Years ago with the advent of low voltage industrial electronics systems, it was though necessary to only ground one end of the shield because if ground loops or different potentials existed between ground points existed, it could cause problems, since then, it is customary to carry out something called equi-potential (ground) bonding this entails bonding all metallic parts of a machine to a earth ground point Star point. If done properly, this virtually eliminates ground loops.
And If this is done, then a preferred method of grounding both ends of a shield can be carried out.
Siemens published a paper on this subject.

.

 

Noise problems are quite often poorly thought-out Circuit designs.
All Outputs, of any kind, should be designed with the lowest Impedance practical.
This will do more to reduce or eliminate Electrical-Noise than almost anything else You could do.

Also, any Digital-Signal should have its "sharp-corners" purposefully filtered-out.
This is, of course, relative to the operating Frequency.
"Removing the sharp corners" will cut-down on "Ringing",
which, otherwise, might not be creating an obvious problem with the functioning of the Circuit.

Then there's Input-Filtering.
All Inputs should be attenuated, ( Filtered ), above their normal expected operating Frequency.

Following these general guidelines, utilizing simple "Twisted-Pairs",
should eliminate all but the worst disruptive Noise problems.

Of course, it's best to separate any Wires carrying Current over maybe 1-Amp,
from Wires carrying "Low-Voltage" Signals or Data.
Run them in separate Cables, and keep them separated by at least a few inches.

Any time you have a bundle or Cable of Wires,
the individual Conductors should always be in the form of "Twisted-Pairs"
.
.
.

 

Certainly noise from PWM speed control has been a problem in some instances, and might be a problem in this instance unless prevented. Shielding of power wires is usually a good choice, along with power wires being in separate cables from signal wires.
For sensors, shielded pairs, either in a multi-pair cable,or by themselves, just one shielded pair, is the best choice. If a sensor is also "grounded" to the noisy conductive frame of a noisy machine, then there may be a problem if the shield is connected to that ground point.
Normally the shield of a sensor cable should connect to the sensor shield point in the sensor connector, and then also to the instrumentation end shield common point. Sensor shielding should NOT be connected to other grounds between the sensor and the input. This is important if terminal strips are involved in the system.
My cheating trick has been that while shields are connected at both ends in the initial construction, the option of disconnecting them is provided, because occasionally a shield needs to be disconnected at one end.
A special caution for current loop signals is to always run the pair to the input resistor before the common is tied to other commons. If you are not using current loop ( 4 to 20 mA) sensors this does not apply.
For power wires, shielding is needed if the power is from an electronic control, either PWM or phase control with any sort of motor driver. Often the manufacturer has recommendations, usually they are the best advice. Brush type motors always have noisey power wires that should always be twisted pairs and shielded if there are any sensors in the drive system. The power shields should be grounded to the power system safety ground point.

 

If you follow the equi-potential bonding practice, you do not have to worry about where the shields are grounded.
See ch6.
https://forum.allaboutcircuits.com/threads/another-shield-grounding-question.184947/#post-1709358

 

If you follow the equi-potential bonding practice, you do not have to worry about where the shields are grounded.
See ch6.
https://forum.allaboutcircuits.com/threads/another-shield-grounding-question.184947/#post-1709358

Adequate bonding of different sections is not always under the controls persons control. AND, while it may be possible to have "all potentials equal" the probability is that some will be "more equal than others." (Apologies to George Orwell)(all animals are equal, but some are more equal.)

 

Adequate bonding of different sections is not always under the controls persons control.

If you are wiring the enclosure, it usually is!

 

If you are wiring the enclosure, it usually is!

You're forgetting he learned from Jim Moore.

 

If you are wiring the enclosure, it usually is!

In most of our machines, ALL of the analog inputs come from devices outside the enclosure. Pressure transducers, temperature sensors, displacement transducers, and load cells. Sometimes encoders, which are not analog but can suffer from noise.

 

In most of our machines, ALL of the analog inputs come from devices outside the enclosure.

Yes, I agree, but I wire the external machine together with the enclosure accordingly, i.e. equi-potential bonded ground throughout the system. !