If one routes one signal in a multiwire cable that quickly transitions (e.g. a digital signal), a 10mV to 500mV spike will inductively couple to all other wires in the cable, at each transition. This means that the 10mV to 500mV spike is added to the other signals, and results in an error equal to the magnitude of the coupled spike. This is called "cable cross-talk" and you can see this for yourself by doing the experiment in the illustration below. For example, a 4 foot .05" spaced ribbon cable with 15 wires and a +/-5V square wave on wire #1 will cause +/-150mV spikes to be added to wire #15. This means that the signal on wire #15 has a maximum accuracy of 150mV, due to the cable cross-talk alone. This experiment is easy to reproduce with a function generator; oscilloscope or data acquisition system; and a multiwire cable. In fact, to measure the cable effect with an existing data acquisition system, we recommend that you ground the most sensitive input to your data acquisition system at the sensor (far from the computer), digitize, and view the "grounded" 0V signal. What ever "stuff" you see on the computer screen from this grounded input is being added to your sensor signal when it is not grounded, and therefore sets your maximum possible accuracy. In addition to cable cross-talk, noise can originate from ground loops, electromagnetic fields (emf) in the air, or emf's generated from millions of simultaneously switching transistors inside the computer (at MHz rates).
The instruNet solution is to place the instruNet network device within inches of the sensor, if possible, and then run single wires between each sensor and each screw terminal. Running analog signals across a room, in a multiwire cable, typically leads to ten's of milliVolts of accuracy. instruNet accuracy, on the other hand, is limited by the background radiation within the signal conditioning operational amplifiers themselves - which is typically ten's of microVolts.