output transistor will protect it but can destroy the controlling transistor. A resistor in series to limit the current through the transistor plays hell with the current limiting circuit because the output can't drop to zero.

So it looks like I need a diode in series with the PSU's output, and was reminded of the reverse P-MOS trick. I'm pretty sure it will work, so I suppose the question is, what can go wrong?

 

[kubeek]

You have one flaw in there, in case of overvoltage in either polarity on the output you are very likely to exceed the Vgs rating of the mosfet. I suggest you use two zeners back to back between source and gate and then a resistor to ground.

 

[crutschow]

That won't protect against over-voltage.
And it only will protect against reverse voltage if the power supply is not on.

 

[throbscottle]

Yep. Just been trying it in LTSpice. IMA dumnass... Any other "smart diode" type circuits I can try?

 

[crutschow]

Are you trying to protect against over-voltage or reverse voltage?

 

[throbscottle]

It's correct polarity but applied from an external source to the PSU's input, so that if the supply is connected to something that retains charge like a large capacitor, but with no load to draw current, it can survive being turned down by more than the output transistor's Vbe (eg, I turn off the mains, or I just want to adjust it a lot).

The existing diode scheme is alright for a relatively small cap on the output, but not a large one, unless I use a power transistor for the one coming from the op-amp's output to the pass tranny (I'm sick of blowing up BCX38C's, so currently I've a BC137 strapped to a BC550 as a power darlington), which is better, but still limited in terms of what it can protect against. I could just use a Schottky in series with it I know, but it might cause the same nightmare problem where it interacts with the current limit and gets stuck in a destructive loop, that has been happening with the current limited version of the existing scheme, so it would be nice to have something that can reduce it's own forward voltage when it conducts. I found these: ZXGD3102T8 but don't know where I could buy one. TBH I was thinking it should be pretty straightforward to get the same result with a couple of extra transistors - but I was thinking it when I went to bed so haven't tried to cot me up with anything yet.

I'm using crowbar protection against reverse voltage applied to the input.

 

[throbscottle]

Right, come up with a design which uses the idea of detecting when the P-MOS's body diode is conducting, like the Diodes Inc. chip does, and turns it on if there's a voltage, off if there isn't. It's rather sensitive to resistor values and the c.c. source value, but it's a start, works correctly if a resistor is connected to the output instead of a cap. Transistors in the model are almost completely arbitrary...

 

[throbscottle]

Ok now I've played with the sim some more, I see that some mosfets exhibit more conduction due to the intrinsic diode when it's supposed to be on. Also the voltage dropped due to Rds(on) can exceed the diode voltage, so again it takes over. Hmm, have to see what else I can come up with.

 

[dr pepper]

I've considered using a smart diode circuit made from 2 fets for a charging system on a boat, so that the domestic battery could be connected to the alternator without voltage drop when the engine is running, but as yet I havent got around to it, nad to be honest if I was to do it seriously there'd be a pic or 'duino in it.

 

[throbscottle]

2 FET's? Care to share the goodness?

I would use one of these but the voltage rating is too low SM74611
Also found this LM74610-Q1, also there's a similar one from LT
Looking cursorily, there doesn't seem to be anywhere I can get them from though

 

[crutschow]

2 FET's? Care to share the goodness?
................

Two MOSFETs connected in series source-to-source will block voltage in both directions if the gates are connected to the common source node.
If the gates are biased ON then the MOSFETs will conduct in both directions.

 

[AnalogKid]

The intrinsic diode in a power MOSFET is a semi-conductor - it conducts current in one direction only. But the FET itself, when on, is a conductor - it conducts current in both directions. The old solution is to have two P-FETs in series with both gates toed to GND (assuming GND is fewer than 20 V away). Old Siliconix trick from the 60's. The new solution is an "ideal diode" controller from Linear Tech.

ak

 

[throbscottle]

(Zapper) Oh yes of course. Don't think I'd gain anything from that method. Brain fried. Been deep in psu building and testing and blowing up transistors and getting frustrated and staying up until 2am all week. Week off work and I'm knackered!
Started wondering if there's another way around this - if the controller "knows" about the diode drop, it can keep reducing the output voltage by that amount after zero is detected at the output. Hmmm...

(edit) OTOH that might give a dead spot in the start of the voltage control, I think

 

[throbscottle]

The intrinsic diode in a poser MOSFET

What, they wear fancy clothes and shiny shoes?

 

[crutschow]

...................
The existing diode scheme is alright for a relatively small cap on the output, but not a large one, unless I use a power transistor for the one coming from the op-amp's output to the pass tranny (I'm sick of blowing up BCX38C's, so currently I've a BC137 strapped to a BC550 as a power darlington), which is better, but still limited in terms of what it can protect against. I could just use a Schottky in series with it I know, but it might cause the same nightmare problem where it interacts with the current limit and gets stuck in a destructive loop, ......................

I don't see a problem with putting a Schottky diode in series with the emitter of the Q2 series regulator transistor to protect against voltage feedback.
Don't see how that would interfere with the current limit.
What is the current limit circuit?

 

[BobW]

This is the circuit that they use on the Raspberry Pi for reverse voltage protection:

 

[throbscottle]

I don't see a problem with putting a Schottky diode in series with the emitter of the Q2 series regulator transistor to protect against voltage feedback.
Don't see how that would interfere with the current limit.
What is the current limit circuit?

It's pretty standard - high side current detection produces a voltage, compared to a reference using an op-amp, pulls down the control reference voltage. I had a 68R resistor connected in series with the collector of Q3 so that if a large cap was connected to the output it would limit the current through Q3 when the voltage was turned down (the zener in the illustration represents a reference with a pot across it). But what I found was that (I think this is what happens) when the current limiting kicks in, it can't pull the output to zero, so Q3 is turned hard-on, and even with the load removed the limiter stays active. Not sure how but it always destroys the pass transistor (which is actually the output side of sziklai pair). So I was worried the same thing will happen with a series diode, though now I look at it properly, you're right. I must have been thinking the diode drop went the other way, or something :/

 

[throbscottle]

This is the circuit that they use on the Raspberry Pi for reverse voltage protection:View attachment 96859

Much simpler than my one, looks like it will work better too. Mine doesn't provide enough gate drive (in simulation anyway).

 

[dr pepper]

The restricting thing with a 2 fet diode is the max voltage you can operate them in reverse is the max gate voltage, usually 20v or so.

 

[AnalogKid]

The restricting thing with a 2 fet diode is the max voltage you can operate them in reverse is the max gate voltage, usually 20v or so.

Don't think so. As long as the gates are clamped to the sources with something like a 12 V zener (and a parallel resistor to turn off the FETs), a series gate resistor can go to whatever the power supply return is. I've done this with 270 Vdc and p-channel fets in the + input, and 360 Vdc with n-channel FETs in the return.

ak