Is there a point to output ESD diodes?

I found when doing some modeling of input protection diode networks for logic gates, that the Schottkys had higher voltage drops at large currents than 1N4148 . This despite that fact that of course the Schottkys start to turn on at lower voltages. So I have been using the

1N4148 diodes now for all inputs due to the lower capacitance and lower voltage drop during surge currents.

Now the other issue is how to prevent the rails from rising above 5V when continuous overvoltage is present at an input or output. This is difficult. A 6.8V TVS across the rails isn't guaranteed to turn on until about 7.14V. That plus the 0.6V from an input diode means the input could get to about 7.74V before clamping. That's outside of device absolute ratings. Oh, I put a resistor between the diodes and input, so it won't exceed the internal clamp currents. But it still can see an input voltage higher than the ratings, because for continuous overvoltage, the internal diodes really won't get involved anyway. It will be the outer diode diverting current to the rail. So the input and associated FET gate will see 7.74V to ground.

What can one do? It seems a zener is better, though not as hefty for surges. So parallel a 5.6V zener with a 6.8V TVS? Then with the 5% tolerance on the zener, that gets us down to 5.88V+0.6V=about 6.5V. That's much better.

But we've now got a lot of protective hardware!

I see only RC networks on most computer products input circuitry, like disk drives for instance. Perhaps they aren't considered "external world" connections, since they are meant to be connected by a tech. following proper ESD precautions, and after that will never get zapped.

Trying to decide whether to go all the way to 100% bulletproof protection on external world connections vs. some reasonable compromise often kills more of my time than designing the cirsuits in the first place!

Good day!

--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarle@sandia.gov -- NOTE: Remove "BOGUS" from email address to reply.

Hello Chris,

This is usually spec'd in the data sheet. For logic it is often called "DC output diode current" under abs max ratings. If in doubt, I'd contact an applications engineer at the manufacturer.

For lines going to the outside world extra Schottkys are often indispensable, considering the moment of truth when the test lab guy puts on his cowboy hat and wields that big honking ESD zapper gun.

Regards, Joerg

In article , Chris Carlen wrote: [...]

One thing you can do to protect inputs from the outside world is use a discrete circuit for the receiver. This is a good idea for extreme cases.

Some regulator/reference circuits will sink as well as source current. This can be useful for low current circuits. For higher current circuit the heat issue is too much.

You can place a "posistor" in series. These are small PTC thermistors. They will disconnect the input after, lets say, 11mA flows for a while. This can help on the thermal issues.

[...]

First you need to come up with a reasonable model for the outside world. On the protect I'm currently on, it includes a idiot with some 440V, 3 phase wires and a soldering gun. Yours may include a scientist with wool pants and a plastic chair.

--
--
kensmith@rahul.net   forging knowledge

Yes, this is exactly how I do it.

Now if the line to be

Yes.

That's my feeling as well. I hadn't considered the FET approach. I have used crowbars at times. On one buffer, I have 200mA SMD fuses so if the overload into the output is continuous instead of just ESD, it will blow to preotect the rest of the system.

Good day!

--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarle@sandia.gov -- NOTE: Remove "BOGUS" from email address to reply.

Hi Chris,

Schottkys drop a lot when large currents are flowing and the diodes are tiny little versions. What I usually do in situations where things have to be very tough: A SOT23 (or bigger) diode pair CR1 like usual, to VCC and GND. A resistor R1 from CR1's center to the outbound connector and another resistor R2 from CR1's center to the device pin. Even if you'd take an ordinary silicon diode pair such as the BAV99 for CR1 to save cost then R2 will prevent the diversion of lots of current into the device port.

R2 can be small since the voltage divider formed by R2/CR1 under overload isn't allowing much into the device's diodes because its 'source voltage' is going to be less than a volt. Now if the line to be protected is an output that must drive a stiff load R1 will be a problem. The smaller you have to make R1 the larger the protection diodes have to be in terms of peak current capability. You can devise very fancy protection schemes beyond the diode approach but when dealing with multiple data lines the cost situation can get out of hand.

Protecting the rails is pretty easy. For example, a big old FET driven by a reference such as the TLV431 (that one costs under a quarter a pop). Set the resistor divider to wherever you want the rails to top off. Just remember that prolonged exposure would cause a FET to start glowing pretty quickly unless cooled. If you want the fuse to blow under that condition you can fire a crowbar circuit instead of holding things down with FETs.

Zener, TVS, MOV are not really to my liking in a situation where the voltage differential between normal operation and catastrophe is around a volt or so.

Regards, Joerg

Hi Chris,

The FET approach is more like a shunt regulator. In mission critical designs you probably want both, the FET solution to handle brief overages and a crowbar in case things do not turn back to normal within a reasonable time. Meaning before plumes of smoke come out ;-)

Regards, Joerg