140402-2345 EDT
I agree with the above comments for the most part.
It is absolutely essential that a proper EGC be provided. It will also help if it is greatly oversized. A supplemental ground rod at the CNC might marginally reduce noise problems.
If there is no connection to anything external to the CNC machine except input power and the necessary EGC, then the machine is essentially a shielded box, and that could be raised to 1,000,000 volts above earth ground (assuming the isolation resistance between the input energy and the machine components could tolerate this voltage) without a noise problem. This might require filtering of line noise at the input to the machine, and obviously the EGC connection to the machine is related to some point in the energy source which does not connect to earth. This is solely to illustrate the problem. Internally the electronics being in a shielded box shields it from noise other than conducted thru the power connection.
In all probability the noise problem that is of concern is communication to the machine for data or other reasons. The way to solve the communication problems is by dielectric isolation in the communication path. A simplified discussion for machine tool type people is at my web site
http://beta-a2.com/noise_grnd.html .
I had one customer with two nearly identical HAAS machines spaced about 1 foot apart, and fed from the same 240 bus. One machine was a 2005 model and the other about a year newer. RS232 drip mode communication to the newer machine was impossible. With servos off communication was possible, but you can not machine a part with servos off. The HAAS service people said there was nothing wrong with the machine because their laptop communicated perfectly. Therefore, HAAS said the problem was with the RS232 source. You could interchange the cables between the two machines and the problem stayed with the one problem machine. Yes it is a machine problem.
Installation of our I232 isolators in the path with the problem machine solved the problem, and further allowed transmission at 115.2 kbaud over 100 ft of cable.
The problem was the newer machine had brushless servos that created a lot of internal noise between the RS232 common and the machine chassis. The older machine used DC servos. Note, both machines had separate proper EGCs and supplemental ground rods.
The RS232 common wire (pin 7 or 5) in now way was adequate to tie the two different systems (computer and CNC) together with a sufficiently low impedance to prevent noise between the two references from producing communication errors.
The CNC reps and manufacturers basically do not know how to solve RS232 communication problems, or they would never suggest a ground rod as a solution.
The reason the HAAS serviceman's laptop did not have a problem with communication to the problem machine was that the laptop was essentially floating from the building EGC and the RS232 computer pin 5 to CNC pin 7 common connection established the same reference potential at both the computer and CNC.
.