Reverse polarity protection

[Haplo]

I have a question which is hopefully not too difficult to answer:

In a project, I have three devices (lets say Dev1, Dev2 and Dev3). These are daisy chained. Dev1 connects to port 1 of Dev2, and port2 of Dev2 goes to Dev3:

Dev1->Dev2P1----Dev2P2->Dev3

Dev1 is the supplier of power to the whole thing.
Now the problem: Both port 1 and port 2 on Dev2 use the same type connector (RJ12) but reverse wiring. This is something I've inherited from an older design and can't change because of the compatibility issues.
Now if someone plugs things backwards (Dev1->Dev2P2----Dev2P1->Dev3) then Dev2/Dev3/Both will smoke. Althuogh everything will be clearly labeled, I can't take the chance.

Is there anyway to protect the circuit against this? I can't use a series diode because the supply voltage from Dev1 is already 5v and I can't afford the voltage drop. There are other protection schemes like this one:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/360

But not useful for my case since I have two possible entry points for power, not one.

I'd appreciate any help.
 
 

[frankb]

See this link for a scheme using a single p-channel mosfet. The penalty in voltage drop is on the order of millivolts for a typical embedded control circuit. [url]
http://electronicdesign.com/power/fet-supplies-low-voltage-reverse-polarity-protection[/url]
I've used it many times with good results.

[PCM programmer]

I recently did a project that runs off two AA Alkaline cells.
So it starts at 3.2v and lowest we can allow it to go is about 2.0v.

We couldn't accept any voltage drop, so we used the "reversed FET"
method. This is shown in Figure 5 of this Maxim appnote:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/636
The Drain goes to the batteries (or input voltage). The Source is
the output. The Gate goes to ground. Pretty simple.

I used a IRLML6401 Logic Level P-Channel FET. It's relatively
inexpensive. On a reel, it costs $.20 (US) from Digikey.
http://www.digikey.com

You'll notice in the Maxim appnote that they worry about using this
circuit at low voltages. That's because the FETs have a silicon diode
as their internal diode. So they might not fully turn on at low voltages,
due to the 0.7v (or so) drop across the silicon diode. I fixed this
problem by putting a BAT54 schottky diode in parallel with the FET.
ie., the schottky diode is in parallel with the internal silicon diode
inside the FET. In my notes on the schematic, it says "D1 typically
drops 0.25v at 1 ma". So there's minimal voltage drop across the
schottky diode during the "turn on" period. Enough voltage can get
across to the Source pin (the output pin) on the FET, to ensure that
it turns on properly. BAT54 from Fairchild is only $.05 (US) from
Digikey in quantity.

Your application uses +5v, and if that is regulated, then I don't think
you'll need the paralleled BAT54 diode. I don't know what your
current is, so you might check whether the IRLML6401 is appropriate.
What I do, is just look at the catalog page from Digikey, and pick a
suitable Logic Level FET that has a reasonably low RDSon, and is
cheap, and is in a SOT-23 package (which is what I like to use).
And, it has to be available.

On Edit:
Frankb got in his reply first while I was typing in my massive tome.
Oh well. At least my idea of a paralleled schottky diode may
help someone in the future. (Also fixed link to Maxim website)..

[Haplo]

Thank you both for your answers. I think I'll use PCM Programmer's suggestion on port 2 of Dev2.