Over the past 20 years, digital media has made possible the economical delivery of music with very high dynamic range. Dynamic ranges above 90 dB are now common, and new technology is pushing toward 120 dB and beyond. Dynamic range, generally measured in dB, is the ratio of maximum undistorted signal (full-scale or onset of clipping) to residual noise (noise floor). How much dynamic range is enough depends on many factors and is usually a subjective decision, but consider that up to 120 dB of dynamic range may be required in "audiophile" sound systems.
Because the dynamic range of an entire signal processing chain can be no better than its weakest link, coupling of "ground noise" and other interference must be prevented all along the signal path. Generally speaking, each downstream device must have a dynamic range 10 to 15 dB better than the source in order to preserve the original input's dynamic range.
A predictable amount of "hiss" is inherent in any electronic device, but interface-coupled ground noise-usually heard as hum, buzz, clicks or pops-is generally much more irritating. Ground noise also contains considerable ultrasonic and RF energy, which can cause less obvious problems such as spectral contamination or digital jitter-related artifacts. In the vast majority of systems, ground noise and other interference enter the signal path through unintentional coupling in the signal interfaces. This article briefly explains how balanced and unbalanced interfaces work, how to troubleshoot a system easily and how to find solutions to noise problems.
WHY GROUNDING?In the context of audio systems, the term "ground" can be both confusing and misleading. To an electrician, ground involves an earth connection. To an electronics engineer, ground is simply a "zero volt" reference point.
Utility power lines are earth-grounded to provide a direct path to earth for lightning strikes. For the same reason, code requires any other wires (phone, CATV, etc.) to be protectively earth-grounded before they enter a dwelling. Fig. 1 shows how AC power is supplied through a typical single-phase, three-wire service to outlets on a typical branch circuit.
If it is powered by the AC line and has exposed conductive parts, any piece of equipment can become an electrocution hazard should its internal insulation fail. Therefore, most modern equipment has a third wire that connects the chassis or other exposed metal to the safety ground contact of the AC outlet, which is connected to neutral at the breaker box by either a green wire or metallic conduit. Now, if insulation fails, a high "fault" current flows in the safety ground circuit, the breaker trips, and power is quickly removed from the branch circuit. Though the white and black wires are part of the normal load circuit, the green wire is intended to carry ONLY fault currents.
Power line safety is extremely important. Never, ever use devices such as three-prong to two-prong AC plug adapters, sometimes called "ground lifters" to solve a noise problem. Remember that signal cables which connect equipment together can also carry lethal voltages throughout the system if just one ground-lifted device fails. The legal liability that results from using a ground lifter can bankrupt a business.
Equipment supplied with a two-prong AC plug has been specially designed to meet strict UL requirements for it to remain safe even if internal components fail. Only equipment originally supplied with a two-prong plug should be operated without safety grounding.
THE POWER LINE AND NOISE CURRENTSAn unavoidable consequence of AC powering system equipment is that significant noise voltages will exist between the chassis grounds of any two devices, whether safety-grounded or not. This must be accepted as a fact of life. This voltage is the dominant noise source in most systems, not noise picked up by cables as is so widely believed.
AC power normally contains a broad frequency range of undesirable noise in addition to the pure 60Hz sine wave. This buzz-and-beyond noise is created by devices that draw current in brief or intermittent pulses, such as power supplies in electronic equipment, fluorescent or dimmer-controlled lights, and intermittent or sparking loads such as switches, relays or brush-type motors. Except for extraordinary cases, attempts to reduce system noise by treating the power line itself are generally disappointing (more on this later).
Power line voltage causes small currents to flow in parasitic capacitances from each leg of the power lines to the equipment's chassis. This noise current then flows in system ground wires that, because of their resistance and inductance, develop ground noise voltages across them.
As shown in Fig. 2, these parasitic capacitances can be either intentional, such as power line EMI filters, or unavoidable, such as the interwinding capacitances of any power transformer. Because the coupling is capacitive, high-frequency noise is favored and the exact path taken by the noise current depends on whether the devices are grounded or floating.
Grounded devices use three-wire power cords. Noise currents flow through each device's power cord ground wire, producing noise voltages across them, to the outlet grounds. The chassis of each device will have a noise voltage with respect to the outlet ground. Depending on the physical distance between the two devices, there will be additional noise in the safety ground wiring caused by noise currents from other loads on the branch circuit. It's common to see 1 volt or more between two distant AC outlets. A current of 100 mA or more may flow in any wire, such as the shield of an audio cable, which connects these two points, completing a visible ground loop.
Floating devices use two-wire power cords. Since they're not grounded, a measurement of their chassis voltage with respect to safety ground may indicate up to 120 volts AC. Any wire that connects a floating device to ground, or two floating devices to each other, will complete an invisible ground loop. The loop current is small, usually less than 1 mA, and will cause an unpleasant but harmless tingle if it flows through a person. See Fig. 3.
THE PIN 1 PROBLEMWhen current flows in a resistance, a voltage drop occurs across it (Ohm's law). If equipment makers forget this when designing the internal shield grounds of inputs and outputs, their error effectively turns the shield connection into an input, a serious problem that Neil Muncy named the pin 1 problem. Because noise currents (which normally flow in cable shields) are allowed to flow through sensitive PC board traces or internal wiring, ground noise couples into high-gain circuitry. The bad news is that the problem is widespread. The good news is that simple tests can find it and simple modifications can fix it.
RF INTERFERENCEElectromagnetic compatibility (EMC) is the modern buzzword for equipment's radiation of and susceptibility to electromagnetic interference. Such interference isn't hard to find, and it can arrive either through the air and/or be conducted via any wire (signal or power) connected to the equipment. In addition to well-known sources, devices that produce electrical sparks, including welders, brush-type motors, relays and switches are often potent wide-band interference transmitters. Lesser-known sources include arcing or corona discharge in power line insulators (common in humid areas). Because power and telephone lines can act as huge antennae, they easily bring AM radio interference inside. But the most troublesome sources may already be inside, or worse yet, may be a part of your system. The most common offenders are light dimmers, fluorescent lights, TV or computer CRT displays or any piece of equipment using a switching power supply.
There are no simple, standardized tests to assess equipment's susceptibility to RF interference. Sadly, the performance of most commercial equipment will suffer because little or no attention was paid to susceptibility issues when it was designed. For example, the same design error that creates a pin 1 problem also creates an RFI problem, allowing RF picked up by the cable shield to enter sensitive circuitry. Symptoms of RF interference can range from actual demodulation of signals heard as music, voices, or buzz to subtle distortions, often described as a "veil" or "grain" in the sound.
BALANCED INTERFACESBoth balanced and unbalanced interfaces use two wires to convey the signal. Each is defined only by the impedances of the two wires to ground. In a balanced interface, they are nominally equal. In pro audio, interfaces customarily use two twisted conductors plus an overall shield. In theory, a balanced interface is noise-free perfection, completely rejecting the effects of ground noise between driver and receiver as well as the effects of magnetic or electrostatic fields on the cable.
A balanced input uses a differential device, whether an amplifier or a transformer, which responds only to the difference voltage at its inputs. By definition, it does not respond to common-mode (identical) voltages at both inputs. This property is called common-mode rejection. Common-mode rejection ratio, or CMRR, compares the response to differential and common-mode inputs. A simplified balanced interface is shown in Fig. 4.
In reality, a balanced interface is a system consisting of driver, cable and receiver. Real drivers and receivers have common-mode input and output impedances, shown as Zcm and Zo/2 respectively. The ratios and matching of these impedances, and those of the cable, are the major determinants of system ground noise rejection or CMRR. Contrary to popular belief, symmetrical (equal and opposite) signal voltages on balanced lines are not required and have absolutely nothing to do with noise rejection.
In real-world systems, CMRR must be maintained with equipment that's mass-produced and must be freely reconfigurable. To do this, driver common-mode impedances must be as low as possible, and receiver common-mode impedances must be as high as possible. Nearly all electronically or servo-balanced receivers have such low common-mode input impedances that system CMRR is exquisitely dependent on impedance imbalances in the driving output. Therefore, an input that claims 90 dB of CMRR (when driven by a perfect lab source) may have less than 50 dB when driven by an actual balanced output and less than 20 dB when driven by an actual unbalanced output. A wonderful benefit of transformer-balanced inputs is that their CMRR is immune to this degradation. They can maintain over 90 dB of rejection when driven from virtually any source, balanced or unbalanced. (A new active input circuit having this same immunity has been patented by this writer and is commercially available as the InGenius IC.)
UNBALANCED INTERFACESIn an unbalanced interface, the impedances of the two signal conductors to ground are unbalanced since one of them is grounded. Unbalanced interconnections are typified by RCA cables. Their use in a studio can be problematic because they lack any inherent ability to reject the effects of ground noise.
As shown in Fig. 5, noise current flows in the wire connected between points A and B. Because the wire has impedance, a small voltage drop will appear across it which becomes part of the signal. Because the wire impedance is common to both signal and noise current paths, this mechanism is called common impedance coupling. Noise caused by this coupling can even be predicted if noise currents and cable characteristics are known.
If both devices are grounded, either directly or indirectly through other signal interconnections, noise coupling can become very severe because ground noise on the building safety ground wiring is effectively forced across the shield. Unlike balanced systems, noise induced in unbalanced cables by nearby magnetic fields cannot be nullified by the receiving input. Cable shields, whether copper braid or aluminum foil, have virtually no effect on audio frequency magnetic fields.
TROUBLESHOOTINGMore often than not, the elimination of system hum and buzz involves a long series of trial-and-error experiments that end only when someone says, "I can live with that." Ground noise is very often the most serious problem in an audio system. I'll briefly describe a troubleshooting procedure that requires no test equipment except ears and some simple test adapters. It's effective, simple, and can be used in balanced or unbalanced audio and/or video systems.
Gather as many clues as possible before changing anything. Ask important questions like "Did it ever work right?" Sketch a block diagram of the system that shows all interconnecting cables, noting unbalanced inputs or outputs and equipment that is grounded by its power cord or other ground connection. Use the equipment's own controls, and some simple logic, to provide valuable clues.
The heart of the procedure involves the test adapters, which are easily made from standard connectors and a couple of resistors. These are temporarily inserted into the signal path while listening to the system output. If clues haven't indicated a specific portion of the system, start at the power amplifier inputs and proceed backward toward the signal source. The procedure reveals the precise location and nature of the problem, whether it's common-impedance coupling in an unbalanced cable, shield current-induced coupling in a balanced cable, magnetic or electrostatic pickup by a cable, or a pin 1 problem in the equipment.
MAKING IT RIGHT1. Don't assume that all AC outlets are wired properly! A swapped neutral and safety ground connection can create incredible ground noise problems.
2. Don't make any unnecessary ground connections! In general, extra ground connections simply create higher ground noise currents.
3. Test the system methodically. Install high-quality ground isolators at problem interfaces. High-performance audio transformers are especially well-suited: They're passive, reliable, inherently suppress RFI as well as ground noise, and their distortion is very low and particularly benign.
4. Give special attention to unbalanced/balanced conversion interfaces! How these connections are wired can make a huge difference in ground noise.
5. Use effective treatments to eliminate RFI problems. Whenever possible, find the offending RF source and treat it. A portable AM radio, tuned between stations, can serve as a homing device to locate a defective fluorescent fixture, for example. If the offending RF is above about 20 MHz, try installing ferrite clamp-on chokes at ends of cables. For lower frequencies, install small lowpass filters at inputs or outputs of susceptible equipment. Contrary to widespread belief, it's simply not possible to short out RFI with heavy ground wires.
6. Keep unbalanced cables as short as possible. Longer cables increase noise coupling. Never coil excess cable.
7. Use unbalanced cables with heavy gauge shielding, especially for long runs. Using heavy braided copper instead of foil and drain wire can reduce noise by 15 dB. This is the only property of cable that has any significant effect on audio-frequency noise.
8. Maintain solid connections. Hum or noise that changes if a connector is wiggled indicates a poor contact. Use a good commercial contact fluid and/or gold plated connectors.
9. Be sure all balanced line pairs are twisted. Twisting is what makes a balanced cable immune to noise from magnetic fields. Wiring at terminal strips and connectors is vulnerable because of this. In hostile environments, consider star-quad cable; it's about 40 dB more immune.
10. Don't forget the pin 1 problem. Lots of commercial equipment, some from respected manufacturers, has this designed-in problem. If disconnecting the shield at an input or output reduces noise, check the equipment at both ends.
SOLUTIONS OF DUBIOUS VALUEMany engineers believe that, since most system noise is coupled from the AC power line, some form of power line treatment will solve their problems. This belief, combined with large doses of vendor-driven hype and paranoia, is selling lots of this stuff. In my opinion, most are unnecessary and many actually make matters worse. Power isolation transformers, for example, have acquired a magical reputation, and misinformation about them abounds both in literature and seminars. The major problem with them, and various filters, conditioners and suppressors, is that they dump the noise that they presumably remove into the safety ground wiring. This often aggravates high-frequency noise problems. Dumping surges and spikes can cause such severe ground voltage transients between equipment that input or output circuitry is actually damaged or destroyed. However, when these devices are installed at the power service entrance, where they're tied to the common point for all building and earth grounds, they can be beneficial.
Symmetrical or balanced power systems are sometimes touted as the solution to ground noise. The basic concept, although seductively appealing, is seriously flawed. Even its proponents admit that noise reduction is usually less than 10 dB and rarely exceeds 15 dB. Many of the benefits ascribed to symmetrical power are due to simply powering all system equipment from the same outlet strip or dedicated branch circuit, which is a good idea for any system.