One End Only
(first published in 'Line Up' magazine)
Some practices get drummed into us by those who have gone before and it is often wise to accept their experience. We have all been taught, or have learned for ourselves, how to coil cables so they don't appear to have only one end next time they need to be used. Another piece of traditional wisdom is that "ground loops", created when cable screens are connected at both ends of a circuit linking grounded equipment are bad and will cause problems. Whilst still an important concept, it needs to be re-examined in the light of today's equipment.
A handy shorthand for the concept of leaving the screen unconnected at one end is "one-end-only" (o-e-o). This principle has never been applicable to portable and self powered equipment as a cable screen connection is the only simple way to ensure that the point used as the ground reference for a mixer gets to microphones and to a portable VTR machine. General purpose balanced line XLR style cable for studio or location use needs 'hi', 'lo' and 'screen' connections at both ends so they can be used with confidence. The decision of whether to adopt o-e-o needs to be made when wiring fixed installations with multiple mains powered units having lots of interconnections between them.
Classic o-e-o systems
If we look at the type of installations o-e-o was designed for, it is easy to see why it enjoyed such popularity.
Broadcast complexes are often spread over many floors and sometimes amongst different buildings so to minimise noise problems, very substantial technical earth distribution systems can be used. Even the stoutest technical earth system cannot avoid some resistance and inductance. This means that as soon it does something useful and carries any current, the earth on, say, the studio A mixer will not be at the same potential as that on the MCR matrix. If an audio cable screen links these points, it will carry a current that often tends to be the fundamental and harmonics of the mains supply frequency. Testing (1) has shown that in many installations mains frequency screen currents can be around 100 mA, though in some cases they can exceed 1A!
It might be assumed that with balanced interconnections and good symmetry (common mode rejection), screen currents should not introduce noise into the audio but experience proves it happens. To see why, it is necessary to look at how many pieces of equipment are designed and the way input and output connections are linked to the internal electronics.
Sorts of ground
To properly understand the alternative methods of connection, we must first be clear what we mean by "ground". A true ground is something at the same potential as planet earth. Of course there can be potential differences between different areas of ground so for a particular installation we consider true earth to be what is provided by an efficient earth stake beneath or adjacent to the building. This true ground may also be used for the electrical supply safety earth that, as the name suggests, is intended to prevent risk of electric shock by keeping exposed metalwork at the same potential as people standing on earthy floors or touching other ground points such as building frameworks.
A second ground is the "relative ground" which exists in a control or equipment room once the true earth has been distributed within the building. If the building earth system is adequate, this will be very similar to true earth, but resistance and inductance mean some small voltages at power and RF frequencies will inevitably be present. Although battery powered portable equipment is normally used without any absolute ground reference, there will be a point that can be considered "relative ground" for the purpose of applying the principles described here. This might be a mixer or recorder earth terminal or any other point that is central to the system.
Within items of equipment, a third ground, sometimes called "analogue ground" (AGND) or "B0" is used as a convenient local point against which to refer the internal analogue signals and to carry the zero volt reference from unit's power supply. Digital equipment, may have a separate internal "digital ground" and some equipment brings one or more of these grounds to external terminals. Unless an installation has a clearly defined strategy which justifies an alternative, the normal setting for modern equipment is to link all its ground terminals together.
Mains powered devices with metal cases have a green/yellow wire or the designated earth pin of an IEC or other connector and this should always be wired to the ground that is distributed with the electrical supply. At one time some organisations used the separately cabled "technical earth" for this safety connection instead of the one distributed with the mains supply. Today this is frowned upon as it can create situations where the safety connection gets severed without a high probability the supply will also being interrupted. Safety earth and chassis ground are alternative names for the same thing. Safety regulations alone are sufficient reason always to bond the chassis to the safety earth connection but there is also a system reliability can also be affected. If the casework of equipment with capacitive mains input filters is allowed to "float" it will settle at around half mains voltage. When a signal input or output is connected the filter capacitors of a 230 volt piece of equipment are charged to something over 100 volts and can be discharged through the signal connections with catastrophic results to its own inputs and outputs and to the device to which it gets connected.
The ground not yet mentioned is that for the incoming and outgoing cable screens. Traditional thinking saw a need to keep the ground that equipment issued to the external cabling as clean as possible. This thinking probably came from the need to ensure phantom power and the ground to which it was referenced was a clean as possible. Connecting the mic circuit screen to a ground point that might have a noise voltage with respect to the point to which the phantom supply was referenced would cause that noise to appear superimposed on the phantom power and therefore be a source of noise on that mic input. The ideal point to be used was often seen as the analogue zero volt point in the mic amplifier module as this was where the power supply clean 0V point was available so it was where the mic cable received its ground.
To get each mic cable screen back to the mic amplifier required an XLR connector panel to have pin 1 isolated from the chassis then wired through the screen of internal cabling and eventually to the mic amp analogue ground. Similar isolated screen connections were also used with multipin connector panels. Once this convention was justified for microphone inputs, it was consistent also to apply it on line level inputs. Such schemes were used in large amounts of successful equipment with world renowned names and these units still work extremely well today - if installed correctly.
Internal ground wiring
The correct method requires knowledge of the internal design, though good product literature should avoid the need to take off lids. Most professional equipment has balanced inputs and outputs but the internal electronics are normally unbalanced with a single conductor carrying the signal against an internal analogue ground reference. When the cable screen is carried through to the printed circuit board for the analogue electronics, the internal analogue ground carries any noise current passing down the cable screen.
We have seen how in a large installation with an earth distribution system such as that shown earlier, cable screen currents of 100mA may occur. The design of modern equipment with fine PCB tracks and ribbon cables creates good opportunities for this current to develop noise voltages within the input stage and other parts of the device.
Because parts of the input section use ground references from around the input PCB and other sections of the device all have their own reference points, the noise voltages get mixed with the programme signal at a variety of points. The result is progressive degradation of the noise performance.
Currents flowing through the output terminals can also make their own contribution to this degradation which occurs regardless of whether the unit uses transformers or any other sort of input and output circuitry. Breaking the screen connection eliminates these screen currents, removes the noise and explains why the traditional o-e-o principle is so effective. Unfortunately, this solution which worked well for many years now starts to show a limitation as more equipment is digital and as we impose tighter controls on how equipment functions at radio frequencies.
A concentration on screening to avoid hum, can make us forget that a prime reason for screened cables, connectors and equipment housings is to keep out radio frequency fields. The traditional concern with most analogue equipment was to prevent it picking up and demodulating radio signals. The increased use of digital technology and switched mode power supplies means there is now a much greater risk of audio equipment being susceptible to incoming RF signals and it can also be a source of RF radiation. For the last few years it has been a legal requirement that all goods meet international standards to keep the ether free for communications purposes and this has meant designers have had to take much more seriously the avoidance of electromagnetic radiation and susceptibility to it. Meeting these EMC standards has meant re-thinking some of the traditional concepts used in audio engineering and as legislation was introduced some otherwise adequate products had to disappear from the market.
Consider once again our traditional unit with external cable grounded at the PCB but now instead of thinking about hum currents developing volt drops around the PCB, imagine the cable screen with an RF current on it. This RF may also produce volt drops that can get mixed with the audio in the same way as the hum currents and demodulation may occur giving radio breakthrough. Sadly, things can be worse as RF has a secondary mechanism as well as volt drops by which it can get into the audio electronics. At hum frequencies, the stray capacitance between cable cores, connector pins and PCB tracks presents a high impedance that can be ignored. At radio frequencies the impedance of this stray capacitance can be low enough for it provide an easy route for the RF to couple into many more parts of the circuit. This gives the system an even greater chance of the audio being degraded with whistles, buzzes and even foreign radio stations - British ones being too shy to intrude!
A device such as a digital reverb unit has a mixture of analogue and digital processing. The digital parts have fast clock pulses with many high frequency components. The stray capacitance that allows high frequency noise on the cable screen to get inside equipment can also provide a means of escape to the outside world of internal RF signals. Cable screens that originate from an internal PCB if carried through to the outside world create a reasonable transmission mechanism to broadcast this RF noise and probably create problems in other equipment. Such a unit is obviously going to have difficulty meeting the permitted radiation specifications which, with few exceptions, are measured with external cabling connected. PCB designers have a number of strategies including careful use of ground planes to minimise coupling of RF signals into outgoing circuits, but care with the way they deliver I/O connections is a fundamental one.
Designers have therefore adopted an RF oriented approach for equipment design rather than the earlier one which focussed more on hum avoidance. This has brought about a fresh view of the way to handle cable screens. They are no longer thought of as there to provide cleanly distributed ground references to microphones and other equipment. Instead they are part of an overall RF shielding system which inevitably also shields lower frequency noise.
To avoid bringing RF into sensitive circuit boards and to prevent cable screens distributing RF from PCBs, designers no longer bring them directly to the PCB. All external cable screens are connected to the chassis as soon as they enter or leave the equipment. The entire enclosure and cable screen can then be considered a Faraday cage, with as few gaps as possible to minimise the leakage of RF in or out. Slots for ventilation and for removable media need careful design to minimise the RF that might leak through. Although digital products often require sturdy, power and DC busbars, the traditional worries of noise superimposition on phantom supplies is generally avoided as it is no longer the norm to build large mixing consoles with long earth busbars and their inevitable volts drops forming part of the audio signal path.
This Faraday cage concept also gives an important benefit in installations where a hum current might circulate along a cable screen. Any hum current arriving at the input connector is not carried through to the PCBs with active electronics. Instead it "disappears" into the very low impedance path provide by the mains safety connection to the exterior case of the equipment which prevents it from producing any significant volt drops and coupling into the audio.
The optimum rejection of RF demands that all parts of the cable, equipment and connectors are shielded as fully as possible. When XLR style connectors are used the metal shell normally contacts the chassis connector achieving full screening of the shells without having to wire to any shell terminal that may be provided. In a system connected for optimal RF rejection, cable screens will always be linked to XLR pin 1 or the equivalent pin on a multi-way. With modern equipment no harm will result in also linking this to the shell but such links can cause problems with older systems that takes the pin 1 connection to some point further inside the unit. This can create a internal ground loop with the screen of the internal cable now being linked to chassis both through an analogue 0 volt point on the PCB and also to the chassis at the terminal panel.
Maximum flexibility is therefore achieved when the shell connection is left floating in cable XLRs, though it is sometimes argued that the lack of a shell connection is undesirable when in-line XLRs are plugged together. In-line XLRs will therefore have limited RF screening and users must weigh up the small risk this creates against the likelihood of the connector being used with "legacy" equipment and also the further risk their earthed connector might come into contact with something that induced a new screen current!
Neil Muncy (1) has carried out tests injecting 60Hz, 600Hz and 6kHz signals into the cable screen of a pair of units which had chassis bonded cable screens and a variety of cable types. As might be expected, his results differ with cable type but the important trend was that the performance of a system with an intrinsic noise floor of -96dBu (30KHz bandwidth) was either not degraded, or by only two or three dB for most hum signals.
All new equipment must meet EMC performance limits and this is usually achieved by connecting cable screens to chassis ground at the point of entry. Even when there is a risk of circulating hum currents, Muncy's tests show that such interconnected equipment should not suffer when screens are fully bonded at both ends. Indeed, failure to bond screens at both ends can also result in equipment failing to meet the claimed EMC specification. The inductance of even just a few metres of cable screen can be enough to mean that the open end of a broken screen cable may not be adequately shielded to prevent RF being radiated.
As a contrast, some tests were conducted for this article involving the injection of 100mA 50 Hz noise currents directly into the XLR pin 1 connectors of some units using the "legacy" concept of carrying the cable screen to the PCB. As one might expect, results varied with the unit under test and which input or output XLR connector was used. However, the results showed a consistent pattern with noise floors degraded to an extent that was unacceptable for high grade audio. It therefore follows that with "legacy" equipment, these currents must be avoided by lifting cable screens.
Because there is not yet consistency in the way equipment internally connects cable screens there can be no absolute rule. However a good principle for the foreseeable future is to bond screens at both ends unless you experience performance problems with equipment which carries XLR pin1 into its innards.
References: (1) Muncy, Journal AES, Vol 43, No.6 1995
All material is copyright PHM © 2004.
P H M (P H Music) :
Ramsbottom : UK
tel: +44 (0)7799 621954