Reading analogue sensors in a noisy environment

Question

I need to read the values from several analogue sensors mounted on a circuit breaker in an electrical substation. I'm not working with any high voltages, but the circuit must still work as the breaker switches normal operating current (not fault current). The cables to the sensors will be up to four metres long. Some of the sensors have two wire 20mA current loop interfaces. Others are potentiometers. I'm aiming to sample at around 2kHz with 8 bit precision or better.

I'm not familiar with the dealing with noise from other equipment. I'm assuming I'll need to consider interference from both electric and magnetic fields.

1. Do current loops sensors tend to perform well in the vicinity of electric and magnetic fields? Should I use a twisted pair to reduce the net "loop area" that magnetic fields pass through? Should I have a grounded sheath to reduce capacitive coupling to the HV equipment?

2. What might I need to do for the potentiometers? Might a single ended signal work? Should I use a voltage dependant current source to convert the output to a 20mA current loop? Or perhaps use an inverting/non-inverting op-amp pair to create a differential output and use twisted pair cable? Or even mount discrete ADCs next to each one and use RS485 to send the signal digitally?

Answer 1

Lets examine the magnetic interference.

The Vinduce into a rectangular loop, from a straight wire located in the plane of the loop, and ignoring natural-log coefficients that may obfuscate the causality, is

Vinduce = [ MU0 * MUr * Loop_Area / (2 * pi * Distance) ] * dI/dT

In air, this simplified to

Vinduce = 2e-7 * Area/Distance * dI/dT [volts]

Assuming some evil spikes on the current waveform (from motors, or poorly filtered switching regulators) at the 100amp/100 microsecond size, using non-twisted pair sensor cables of physical nature 1meter long by 2mm signal-return spacing, and this sensor cable located 1cm from the evil spiky breaker&cables, we compute this level of trash

Vinduce = 2e-7 * (1 meter * 2milliMeters) / 1cm Distance * 100 amp/100uS

Vinduce = 2e-7 * 0.002 / 0.01 * 1e+6 amp/sec

Vinduce = 2e-7 * 0.2 * 1e+6 = 0.4 * 1e-1 = 0.04 volts of trash.

In a 2.5 volt sensor signal, the (unfiltered) trash causes 2.5/0.04 = 62:1 ratio, thus about 6 bits may be possible. Heavy filtering may improve the accuracy.