On a PCB I'm working with, I have analog feedbacks running through op-amps into a 3.3V microcontroller. Thus, signals outside 0-3.3V are useless, and potentially dangerous to the microcontroller. The op amps have +-15V rails, because those rails were available and convenient. However, that means it's possible for my op amp to destroy my microcontroller by railing out positive. In perfect operation this should never happen, but I'm considering edge cases. To improve reliability, I'm considering connecting my op amps to +3.3V and -15V.
Is there any reason I should not run an op amp off asymmetrical voltage rails?
Obviously, railing out negative could also destroy the processor. I'm only considering the positive-rail case at the moment.
Best Answer
As long as the input pins or the output pin(s) of a typical rail-to-rail op amp are between the rail voltages, the "ground" reference is not actually locked to midway between the rails. (n.b. IIRC there are some op-amps which actually have a separate ground pin)
To illustrate, let us consider a rail-to-rail input/output (RRIO) op-amp in a voltage follower configuration:
We supply the op-amp rails with +/- 9.15 Volts, and inject an AC signal with a DC bias of 7.5 Volts and a peak-to-peak AC amplitude of 3.3 Volts.
The output would traverse between +9.15 Volts and +5.85 Volts, which is within the operating range of this hypothetical op-amp. The rest of the operating range, from -9.15 Volts through to +5.85 Volts would be unused so long as the signal stayed within parameters.
The mid value, 7.5 Volts, can thus be treated as the "ground" for this signal.
In other words, the op-amp doesn't care about the symmetry of the rails, it doesn't know where "ground" is.
For a real op-amp rather than an ideal one, there may be some minimum headroom for the output swing, from a few millivolts to a couple of Volts. So, this needs to be accounted for in the strategy laid out in the question.
In addition, the MCU pins may not take kindly to a traversal below the MCU ground rail. Hence, a clipping diode to the ground rail, on the input side, would be a good idea.