100124-1256 EST
I ground shields at one end or the other, both ends, neither end, or no shield.
What does a shield do?
Consider a microphone input. High impedance dynamic microphones produce small output voltages for full scale. One I have just measured is about 100 MV + to - peak talking close and moderately loud. With 2.5 ft of open clip leads and floating shell on the microphone the high frequency noise is about 10 to 20 MV. Putting my hand on the mic shell reduces the noise to about 5 MV.
With a shielded cable this noise will be in the microvolt range. Note these are 2 terminal connections to the microphone. Thus, one is the shield and connected to the microphone housing, and the center conductor is the signal relative to the shield. Not a good idea to physically ground both ends of this shield because any current thru the shield will be a signal added to the desired signal. Usually the microphone end is essentially insulated from earth by a wood floor or carpet.
Another way to handle the microphone is to have the transducer, it is really a small speaker in some cases, have two leads isolated from the housing connected to two wires in the shielded cable. Now it is a three wire cable as far as separate conductors are concerned. If the two signal wires are twisted, then these tend to cancel out both magnetically and capacitive coupled noise if these two wires feed a differential input amplifier. Now if you earthed both ends of the shield, and there was little current in the shield, then there would be negligible noise coupled to the signal lines.
Electrostatic shields reduce capacitive coupling to components inside of the shield. Electrostatic shields do little for magnetic fields. Twisted pair wiring reduces magnetic coupling if the magnetic field is fairly uniform in intensity over a number of pitch periods of the twist.
Ignoring any shielding in a current loop system. If at the current transducer end one lead of the signal pair is earthed, possibly via an EGC, and if at the receiving end there is also earthing of one lead, then you have a noise problem from the difference in ground potentials between the two ends. One end or the other needs to be isolated from EGC, ground or earth.
Twisted pairs are probably more importaant in a current loop system than shielding. I suspect the best end to single point shield a cable for a current loop system is at the receiving end. It would be ideal if this shield could extend around the transducer and be insulated from ground at that end. For most application may not matter.
.
