Protection against mis-wiring - logic signal vs 24V

[AndyC_772]

Hi,

I design a lot of circuits that are powered from a DC supply, typically a 12V or 24V battery. They usually connect to the outside world via one or more flavours of low speed, wired serial connection, such as RS232, CAN, RS485 and the like.

Sometimes mistakes are made when installing products, which can result in the cable being mis-wired in arbitrary and creative ways. Most of these errors won't do any harm, as I/O devices tend to be quite robust and incapable of causing each other any damage, but the power supply is an obvious exception. Connect +24V to any kind of logic input and it probably won't react well.

Sometimes I use a combination of series resistors and a TVS diode, which works fine if the 'hazardous' voltage that might get applied isn't too high - but the power dissipation becomes unmanageable if a pin has to survive a higher voltage and/or have a low series resistance to the outside world for functional reasons.

Another option is a TVS diode plus a PTC thermistor. This is fine provided the device itself doesn't have to operate in a hot environment, but most of mine do.

This seems like a fairly common requirement which could no doubt be met by some clever little IC containing back-to-back series MOSFETs, current measuring, a latch, noise filter, time delay and some simple control logic. I'm sure I used to know of such a device, but that was a long time ago now and I can't even think what it might be called. "Electronic fuse", or "solid state fuse", or "<some other kind of> fuse", probably.

Does anyone have a favourite idiot-proofing chip that they'd care to recommend please? Anything that'll stop, say, an RS485 interface from being destroyed by a 24V battery would be ideal.

[ConKbot]

Depending on how much capacitance and/or parasitic series resistance you can tolerate, JFET or depletion mode MOSFETs current clamps can limit currents so clamping and steering diodes aren't overwhelmed. 
Apex appnotes 53 has an example of the bidirectional clamp. The more voltage dropped across the resistor, the more the FETs are turned off, and they push out into the constant- current linear region of operation. But you have two RDS-on's and the series resistor in your circuit now.

Takes up a good bit of space, but it looks like microchip and IXYS make sot-223 FETs still, with 250-350v ratings, for reasonable prices.

I've used this for a 3.3/5V optoisolator input as a "just in case someone applies (US, 120V ) mains" strategy. Sure enough bit was put to the test weeks later, and the input still worked afterwards.

Appnotes covering PLC and other industrial electronics may have more in depth examples and a better variety of techniques used.

[nali]

I had this very issue with equipment fitted to buses using RS-485. In the end I just did a simple retrofit box consisting of a couple of TVS & polyfuses in the A/B lines which cured the problem.