We're in the midst of designing a device for measuring analog signals and have been testing our initial prototype. We had a single-ended gain stage before the ADC, but due to the connections in the rest of the system we are experiencing ground loops through the signal cable. These are causing offsets and throwing a heap of noise onto the signal.
Our initial solution has been to use an instrumentation amp as a pseudo-diff amp with a pulldown on the inverting input, as shown below.
I understand R3 prevents the inputs from floating when no device is connected while allowing the GND of the two devices to be at different potentials. I'm having a hard time figuring out why the value is so low though (the design is recycled from an older device). The signal we're interested goes down to a few mV and offsets of 0.1mV are acceptable. With input bias currents in the nA range for IN-amps I would think R3 could be much larger.
This comes about because there is a use-case for having numerous (10-20) of these amplifiers in parallel. With 10-20 R3's in parallel the impedance between the two system grounds will drop to 5-10 Ohm and the 'break' in the ground loop will be much less effective.
What might be the reason for keeping R3 low? I'm hoping to optimize it for some larger value but first need to understand the trade-offs.