Exploiting balanced circuits: balanced systems can fight unwanted noise and coupling, but test equipment is still geared toward the unbalanced.

The drive of progressive miniaturization through tighter packing densities and integration places a great premium on isolation. Internal parts must be packed progressively tighter or be more highly integrated. Achieving adequate isolation between different sections of a system has become a major design issue, placing a limit on the level of performance that can be achieved.

Screening is expensive and consumes otherwise useful space, and adds weight and cost. The use of balanced signals can provide a useful degree of isolation without severe penalties in space or cost. The RF/microwave industry has long concentrated on unbalanced techniques, but now balanced techniques are becoming popular. How do they work, and what are their idiosyncrasies?

Compared to regular unbalanced designs, balanced signals can help in two ways. First, they are less susceptible to unwanted signals and noises. Second, they are less prone to coupling their signals onto other paths. These benefits are cumulative; you can balance your susceptible signals and also balance the intentional paths of your intruder signals and expect to get added isolation.

EMC and emissions both benefit from the balanced approach independent of improvements in screening. High-speed digital designs now use balanced data and clock paths such as LVDS, and wireless device chipset manufacturers use balanced techniques to allow them to collect functions onto the same die, while still achieving adequate isolation to meet system requirements. An abundance of design information and equipment supports unbalanced techniques, thanks to the RF/microwave industry's long history with the technique. But balanced systems are less supported by experience, literature and test equipment.

The pitfalls of balanced systems mostly relate to common-mode issues--you are handling a signal path with two dimensions of freedom, but your intended signal determines only one. There is nothing too difficult here, but some of the foibles related to balanced systems can be counterintuitive to anyone coming from the world of unbalanced systems.

What is a balanced circuit? Balanced circuits are, quite simply, symmetrical. In the words of Henry W. Ott (1), "A balanced circuit is a two conductor circuit in which both conductors and all circuits connected to them have the same impedance with respect to ground and to all other conductors." But "balanced" is not the same as "differential." A single-ended circuit is still differential--the system ground is the other wire. However, in a truly balanced circuit there is no current in the ground system, because the two return currents in the ground are equal in magnitude but opposite in sign, and therefore, theoretically, cancel one another.

Fair and Balanced 

What are the advantages of balanced circuits? And how does a balanced circuit reject ground noise? Let's first look at a conventional single-ended circuit with ground-induced noise. See FIGURE 1.

[FIGURE 1 OMITTED]

Notice the noise source located between two parts of the circuit. It could represent different locations on a PCB, or perhaps the connection between a remote sensor and a measuring circuit. It is easy to see that the output voltage, [V.sub.L], is simply

[V.sub.L] = ([V.sub.S] + [V.sub.N])[R.sub.L]/([R.sub.S] + [R.sub.L])

Noise has been added to the signal, reducing the signal-to-noise ratio. In many situations this effect is detrimental to the system performance and may be unacceptable. FIGURE 2 shows an alternative way of connecting these two points. Here the signal is split into two equal parts, with the electrical center grounded. The source and load resistors are duplicated, and the output signal, [V.sub.L], is taken across both of the load resistors.

[FIGURE 2 OMITTED]

0 = [V.sub.S1] - [i.sub.1] ([R.sub.S1] + [R.sub.L1]) + [V.sub.N]

Figure 1 is slightly tougher to analyze than Figure 2. Kirchoff's Second Law tells us that the voltages around the [i.sub.1] and [i.sub.2] loops add up to zero. So that

0 = [V.sub.S2] - [i.sub.2] ([R.sub.S2] + [R.sub.L2]) - [V.sub.N]

[i.sub.1] = ([V.sub.S1] + [V.sub.N]) / ([R.sub.S1] + [R.sub.L1])

and

[i.sub.2] = ([V.sub.S2] - [V.sub.N]) / [R.sub.S2] + [R.sub.L2])

then,

[V.sub.L] = [i.sub.1][R.sub.L1] + [i.sub.2][R.sub.L2]

[V.sub.L] = [R.sub.L1] ([V.sub.S1] + [V.sub.N]) / [R.sub.S1] + [R.sub.L1]) + [R.sub.L2] ([V.sub.S2] - [V.sub.N]) / [R.sub.S2] + [R.sub.L2])

In a perfectly balanced system, [R.sub.L1] = [R.sub.L2] = [R.sub.L] and [R.sub.S1] = [R.sub.S2] = [R.sub.S] then

[V.sub.L] = [R.sub.L] ([V.sub.S1] + [V.sub.S2]) / ([R.sub.S] + [R.sub.L])

Now the noise voltage has been completely cancelled. Of course in a practical application, the resistors are never exactly equal and the noise is not completely cancelled. Rearranging the expression for [V.sub.L]:

[V.sub.L] = [V.sub.S1][R.sub.L1]/([R.sub.S1] + [R.sub.L1]) + [V.sub.S2][R.sub.L2]/([R.sub.S2] + [R.sub.L2]) + [V.sub.N]{[R.sub.L1] / ([R.sub.S1] + [R.sub.L1]) - [R.sub.L2] / ([R.sub.S2] + [R.sub.L2])}

The part of the equation being multiplied by VN is the common-mode rejection ratio or CMRR. The CMRR is the ratio of differential gain to common-mode gain:

CMRR = {[R.sub.L1] / ([R.sub.S1] + [R.sub.L1]) - [R.sub.L2] / ([R.sub.S2] + [R.sub.L2])}

For example, if [R.sub.L1] was 1% higher than nominal and [R.sub.L2] was 1% lower than nominal, the CMRR would be 0.005 or -46 dB. The rejection of ground noise is not the only advantage of balanced circuits. Some noise can be coupled capacitively, possibly from a nearby cable, or even inductively. FIGURE 3 shows how capacitive coupling might appear.

[FIGURE 3 OMITTED]

Even without resorting to circuit analysis, it can easily be seen that the current flowing from [V.sub.N] through the capacitors and into the load resistors make both [V.sub.L1] and [V.sub.L2] more positive. The result is a cancellation of noise voltage (similar to the cancellation of ground noise), provided that the circuit is resistively balanced and the capacitors have the same value. Another way of looking at this circuit is to redraw it as a bridge, as in FIGURE 4. If [R.sub.L1] = [R.sub.L2] and [R.sub.S1] = [R.sub.S2], it will be balanced when C1 = C2, and the noise output ([V.sub.L]) will be zero.

[FIGURE 4 OMITTED]

These are the basic advantages of balanced circuits: They are less susceptible to ground noise and capacitive coupling. Similar arguments can also be made to show that inductive coupling and electromagnetic coupling are reduced using balanced circuit topology. Reciprocally, balanced circuits also reduce radiation of these fields over unbalanced circuits.

The main disadvantage to using balanced circuits is that they require the use of more components. It is also very possible to connect up a balanced system incorrectly, thus inverting the signal. The decision to use balanced or unbalanced topology will need to be a considered on a case-by-case basis. There will always be a proportion of applications that do not require balancing, or for which it is unsuitable. There will also be cases where it is essential.

There are two types of subtle pitfalls to using balanced circuits. First, the benefit of balanced signals stems from cancellation of unwanted signal flows. This is a subtractive process; if things are well balanced, the reduction can be very great. A small imbalance can then cause a relatively large increase in unwanted signal flow. It is important to calculate, experiment with or simulate deliberate Imbalance in order to determine if the system will be robust enough to tolerate the expected component tolerances.

Second, balanced signal lines have that extra dimension of freedom: the common-mode signal. We have to plan what we want to do with it as part of our design process. We may need a specific common mode voltage to bias an amplifier, or to position the voltages in the range of an ADC. Alternatively, we may want to remove unwanted common-mode noise. Amplifiers can have good common mode rejection at their inputs, or possibly even have common-mode gain. Some differential filter and attenuator topologies act on the differential component but not at all on the common mode component. Two single-ended filters or attenuators act on both paths, but may cause common-mode to differential-mode conversion if they are not well matched.

FIGURE 5 shows two possible filter and attenuator configurations and details the differences between their effects on common-mode signals and differential mode signals. FIGURE 6 shows two common amplifier configurations and explains their effects on common-mode and differential-mode signals. Managing common-mode signals is easy; the key lies in understanding that they must always be considered.

[FIGURE 5-6 OMITTED]

By Any Measure

The RF industry's focus on unbalanced systems has created a vast legacy of test equipment with unbalanced ports. Oscilloscopes normally feature an "A minus B" mode that allows the use of two channels as one differential input. (Remember to occasionally put both scope probes on the same signal to check that both channels are adjusted to match well and leave a small enough residual).

However, the CMRR of measurements made with this arrangement may be limited and ill-defined. Available balanced probes offer specified common-mode rejection, and only take up one oscilloscope channel per balanced signal. These offer more dependable measurement accuracy.

Recently, network analyzer systems have begun targeting the growth in balanced techniques. But the rest of the RF test world has not been so helpful. Spectrum analyzers, signal generators, noise figure instruments, etc., are not yet available in balanced circuit versions. This forces us to use balanced-to-unbalanced converters, or "baluns." These can be simple transformers, but care is needed to use designs that cover both the wanted frequency range and the necessary impedance range. There is no such thing as a universal balun. Balun loss affects measurement calibration and noise floors, so pairs of similar baluns must be measured back-to-back to assess their performance.

There is no standard impedance or connector for balanced RF signals, so custom arrangements are inevitable. Using unbalanced instruments with adaptors allows customization of the balanced interface and suits the necessary mixture of unbalanced and balanced measurements.

For now, you may have to use a customized approach each time you utilize a balanced circuit topology. But as the drive toward miniaturization continues and isolating sections of a system becomes more problematic, test instrument vendors should eventually address the need for balanced circuit tools.

REFERENCES

(1.) Ott, Henry W., "Noise Reduction Techniques in Electronic Systems," John Wiley and Sons. New York, 1988.

ANTHONY LYMER is a senior development engineer at Agilent's Queensferry Telecomms Division. He can be reached at [e-mail]. David Stockton is lead engineer in hardware development for Agilent's noise figure analyzer and the N4000 group of smart noise sources. He can be reached at [e-mail].