> So, if I current limit to a milliamp or less, maybe even .1mA, am I 
> ok here?  I don't see the harm in forward-biasing the input protect 
> diodes with a small current, but then again I don't know much about 
> chip design norms...

"Come with me if you want to live" :-) (Google knows).

Russell will now chant his standard mantra on this subject and people 
will tell you why he's wrong (even though he's not) and tell you that 
they have done it for years with no problems (even though that may not 
actually be true)    (even though they think it is) and ....

In the data sheets there are TWO specs for maximum and minimum input 
voltages.

One spec is in the "absolute maximum"ratings section and this is what 
most people refer to. It relates to the chip not dying and says 
NOTHING about whether the chip will function correctly. It may and it 
may not.

The second spec gives the values allowable during *normal operation*. 
`It gives the values where the manufacturer guarantees that the chip 
will operate as designed. Exceeding these values in ANY manner MAY 
cause misoperation. The allowable values are typically in the range 0 
to 0.3v outside the power supply rails. The intention is that the 
protection diodes NEVER conduct during normal operation.

The reason that you should not allow the protection diodes to conduct 
is that the current flows into areas of the IC that are not designed 
to have current flowing in them. Transistors can be biased on or off, 
nodes that have no formal discharge path can accept charge and form 
spurious FET's that do funny things. This really does happen. The 
effects can be intermittent, variable to random and may well not 
relate in any apparent way to the pin causing the problem.

*HOWEVER* everything MAY work perfectly. Or it may appear to. Or it 
may almost always do so. Or ... .

Many people do this and don't see any problems and therefore tell 
others that it's OK. In practice it very often is. If you can accept 
"probably almost never does totally random things" for your product 
then by all means do it.

In practice it is found that very very small currents almost never 
cause perceptible problems (for some values of almost never). This is 
true enough that Microchip without embarrassment produce application 
notes like eg AN521 in violation of their own spec sheets and even 
defend such things if questioned. Some of their app notes are far 
worse in this area.

The truth is that app note AN521 is almost certainly OK. I would 
personally consider doing what they recommend there. (Note how large 
the resistor values are!). BUT if the product dies or malfunctions you 
should not be surprised.

There is a very safe way to violate the rules above and "almost 
always" live. This works so close to "always works OK" that I have 
never seen it fail - but it may still violate specs, depending on 
which PIC you use. You use a series input resistor as before but also 
clamp each input to Vdd and Vss with a small reverse biased Schottky 
diode so they can only rise 1 Schottky diode drop outside rails. As 
this is less than the internal silicon diode drop it is usually OK. 
Usually.

Another "bodge" which may be cheaper and also works "every time so 
far" but still technically violates spec is to use TWO series 
resistors. Clamp the junction between the two with 2 ordinary silicon 
diodes (1N4148 etc) to rails. If say the second resistor is 100k you 
have essentially zero current flow into the pins when the IC clamp 
diode rises to conduction point as there is eg Vdd+0.6v at both ends 
of it so I = 0. SEEMS to work very well. But may still violate spec.

Some PIC datasheets allow larger excursions than others. I suspect 
this is not a matter of test pr design but just of the designer 
setting a value that they feel is far enough away from full protection 
diode conduction that it doesn't matter.

All the people who care and who can't read data sheets properly will 
now explain why I'm wrong and a few who can read datasheets will do so 
as well. I agree with them conditionally - you can indeed get away 
with it sometimes or even often on many IC's in many circumstances. If 
your IC & circumstance is one of these by all means listen to them :-)



        Russell McMahon




-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist