> A while back I reported that very small currents (<10uA) "into" the upper
> protection diode on an input caused the PIC to draw more current into VDD.
> Recently I have found that current "out of" the lower protection diode
> causes more severe problems.  Currents of about 2mA on the 16C622 can cause
> registers within the chip to change or can cause a chip reset (possibly
> watchdog).  When pulling current out of a PORT B input, I observed
> spontaneous changes to TRISB and OPTION registers as well as reset.

Well now, that's very interesting, becuase (at least for the 16C76
that I use) there is a SPECIFICATION that says up to 20 mA in the
clamp diodes is OK.  So you are saying the parts don't meet spec.
Is the spec for the 16C622 different than the 16C76?

Can you tell us more about what the power supply is to this PIC?

Since the clamp diodes are connected to the supply pins, is the
effect you are seeing feedback into the supply?

For example, if you parallel an external diode from the input to the
supply pin, does the same odd behavior occur?

I wonder if you taking the Vss pin negative by clamping a negative
voltage to an insufficient ground path to Vss.  In other words, what's
the rest of the current loop, and where are the voltage drops?