According to page 4 of this doc from Amtel, the input protection diodes shouldn't conduct more than 1 mA. In broken English: "It is not recommended that the clamping diodes are conducting more than maximum 1 mA"
Also, in this discussion on resistors in series between an output and an input, MarkT said:
MarkT: When you link (an output and an input) with a current-limiting resistor then you introduce low-pass filtering (the R combines with the stray capacitance and the input-capacitange of the pin).
For fast signals that means you need quite a small resistor, such as 220 ohms, for slower signals 2k2 or 10k might be fine.
If you have two devices with separate power such that one may be powered up when the other is down you have to worry about back-powering one device from the other, and the consequent risk of CMOS-latch-up. In this circumstance current is forced through the input protection diodes, and too much is a one way to trigger latch-up mode (which is bad news).
I am using a HX711 with a barebones 328p-pu. The 328 is powered by a 78SR-5/2-C Murata DC/DC converter. The HX711 is driven by a Maxim MAX6250 voltage reference. Grounds connected of course. A 24v wall wart powers both the 78SR and the MAX.
At power on, the MAX6250 "starts up" faster than the 78SR and for about 25 ms the data output pin of the HX711 is about 3 volts higher than the 328's Vcc.
So, I will be very briefly back-powering the 328 through the 328 input pin and its protection diode. I don't know how much current the data output pin of the HX711 is capable of supplying, but to be on the safe side...
...I am thinking of putting a 3300 ohm resistor between the HX711 output and the 328 input, to limit the current to less than 1 mA. Or, 2700 ohms, since there will be a 0.7 v (?) drop across the protection diode [(3 - 0.7) / 0.001 = 2300...say 2700]
Would appreciate your thoughts on this. Specifically, is a resistor necessary here or a good idea? If so, would a different value be a better choice?
Thank you, Dave