by R. N. Marsh
"If the nature of a problem is not recognized, a useful solution will not be available."
-Ralph Morrison, Grounding and Shielding Techniques in Instrumentation
Amplifier performance suffers from more signal contaminants than just the amplifier's noise, frequency response problems, and nonlinear transfer characteristics. How the amplifier handles certain types of signals, called differential- and common-mode signals, can determine the ultimate performance level of a system. Common-mode signals, which are undesirable, can get converted to differential mode and then be amplified along with the desired signal. Insufficient rejection of common-mode signals can cause audible problems. Unfortunately, the way signals are typically handled in audio equipment makes it likely that these unwanted signals will contaminate the desired music signal.
All signals exist as differences of potential. The difference can be between a zero-reference conductor and a signal conductor or between two signal conductors. Signals referenced to a zero-reference conductor, called "single-ended" signals, are most common in consumer audio equipment. Signal differences between two points re moved from the zero-reference conductor, called difference or differential signals, are common in professional equipment. In Fig. 1, e1 and e2 are single-ended voltage differences be tween signal conductors and the zero-reference conductor, and e3 is a differential signal voltage.
Fig. 1--Single-ended signals (e_com, e1, and e2) and differential signals
(e_diff and e3).
A differential amplifier, the configuration used in professional studios, has three terminals at both input and out put, with one terminal of each group grounded (Fig. 2). The input terminals of such circuits are "floating." Whenever there is a difference between the voltages applied to the two ungrounded input terminals (ei1 and ei2), the amplified response appears as a difference between the voltages (eo1 and eo2) at the ungrounded output terminals. Difference signals of this type are known as differential-mode (DM) signals, whereas signals common to both terminals are called common-mode (CM). The common-mode signal is the average of two signals; 1/2(ei1 + ei2) or 1/2(eo1 + eo2).
The desired difference signal, sometimes called the "normal-mode" signal, contains the original in formation to be amplified. A significant advantage of the differential amplifier is its ability to distinguish common mode input signals from differential in put signals.
In an ideal differential amplifier, a common-mode voltage would cause no differential output voltage. The amplifier's output would be a function only of the differential input, and any common-mode signal would be rejected. In practice, no differential amplifier is perfect, because non-ideal conditions exist. A useful measure of performance, or figure of merit, is the common-mode rejection ratio (CMRR). This is defined as the ratio of the common-mode input to the differential-mode input that would produce the same differential-mode output. The higher the CMRR at a given frequency, the better the amplifier.
Common-mode rejection must be preserved in the input stage since common-mode error voltages, once added to the amplified signal output, cannot be separated from it. Single-ended input circuits (Fig. 3A), in which one terminal is grounded and the signal is applied to the other (typical of home audio equipment), have lower CMRRs than do completely balanced amplifiers with floating inputs (Fig. 3B).
For a differential input and a single-ended output, both the differential and the common-mode output error volt ages add to the output signal. With differential loading, only the differential output error signal directly adds to the amplified signal.
In practice, the common-mode rejection of a balanced amplifier stage is never infinite. It is, however, considerably greater when constant-current (high-impedance) sources are substituted for resistor Re in Fig. 2. Imbalances in the differential amplifier stage reduce the rejection of the average or common-mode signal. One common cause of imbalance in audio amplifier designs is device mismatching. If the two devices used in a balanced amplifier do not have identical characteristics, the transfer conductance of that amplifier's two sides will not be equal.
Therefore, there will be a differential-mode component of the output current in response to a common-mode input voltage; this is called common-mode to differential-mode conversion and is the source of the differential error volt ages referred to previously.
Inequality of load impedances (RL1 and RL2 in Fig. 2) also causes differential-mode conversion, even if the two sides have equal transfer conductance. Unequal input impedances from each input (ei1 and ei2 in Fig. 2) to the output common create differential errors too.
The greater the imbalance, the greater the unwanted differential error pickup will be. Imbalances between the two sides of a differential amplifier mismatch the gain presented to the common-mode signal at the two input terminals. This creates a differential error signal which the following stage amplifies.
Then there is the question of ground-signal reference regions, containing points which are supposedly, but not actually, at the same potential. Consider a signal potential that is developed in one ground-signal reference region and must be observed in a second such region. For the output signal e0 in Fig. 4, the zero-signal reference is taken as the zero-reference potential (e1) for the input signal. The potential at point B, as measured from point C, is el (assuming no current flows in R1 or R2, which would signify a potential difference between B and e1). Potential el is just the zero-reference potential of the signal measured from C. The potential at point A (again referenced to C) is esig + e1. Therefore, the potential difference esig can be measured from point C by subtracting the voltage at B from the voltage at A-in other words, esig = eA- eB.
The average (common-mode) value of the input eA + eB is:
(esig + e1 + e1)/2 = esig/2 + e1.
Since the unwanted term el is present, the amplifier in Fig. 4 must reject and not amplify this average value. Here, we can say that the common-mode signal is simply the difference of ground potential between ground-signal reference points. This points up the important fact that all conductive paths, reactive and resistive, must be considered-the amplifier is only one area of concern. For that reason, the common-mode rejection (CMR) demands on the amplifier can be met only by understanding the basic shielding and grounding processes involved.
In the design of amplifiers, single-ended topologies offer no rejection of these unwanted signals, while the differential family of circuits does. Unfortunately, it is too often assumed that the topology will take care of matters, so efforts to improve or optimize the circuit for CMR are rarely seen. Even when an active current source is used to bias the differential input stage, one cannot necessarily assume the matter is taken care of.
Fig. 2--A commonly used differential amplifier configuration, often
known as a long-tailed pair.
Fig. 3A--An op-amp used as a single-ended amplifier. The distinguishing
feature of the single-ended configuration is the through connection of
the zero-signal reference conductor, which means that ground-line signals
(eg) will be amplified. As a result, the output voltage for the single-ended
amp equals the gain, A, multiplied by the sum of the input and ground
voltages:
eo = A(e, + eg).
Fig. 3B--An op-amp used as a differential amplifier. In this configuration,
ground-line signals are attenuated by common-mode rejection, and output
voltage is as follows:
eo = A(ei + eg/CMRR).
The basic method for measuring CMR is to apply a common (mode) signal to both inputs (plus and minus) of the differential stage; the resulting residual output signal indicates the degree of rejection. Note that the input should be terminated with the output of the actual expected driving unit so that the measured CMR will be whatever the system produces. This system CMR could be much different if the amplifier had a short or specific resistor value terminating its input. For example, if a power amplifier were measured for CMR, the amount of rejection could vary significantly depending on the output impedance of the preamplifier. This is because the output impedance is added to the power amplifier input impedance, as seen by the power amp's input transistor. This transistor's impedance will then be out of balance relative to the other transistor of the input pair. It might help if terminating impedances for audio equipment were standardized. With a standard impedance (where appropriate), various components could be combined with minimal CMR degradation; per haps the communications industry standard of 600 ohms should be adopted. Taking its cue from communications equipment, recording studio equipment uses a balanced, 600-ohm impedance for the very purpose of crosstalk and CM rejection. A CMR ratio of at least 60 dB (and preferably 80 dB) at frequencies up to 20 kHz should be a minimum for the highest quality equipment.
Fig. 4--Two zero-signal reference regions; see text.
Fig. 5A--An op-amp connected as a differential amplifier can reduce
sensitivity to ground-line signals.
However, common-mode rejection is seriously degraded by signal-source resistance (Rs). Output voltage in this setup is:
eo = e,(- R3/R1) + (eg/CMRR), where CMRR s (R1 + R3)/RS.
Fig. 5B--The topology of an instrumentation amplifier offers high accuracy
and a common-mode rejection ratio exceeding 100 dB.
Here, output voltage is:
eo = [1 + 2 (R1/Rg)] (ei2- ei1).
In many cases, a single op-amp configuration used as a differential amplifier (Fig. 5A) to accept a balanced cable from a balanced source is insufficient to achieve good common-mode rejection. For improved CMR, a three-amplifier circuit known as an instrumentation amplifier (Fig. 5B) can be used. This consists of two noninverting amplifiers having a common gain-set ting resistor, followed by a difference amplifier. Input signal ei is duplicated across gain-setting resistor Rg via feedback that forces the differential in put voltages on A1 and A2 to be zero.
This is accomplished with current sup plied through the R1 feedback resistors. The outputs of A1 and A2, how ever, are amplified replicas of ei. Note that these outputs are of opposite phase. Also present at the Al and A2 outputs is the common-mode voltage of the inputs. The difference amplifier, A3, nulls this common-mode voltage while the opposing-phase ei signal components are combined; the result is a ground-referenced, amplified version of the differential input signal.
Common-mode rejection is important because CM error signals are un wanted signals that are not a part of the original. In addition to the imbalances already discussed, the ways in which signals enter the amplifier have a lot to do with CM rejection. Common-mode voltage is produced when a signal's electric or magnetic field is coupled to other circuit conductors such that currents flow in-phase on all wires.
Such field strength is converted to an open-circuited, available voltage. This voltage coupling can occur via a shared, common-ground impedance (conductive coupling). It can also occur when electrical or magnetic fields travel via nonconductive paths (field coupling). Examples of field coupling include wire-to-wire and cable-to-cable coupling (especially when the wires or cables are bundled together).
Other examples are emissions from fluorescent lamps and other appliances.
Part of the cure, as discussed above, is in the amplifier topologies used. If both the amplifier and the sys tem it is used in are optimized, the differential family of amplifiers (push pull, paraphase, and differential) are best for the rejection of these unwanted induced signals. However, we can do more than just use this family of topologies in our search for more cleanly reproduced signals.
When using wire pairs for balanced transmission of the low frequencies (below about 100 kHz), either magnetic-field or electric-field coupling will predominate, depending on both the wiring geometry and the impedance levels. Below about 100 kHz, we can make the following generalizations:
When the circuit impedance products (both source and load impedance) are less than 300 ohms, the principal coupling mechanism is magnetic-field.
However, when the circuit impedance products are more than 10 kilohms, the principal coupling mechanism is electric-field.
Fig. 6--Relative susceptibility of common cabling circuits to electric-field
and magnetic-field coupling.
Fig. 7--An a.c. power-line filter used to remove common-mode (CM) and
differential-mode (DM) noise.
Capacitor C1 converts DM to CM, which is then cancelled by common-mode choke T1. Capacitors C2 and C3 filter DM components only; C4 is sometimes added between filter output's hot and neutral lines to keep load-generated noise from being reflected back into the a.c. line.
Figure 6 shows some of the many possible cable constructions and their relative susceptibility to both magnetic- and electric-field coupling. Consumer amplifier inputs are usually con figured to accept single-ended input cables such as those shown in Figs. 6A, 6B, and 6C. In coaxial cables (Figs. 6A and 6C), both an inner wire and an outer cylindrical conductor carry the signal currents (source to load and return). Inasmuch as the outer conductor is usually grounded at the source, load, and other intermediate points (Fig. 6C), ground-current loops and common-mode currents caused by coupling of external noise sources are also carried on the outer coaxial conductor. Since both the desired signal and the undesired signal are carried on the same outer conductor simultaneously, some level of electrical interference certainly will be introduced into the system.
With differential amplifiers, we are able to use the other wiring methods of Fig. 6 for improved rejection. However, in the presence of severe electromagnetic interference (EMI), more complex configurations may be necessary. One such configuration would be triaxial cable, in which one signal conductor surrounds the other as a shield while an additional, grounded shield pro- [… this article, to be continued...]
(Source: Audio magazine, Feb. 1988)
Also see:
Nulling Out Amp Distortion by David Hafler (Feb. 1987)
The Magnavox 16-bit Series -- Making Good Players Better (June 1987)
= = = =