AVR-Chat

Larry Barello wrote:
> So, what happens when you take out the 1k resister?

instead of 9mA, we get 66mA into the negative diode.

> How about changing the 100k to 10k?

10K/10nF/1K results in a peak of around 55mA.

> Is there a freeware version of spice I can get a hold of?

Absolutely. Look at SwitcherCAD III, from Linear Technologies:
http://www.linear.com/company/software.jsp

It is a very nice program, and it's well supported and maintained.

It'll even run under Wine on Linux.

> I think your network forms a capacitive voltage divider. 

Exactly.

> The resistance helps cover any parasitic inductance in the capacitor (thus limiting the
> dV/dT on the cap.  The second resistor just limits the rate the capacitor
> can be discharged by the input protection diodes.

Yes. In fact, for back-of-envelope calculations you can assume exactly 
this: that the capacitors form a voltage divider, and that the second 
resistor is responsible for limiting peak current in the case of ESD. 
Naturally, the series combination of the resistors is responsible for 
sustained over-voltage protection.

> With the 100pf model, the peak voltage on the 10nf cap is something like 7
> volts (no load) - no wonder the peak current through the 1k is < 1ma!
> Steady state, the current is determined by the big resistor, so 7.5kv/100k =
> 75ma - kind of high for an AVR (and will vaporize your input resister!)

That's right. I was thinking that it would be hard to maintain a 7.5KV 
input voltage in an automobile. I wasn't even trying to protect against 
direct connection to the spark plugs.

For automotive purposes, it seems as if you can maintain protection 
against (say) +/- 100V or so continuously, and a few KV of ESD and other 
spikes, you can survive OK.

> Anyway, thanks for the tip on making more rugged input protection.

You're welcome. I got tired of my knee-jerk habit of using input diodes, 
and decided to try simulating what was happening. I also remember the 
conditioning that was used for the digital inputs on the old 
Bally/Midway microcontroller-controlled pinball games (which I 
maintained for a few years). They got away with a 1K/470pF network on 
every input to a couple of 4049s used to read the switch matrix. They 
didn't have to deal with ESD, though.


-- 
Ned Konz
ned@bike-nomad.com
http://bike-nomad.com