Balanced Input
Balanced inputs, when used properly, can clean up hums, buzzes, RFI, and general extraneous rubbish. When not used properly, the balanced-input’s object may be partly defeated, but the connection will probably still improve the amplifier’s and system’s effective SNR.
Definition
To be truly balanced, a balanced input and the line coming in and the sending device must all have equal impedances to (signal) ground, to earth, and to everywhere else. Also, the signal must be exactly opposite in polarity but equal in magnitude, on each conductor.
Real Conditions
In practice, the signal is not of exactly opposite polarity. At high frequencies (and low frequencies in some poorly designed equipment), phase shifts add or subtract up to 90° or more, from the ideal 180° polarity difference. Otherwise the requirement for having a signal of opposite sign on each conductor is usually met. The exception is when one-half of a ground-referred, balanced source has been shorted to ground. Not surprisingly, this degrades the benefits of balancing.
Balancing Requirements
8.3.3.1 Input Impedances
The norm in modem pro-audio equipment is 10 kΩ across the line. This is commonly known as a “bridging load.” It is also the differential input impedance.
The common mode impedance, what any unwanted, induced noise signals will see, is often (but not always) half of this, for example, 5 kΩ in this case.
Considering the hum/RF noise rejection capability of an effective balanced input, input impedances much higher than 10 kΩ, say, 500 Ω, would seem feasible and useful in
professional systems. However, if the input resistance is developed by the ubiquitous input bias path resistors connected from each input to the 0-V rail, then there are limits to the usable resistance, before the input stage’s output offset voltage becomes unacceptably high. Although low Voos op-amps exist, a number of otherwise good ICs for audio have execrable DC characteristics, as IC designers do not appear to comprehend that good
DC performance is a most helpful feature for high performance audio. In this case, input impedances above 15 to 100 kΩ are found to be impractical, depending on bias current.
A galvanically floating input (i.e., the primary of a suitably wired transformer) has no connection to signal 0 V (as it has no bias currents), so there can be a very high common- mode impedance, say, l M or more, up to modest RF. This aids rejection.
Conversely, differential impedances of less than l Ok increase the influence of such random, external factors as mismatched cable core-to-shield capacitances.
Introducing Common Mode Rejection
Common mode rejection is an equipment and system specification that describes how well unwanted common mode signals, mainly hum and RF interference, are counteracted when using balanced connections.
Minimum Requirements
At the very least, all the equipment in a system must have a balanced input (alias a “differential receiver”). CMR can be improved and made more rugged when balanced inputs are used in conjunction with balanced outputs (alias “differential transmitters”), but this is not essential.
What Does CMR Achieve?
Common mode rejection action prevents the egress and build-up of extraneous hum, buzzes, and RFI when analogue signals are conveyed down cables, and between equipment powered from different locations—all the more so in big or complex systems. CMR helps make shielding more effective by canceling the attenuative residue, the bit that any practical shield “lets through.”
Sending the signal on a pair of twisted and parallel conductors ensures that this latter residue and any other stray signals that are picked up en route are literally coincident and appear “common mode,” that is, equal to each other in size and polarity. A tight enough twist makes the conductors almost experience interfering fields as if they occupied the same space. This is true below high RF (200 MHz, say), when averaged out over a cable’s length.
In contrast, the wanted, applied signal from both balanced and unbalanced output sockets is distinguished while being no less equal in size by appearing opposite in polarity on each input “leg,” called hot and cold.
CMR also makes shielding more effective by freeing it from signal conveyance, enabling it to be connected at one end only, according solely to the dictates of optimum RF suppression and/or individual system practice. Breaking the shields through connection also prevents (or at least lessens) the build-up of the mesh of earth loops that causes most intractable hums and buzzes. CMR is also able to cancel differences between disparate, physically distant and electrically noisy ground points in a system.
Above 20 kHz, even a modest CMR lessens the immediacy of the need for aggressive RF filtering. RF filtering can take place at higher frequencies, and both the explicit and the component-level effects on the audio may be diminished accordingly.
Figure 8.5 shows the CMV that CMR helps the audio system ignore. Even when connection to mains safety earth is avoided by ground lifting (ground lift switch open) or by total isolation (switch open and ground lift R omitted), considerable capacitance frequently remains, through power transformers and wiring dress.
Overall, the rejection achieved (which is a ratio, not an absolute amount) is described in minus (–) dB. Often the minus is understood and omitted. In plain English, “CMR = 40 dB” means “all extraneous garbage entering this box will be made 100 times smaller.”
What CMR Cannot Do
Like the stable door, the one thing CMR can’t do is remove unwanted noises that are already embedded in with the music. It follows that just one piece of equipment with poor CMR, and in the wrong place, can determine the hum and RFI level in a complex studio or PA path.
The ingress of common mode noise, called mode conversion, is cumulative, as each unit in the chain lets some of it leak through. As a result, the CMR performance and/or interconnection standards of all the equipment in complex systems (e.g., multiroom studios and major live sets) must be doubly good.
The higher CMR of well-engineered equipment (80 dB or more) provides a safety factor of 100- to over 1000-fold over the minimum 40 dB that is common in more “cheerful” products. However, the higher CMRs are more likely to vary with temperature and aging, as with all finely tuned artifacts.
Relativity Rules
The size of common mode (noise) signals is not fixed or even very predictable; they may range from microvolts to tens of volts. CMR is just a layer of protection. Forty dB of
protection is not much against 10 V of CMR, but it is definitely enough for 1 μV.
Sonic Effects of RF
Radio frequency interference is a common mode noise, and sources of RF go on increasing. In a competently wired system in premises away from radio transmitters and urban/ industrial electrical hash, a modest rejection no better than 40 dB has previously seemed good enough to make inaudible induced 50/60-Hz hum and harmonics, and the “glazey” sound of RFI and RF intermodulation artifacts. Unfortunately, RFI artifacts aren’t always blatant, and when any sound system is in use, they’re the last thing that users are likely to be listening for the symptoms of. However, even if there are no blatant noises, inadequate CMR can allow ambient electrical hash to cover up ambient and reverberative detail.
System Reality
The CMRs discussed are those cited for power amplifier input stages. The actual system CMR is inevitably cumulatively degraded by the cabling and the source CMRs. However, it can be maintained by ensuring all three have individually high CMRs and have highly balanced leg impedances. Lines driven from unbalanced sources give numerically inferior results, but often quite adequate ones (subject to appropriate grounding and
cable connections) in low-EMI domestic hi-fi and studio conditions, where equipment connections are also compact, and even in outdoor PA systems, in an open countryside.
Summary
Generally, 20 dB is a low, poor CMR, 40 to 70 dB is average to good, and 80 to 120 dB or more is very good and far harder to achieve in a real system. In a world where some audio measurements have had their credibility undermined, it’s reassuring to know that with CMR, more dBs remain simply better.