# Protecting A/D inputs

I'm looking for ideas on how to protect A/D inputs from disaster. My current application is automotive, but I seem to be blowing up the AD ports pretty regularly.

Zeners are not ideal (they don't always trigger at a voltage low enough to protect the port).

My understanding is that the port is good to Vcc + 0.3V; after that you can blow the port or the whole chip.

If I use voltage dividers and set VREF lower (say 2.56V) using analogReference(INTERNAL2V56) does that give me room between 2.56V and Vcc+0.3 for over-voltages? Then I could use a 4.5V zener without any trouble.

Would something like this work?

http://www.mouser.com/ds/2/302/RB751_SER-104980.pdf

The board will be surface mount. The divider has a 23k5 R so current is limited.

cptdondo: Would something like this work?

http://www.mouser.com/ds/2/302/RB751_SER-104980.pdf

The board will be surface mount. The divider has a 23k5 R so current is limited.

Yes that would work. For protection see:- http://www.thebox.myzen.co.uk/Tutorial/Protection.html and http://www.digikey.com/us/en/techzone/microcontroller/resources/articles/protecting-inputs-in-digital-electronics.html

If I change VREF do I need to protect against V > VREF or is the threshold still V > Vcc?

If I change VREF do I need to protect against V > VREF

No.

or is the threshold still V > Vcc?

Yes :)

Awesome! That makes life a lot easier. :)

One more question:

If I use a 10K current limiting resistor before the diodes would that affect the accuracy?

I am assuming that the internal resistance of AD itself is on the order of 100M so a 10K in series would be less than a decimal point....

If I use a 10K current limiting resistor before the diodes would that affect the accuracy?

It would affect the acquisition time if you are switching between analogue inputs. It could be that the first few samples read on a new channel would be off by a bit.

Right now I throw away the first 8 readings, and then average the next 128. I may need to tweak that with some timing as I learn more, but the idea is to average out the readings in case any noise gets through. The signal is noisy (it has both 1k5Hz and 3kHz PWM mixed in and probably all sorts of harmonics) and I'm interested in long-term (1 sec or so) average.

You are best to reduce the noise at source with a capacitor across the analogue input.

Yes. I have a 2.2uF cap to ground across the input to filter out noise. Not sure if I need the multiple sampling and averaging as well. I just don't have enough data in the real world for my scenario.