SOLUTIONS:
removing ground loops in
distributed computer systems
 

As stated in "Ground Loops and Digital Equipment", the problem with distributed computer terminals is they are often powered on the nearest mains point with a different Ground potential which results in high currents and noise on the signal lines.

The most common solution is to create a 'beefy' earthing system between the terminals and the central processor, or to weaken the local earthing systems so that any common grounding between the terminals and processor manages to "short out" enough of the voltage and noise so as to no longer be a problem. Neither of these is actually tackling the problem, and leaves the user open to further failures should the parameters change within the installation.

There are a few distinct possible solutions. The first is to rely on the computer central processor to provide the ground via the signal cable. This requires that the terminal equipment's usual mains ground be lifted so as to avoid the ground loops, although there may be objection to this as the chassis could become a little live and "sting" (although would never endanger life).

This is accomplished using the circuit alongside showing that the chassis is still brought to earth (to stop static) through a 1k-ohm resistor. Most currents are reduced sufficiently to stop any further trouble. However, certain Health and Safety guidelines will still object to this line of defence and we satisfy them by installing the 90V gas discharge tube across the resistor. Should the chassis to ground voltage exceed the strike voltage the tube will conduct allowing a high current fault path for any relevant protection to act accordingly.

This circuit does not necessarily need to be built into the terminal equipment and comfortably resides in the plug top of the mains lead.

The next option is to install simple optical isolation interface devices. Please note that we are not referring to the expensive route of fibre optics! These opto-isolator devices simply provide, through very inexpensive circuitry, a method for the signal to be transferred across a galvanic barrier (i.e. no path for currents to flow).

"Baseband modems" are an alternative route although we have seen these fail under very noisy conditions. They do, however, save on installation costs as only a two pair cable is required to connect the two sides together. There are also types that are powered from the RS232 port directly making implementation a breeze!

The third option is to use "common mode chokes", and they work extremely effectively in this type of environment. The feature they boast is there is no need to modify either the terminal equipment, or the mains supply (and the equipment retains the earth it was supplied with - so that'll keep H&S happy!).

How these operate is described in the section devoted to them. The main issue here will more than likely be one fo costs as the transformer cores required to make such devices may be more expensive than other options mentioned. The one advantage over the "isolation" methods is it does not flummox the typical inadequate IT person trying to fathom out why there is no signal voltage when the connection is broken.

The last, but most successful method is to re-wire so that all terminals are supplied from the same supply and Earth as the central computer processor. A compromise can be reached where the earth is supplied from the central processor with Live and Neutral from anywhere nearby. Earth leakage currents often prove to be a challenge, but can be overcome.

It appears though, with all this advice, the 'insulting engineers' still get it wrong!

>>


| | Ask a Question |

© 23.12.01