[SPAM bayesian] - Protecting Microcontrollers (was Re:[sdiy] converting a 10v p to p toa 0-5 volt signal) - Bayesian Filter detected spam

[Neil Johnson]

Hi,

Scott Nordlund wrote:
>> I don't think you understand the problem you are trying to solve.
>> Its not the voltage that is the problem - its the current that flows
>> into/out of the pin when the applied voltage is large enough to turn
>> on the internal ESD protection diodes.
>
> As far as I understand (and I'm certainly open to being corrected),  
> you don't want to forward-bias these diodes when the device is  
> powered.  I guess it would depend on the specific design, but in  
> general these diodes are only parasitic structures that exist as by- 
> products of the CMOS process.  Sending any current through them  
> while power is applied risks switching on parasitic thyristors that  
> could short the power supply rails until power is removed  
> (latchup...).
>
> I really understand this only in theory- I'm not familiar with any  
> practical implementations, so I don't know if it's standard  
> practice to circumvent this by isolating the inputs and outputs  
> from the rest of the substrate, etc.  I wouldn't want to risk it,  
> though, if I can satisfy my paranoia with a couple extra components.

Modern devices add the diodes for ESD protection.  As an example,  
read the datasheet for, say Atmel ATmega8 in the I/O section (page 51):

"All I/O pins have protection diodes to both VCC and Ground as  
indicated in Figure 21."

Another example: many Xilinx FPGAs have their programming pins run  
off the 2V5 rails.  But with current-limiting resistors it is  
possible to program them from 3V3 drive (ref. Xilinx app notes on  
programming Spartans).

I know that our chips have ESD protection diodes on all pins, and its  
something that gets tested quite extensively.

> This is the best approach that I know of.  If you use Schottky  
> diodes (for lower voltage drop and fast response time), you'd  
> probably be nearly bulletproof (though I haven't tested this...).

That's generally the idea when you want the input pins to have a nice  
low-impedance drive (e.g. fast signals).

Neil
--
http://www.njohnson.co.uk