I've been putting some thinking into protecting opamp inputs from over voltage and could do with someone confirming my thinking on this. I'm going to assume that this is a complex topic but as naive as ever I want to try and keep it simple if at all possible! I'm working on the basis of a transitive ESD event of 1kV over a number of ns and a stress event of 50V over a number of ms
Example circuit: I actually have a BAV199 dual diode and TI OPA192 Opamp; input will be 0-2V. C1 represents the capacitance on a ADC pin with high impedance input. That has a LSB of 62.5uV and INL of 20.1uV. This opamp doesn't have the typical internal ESD diodes and using a proprietary means of over voltage protection up to the common mode range (V+ - V-) so 10V in this case.
Under normal operation, at peak voltage input of 2V, there is a reverse voltage on the diode D1 of 7V and a leakage shown by this graph:
Leakage current = 0.55nA or 550pA which seems to be good to quite a high temperature
That would induce a voltage drop of 5.5uV across R2. This would increase the opamp offset voltage from typical 8uV to 13.5uV (or max 50uV to 55.5uV.) This is less than 1LSB of the ADC so seems like a reasonable approach to protect the Opamp and, possibly, the ADC.
But...
Those diodes, once they start conducting will push current into the voltage sources which are linear regulators and cannot sink current so may well end up toast - cheaper than the opamps - but also may just end up out of regulation and become over voltage suppliers themselves. If I put a TVS or even a Zener on the regulator output that would just contribute additional and higher leakage current.
Could I put, say, a 1Mohm resistor between each diode and regulator to current limit and thus help with preventing the regulator going out of regulation but without affecting them under normal operating scenarios? Also is it worth protecting the inverting input of the OpAmp with a resistor - with the architecture of this opamp not having internal protection diodes it may not add anything except additional noise?
The choice of 10K for R2, with matching resistance for R4, is a tradeoff of current limiting into the opamp at 50V (0.25W) and additional noise (12.9nV/ √Hz )on a signal under normal circumstances. With a 50V over voltage scenario, that's excessive voltage at the non-inverting input but only 5mA. There will be things here that I'm unaware of so it may be that the better approach is just to go with current limiting and forget the diode protection; within my understanding it seems like a reasonable approach that won't have too much of an impact on the signal quality in relation to the ADC.
