Part 1 considered the principles of how screened cables work to avoid interference from electric and magnetic fields. We now look at some more practical details.
As described in Part 1, we need the screening to be effective over a very wide range of threat frequencies. For screened cables to be effective at high frequencies it is essential that the screen be connected directly to the reference pole (“0V” etc.) at both the sending and receiving ends.
The use of a pigtail to connect the screen reduces the screening effectiveness by allowing noise current in the screen to inject a noise voltage into the signal circuit. For a truly galvanically isolated circuit this may be unimportant, because the isolation minimises the noise current which circulates and flows in the pigtail. However generally pigtails should be kept to a minimum. For wide band data circuits pigtails must be avoided, which can be achieved by clamping the cable screen directly to the chassis or reference or ground point.
There are guidelines in circulation which recommend connecting the cable screen at only one end. It should be clear from the above that this defeats the high-frequency screening benefit of the screen. In the past these guidelines were applied to some kinds of simple electrical control circuits which are inherently immune to high frequency interference, to avoid ground loops – see below. It can also prevent power system fault currents from circulating in cable screens, but this should properly be achieved by ensuring adequate equipotential bonding in the power system.
In power distribution systems there are certain applications where power cable shields must not be connected at both ends, in order to avoid touch potential hazards during faults or lightning activity; as in TT distribution systems. This does not apply to the motor cables of variable speed drives.
Whenever signal cables pass outside of buildings and outside an equipotential bonding area, consideration must be given to safety during electrical faults and lightning strikes when potentially dangerous differential ground potentials may exist.
The simplest kind of analogue interface is shown in Figure 4. This is adequate for many general applications. From the foregoing explanation, you will see that this arrangement has some weaknesses, which may be acceptable where control at high precision and wide bandwidth is not required.
The dashed lines in the controller and drive indicate that the 0V connections of the controller and drive are usually connected to ground, either directly and intentionally or because some equipment in the system has 0V and ground connected internally. In this case there is a risk of disturbance from the following sources:
Precision analogue circuits often provide differential inputs, and sometimes differential outputs. They are commonly provided for precision controllers such as servo drives, and also for sin/cos shaft encoders. When used correctly these give excellent suppression of low-frequency interference. In conjunction with a screened cable this can achieve immunity over the entire noise spectrum. Figure 5 shows how to use a differential analogue input. Note that the signal cores would normally be a twisted pair, which further improves noise immunity by making the route of the two conductors as well-balanced as possible.
In this case we have a single-ended controller output and a differential input. By using two cores in the screened cable we can connect the inverting input directly to the reference pole of the controller analogue output. Any low-frequency voltage induced in the cable screen is therefore rejected by the input, whilst the screen still gives its excellent high-frequency rejection. The differential input cannot reject common-mode voltage at the high frequencies, beyond its bandwidth, where the cable screen works best. The two techniques combined give noise rejection over the whole spectrum.
Grounding clamps as in Figure 4 can also be used to avoid the high-frequency noise coupling caused by the pigtails.
If the controller also offers a differential output then the AI- core can be connected to the AO- terminal rather than 0V at the controller. A special case is if the controller offers a “virtual earth” output, where the AO- terminal is not an output but a sense input. In that case the AO- line must be connected to 0V at either one end or the other, it must not be allowed to “float”.
Digital circuits are not susceptible to disturbance from the kind of low-level low frequency errors caused by ground loops. High-frequency interference in a data link can cause bit errors which are normally detected and rejected, but if they occur too often the channel may close down or give inadequate performance. Shaft encoder circuits for speed/position feedback are particularly inclined to cause noise and vibration in the presence of high-frequency noise. In both cases correct management of the cable screen is essential.
Data links often use a high bit rate. For rates above about 1 Mb/s, the cable has to be correctly terminated in its characteristic impedance in order to avoid data errors from reflections. To maintain matching, only short lengths of exposed cable cores can be tolerated.
The most widely used digital interface for basic local data transmission is based on the RS422 and RS485 standards, which have differential transmitters and receivers. The cable type is not directly specified, and in principle it might be unscreened provided it has the correct characteristic impedance, but usually screened cable is preferred.
The use of a balanced circuit means that injected
noise is rejected to a considerable degree because it is in the common mode, i.e. it affects both the lines equally and therefore does not appear as a signal. However the transmitters and receivers have limits to their common-mode range, so errors do occur if the noise voltage is too high, or too fast-changing, as well as if asymmetry causes the common mode noise to be coupled into the series mode. The standard line-driver chips used in most ports have a common-mode range of about 5V and give major errors if this is exceeded. This can be increased by using galvanically isolated ports, but this is costly.
In Control Techniques equipment the reference terminal is shown as “0V” in Figure 6. In other equipment it may variously be called “G” or “GND” for ground, “SC” for screen, or “reference”. Sometimes it is left unconnected, or even not provided. This may be successful for short links, or where the ports have well-designed galvanic isolation. Generally it is far preferable to connect 0V to the cable screen.
RS485 allows for multi-drop connection of multiple ports. The effect of the minor impedance mismatch at each port as well as the injected noise from each pigtail makes the arrangement increasingly sensitive to disturbance as the number of ports increases. Complete communication protocols using high data rates, such as Profibus, use defined hardware which in that case requires direct clamping of cable screens in the connectors to avoid pigtails, and the correct termination impedance network to be connected at the end nodes only.
Many industrial connectors are designed without proper provision for the management of cable screens because they were not intended to be used at high frequency. For general applications it is usually tolerable to connect the screen through a short pigtail to a connector pin. However it is far preferable to pass the screen connection through the conductive body of the connector so that it continues to surround the signal conductors, as is always the case for a radio frequency connector. If a signal circuit passes through multiple connectors, each with its pair of pigtails, the injected noise at each connector accumulates.
One way of managing screen connections is to clamp the screens together or to a common metal part. Hardware for this is available from suppliers of screw terminal blocks. The idea is illustrated in Figure 7.
The purpose of the clamp is to avoid pigtail screen connections, and therefore avoid the injection of the noise voltage which would appear at the pigtails. It links the screens with the minimum of stray inductance. The small area of exposed unscreened conductor at the terminals here is much less important than the pigtails. The reason is that the unscreened conductors are only exposed to electromagnetic fields in the immediate vicinity of the terminals, whereas the pigtails would be carrying noise current which has been collected along the entire run of the screened cable.
Usually the clamps would be fixed to grounded metal parts, but this is primarily for safety reasons. The EMC benefit is the very low inductance link between the two cable screens.
Ethernet is an exception to all of the above. Modern Ethernet does not need screened cable, but relies on a very well-balanced unscreened twisted pair cable in conjunction with galvanically isolating balanced transformer coupling to give excellent common-mode noise immunity. Also it does not use a multi-drop structure, so the tendency to accumulate noise current at multiple nodes is also avoided.
Having looked at the resistance effect in Figure 3 we are well placed to understand why in some applications it is advised not to connect the cable screen at both ends. The error voltage IR would not appear if the screen were only connected at one end because there could be no current in the screen. This advice might be given in order “to avoid ground loops”. However the cable will have lost all of its magnetic field screening capabilities, which means its high-frequency screening capabilities. This advice is only correct in a very particular situation, when all of these apply:
The most common case for this is within analogue audio systems, where even a small level of mains pickup causes an irritating “hum”. It may also apply in servo controllers with analogue interfaces, but there it is better to use a differential interface as explained above.
Double-screened cable is sometimes recommended, especially with shaft encoder interfaces where there are typically
three balanced pairs of data cores which are screened as pairs, possibly some power cores, and an overall screen.
In principle only one screen is needed for each data pair, and it could be either the overall screen or the individual screens. However the benefits of double-screened cable are:
Henry W Ott: Electromagnetic compatibility engineering: Wiley: ISBN 978-0-470-18930-6
Tim Williams and Keith Armstrong: EMC for Systems & Installations: Newnes: ISBN 9780750641678