IO protect from overvoltage

[Nigel Goodwin]

I thought when you assign a pin on a PIC as an input, it has a rather large input impedance already, or have I got this wrong

Not for a normal I/O pin, but for an analogue input it has to charge (or discharge) the internal sample and hold capacitor, this takes time, and the higher the source impedance the longer it takes.

 

[Hero999]

Even a modest ESD discharge would fry the protection diodes. Discrete zeners usually start at what 1/8 - 1/4 watt?

I don't belive that for a second. Human body model is 100pF with 1k charged to 2kV, that's just 200:mu:J; some how, I think the zener will survive. Those 1/4 and 1/2 ratings are for continious power, there's no reason why it can't stand five or more times the rated value for a couple of mS and even more for :mu:S.

 

[Nigel Goodwin]

Even a modest ESD discharge would fry the protection diodes.

Complete rubbish!.

The series resistor, suitably and sensibly chosen, will limit the current to easily safe levels.

 

[Sceadwian]

The voltage is so high from a decent static discharge and their power handling so low how is a basic current limiting resistor going to save the built in protection diode?

 

[denikar]

Well you need an ESD cap too. The resultant voltage at the pin is reduced via charge sharing between the ESD gun / person and 68nF capacitor.

Overall circuit is like the attached schematic (3 separate analog inputs shown).

 

[Sceadwian]

I'll have to toy around with that, I've never found any decent static discharge models for simulating. Then again I've never looked, and never put something in an environment where it might be an issue.

 

[Nigel Goodwin]

The voltage is so high from a decent static discharge and their power handling so low how is a basic current limiting resistor going to save the built in protection diode?

Do the maths! - any massive pulse like you're suggesting is going to blow a zener anyway! - the protection diodes are very sturdy devices. Don't forget, the inputs are CMOS, so you can use pretty sizeable series resistors! - far more so than with a zener, where leakage will restrict you.

 

[Sceadwian]

I don't means a massive pulse, just a brief EMF pulse. Due to die constraints and to keep costs down the protection diodes on a micro controller are as small as possible, which means the size of the diode junction is tiny, a discrete zener's junction is going to be MANY times the size of a micro controllers zener so it's pulse handling ability is going to be at least an order of magnitude higher if not more even though the average power handling is going to be similar because the micro controller's die disipates the average heat. I unfortunatly don't have any numbers to do the math on but I'll hazzard a guess that the discrete zener's junction is at least 10 times the size of the micro controllers zener, which gives it an order of magnitude improved protection from a spike because there's so much more junction area to share the current. The overall power handling of the two is compareable because the rest of the micro controller die will act as a heatsink for the protection diode but that won't protect it from instantanious spike, which will quickly violate junction temperature limits compared to discrete.

 

[Nigel Goodwin]

I don't means a massive pulse, just a brief EMF pulse. Due to die constraints and to keep costs down the protection diodes on a micro controller are as small as possible, which means the size of the diode junction is tiny, a discrete zener's junction is going to be MANY times the size of a micro controllers zener so it's pulse handling ability is going to be at least an order of magnitude higher if not more even though the average power handling is going to be similar because the micro controller's die disipates the average heat. I unfortunatly don't have any numbers to do the math on but I'll hazzard a guess that the discrete zener's junction is at least 10 times the size of the micro controllers zener, which gives it an order of magnitude improved protection from a spike because there's so much more junction area to share the current. The overall power handling of the two is compareable because the rest of the micro controller die will act as a heatsink for the protection diode but that won't protect it from instantanious spike, which will quickly violate junction temperature limits compared to discrete.

Which is why you have a current limiting resistor to keep any potential pulse within acceptable limits - in just the same way as you have to do for a zener diode as well.

 

[Sceadwian]

Current limiting resistors look like dead shorts to an EMF spike.
I'm having trouble finding models (not sure of the capacitance one should use) to calculate the size of the actual pulse in order to determine junction temperature, but the voltage of a small static spark is 1000 volts and some produced like I get from my car all the time (I hate those) is in the neighboorhood of 10kv's Even a 100k resistor is going to let 10ma's through on the low side, and with 10kv the current pulse would be as high as 100ma's though the duration is going to be virtually instantanious the junction size of the diode suddenly becomes VERY important because even if it's just for a few nano seconds, violating the junction temperature will instantly destroy the device. I'll see if I can find any other information on the actual size of the junction involved in a micro controllers diode compared to say a 1/8th watt discrete but I'm going to guess that 10 to 100 times the physical size isn't going to be over estimating which makes the EMF pulse handling ability of a micro controller diode virtually nill compared to a discrete. If this device is going into a car a little EMF protection is going to be a good idea, especially if it's on a remote sensor.

I used to frequent AVRFreaks.net which is a forum for Atmel's AVR chips and every single time this topic comes up the power users there always warn against trusting the built in clamp diodes on a micro controller to protect from a EMF type pulses because of it's low pulse handling ability.

 

[Nigel Goodwin]

Current limiting resistors look like dead shorts to an EMF spike.
I'm having trouble finding models (not sure of the capacitance one should use) to calculate the size of the actual pulse in order to determine junction temperature, but the voltage of a small static spark is 1000 volts and some produced like I get from my car all the time (I hate those) is in the neighboorhood of 10kv's Even a 100k resistor is going to let 10ma's through on the low side, and with 10kv the current pulse would be as high as 100ma's though the duration is going to be virtually instantanious the junction size of the diode suddenly becomes VERY important because even if it's just for a few nano seconds, violating the junction temperature will instantly destroy the device. I'll see if I can find any other information on the actual size of the junction involved in a micro controllers diode compared to say a 1/8th watt discrete but I'm going to guess that 10 to 100 times the physical size isn't going to be over estimating which makes the EMF pulse handling ability of a micro controller diode virtually nill compared to a discrete.

You've never tried to kill a PIC have you?

 

[Sceadwian]

My car would make short work of any PIC =) I have a pavlovian reaction of stark terror every time I get out of my car they're so bad. I shouldn't complain though all I need to do is tack a grounding strap to the bottom of the car so it drags but I'm too lazy to put it on =\. I try to remember to touch the car with my key instead of my hand when I get out but I usually remember AFTER I get the shock. I'm not sure how to calculate the voltage from the length of the spark in damp air but I've pulled sparks over an inch off my car, it's enough to make me actually drop my keys.

 

[Nigel Goodwin]

My car would make short work of any PIC =) I have a pavlovian reaction of stark terror every time I get out of my car they're so bad. I shouldn't complain though all I need to do is tack a grounding strap to the bottom of the car so it drags but I'm too lazy to put it on =\. I try to remember to touch the car with my key instead of my hand when I get out but I usually remember AFTER I get the shock. I'm not sure how to calculate the voltage from the length of the spark in damp air but I've pulled sparks over an inch off my car, it's enough to make me actually drop my keys.

Change your trousers then! - it's most likely it's YOU that's charged up, caused by the action of your trousers on the car seat.

We had this years ago with one of the engineers here - he could get out of the van, point his finger, and send a 3-4 inch spark to your nose!!.