In many communication transceiver designs, an unbalanced single-ended signal must be converted to a balanced differential signal. This task is often considered routine, and a flux-coupled transformer or transmission-line balun is simply dropped into the signal chain to provide the desired conversion. In reality, the parasitic affects of the sourcing and loading stages can cause unexpected results unless reasonable care is applied to the transforming network. Proper modeling of the impedance environment before committing to any one circuit topology is thus beneficial. The following discussion addresses a variety of alternative solutions for converting an unbalanced single-ended signal to a balanced differential signal.
A balun is a device that converts a BALanced signal to an UNbalanced signal, or vice versa. Baluns can be constructed using a variety of techniques, including magnetic flux coupling or quarter wavelength coupled transmission lines. Such designs are capable of offering broadband performance with very little insertion loss. Transformers are close cousins to wire-wound baluns; both are capable of converting balanced signals to unbalanced signals and transforming impedances to match differing sources and loads. The fundamental difference between wire-wound baluns and transformers is the manufacturing technique.
The classic flux coupled transformer is presented in Figure 1a .
Click to Enlarge Image
Figure 1: Commonly used balun-transformer networks.
a) Flux coupled transformer with optional center tap.
b) Wire-wound transmission-line balun with optional center tap.
c) Three-pole lumped-element lattice balun.
A transformer has primary- and secondary coils. The two coils are DC isolated, and the secondary coil may or may not offer multiple taps. The taps on the secondary can function as return paths in applications such power transformers where multiple loads must be driven, or for feeding DC bias through a center tap—a common practice in RF circuit interfacing. A transmission-line balun, Figure 1b , presents a DC connection between primary and secondary ports, and does not typically provide multiple secondary signal taps. This can be problematic in designs where the output signals need to be level shifted to a different common-mode level. External AC-coupling caps and bias chokes may be required if a common-mode level translation is required when using a transmission-line balun.
Narrowband lump element solutions
In many communication systems, narrow-band solutions are acceptable due to the reasonable fractional bandwidth requirements. Figure 1c presents a lumped element lattice-balun design comprising low-pass and high-pass filter networks that provide 180-degree phase difference between the two outputs. The filter polynomials are selected such that the frequency at which the 180-degree phase difference is achieved corresponds to a half power –3-dB magnitude response. This provides an equal split of the unbalanced input signal power, resulting in low insertion loss and excellent magnitude balance. The 3-pole lattice balun design offers ≈20% fractional bandwidth with ±1dB or better amplitude balance.
Higher-order designs that can feasibly increase the useful fractional bandwidth are possible, but in practice they result in greater insertion loss and thus become unattractive when compared to wire-wound balun-transformer implementations. The main tradeoffs in selecting a lattice balun solution are the number of lumped elements required, the necessary bandwidth of operation, and the actual implementation issues that arise due to component values and circuit layout parasitics.
Coping with PCB parasitics
In any balun design it is important to pay careful attention to circuit board parasitics. Figure 2 demonstrates how board parasitics could cause unexpected results if ignored. We begin with the idealized solution presented in Figure 2a .
Click to Enlarge Image
Figure 2: Improved lumped-element lattice balun design example.
a) Ideal parasitic-free lumped-element balun design for a target frequency of 1120MHz.
b) Effects of practical board parasites on magnitude and phase balance.
c) Corrected design where parasitics have been accounted for, thus restoring decent magnitude and phase balance.
The following equations are used to solve for the ideal inductor and capacitor values for a given source and load impedance at the desired target frequency.
Where RS is the source resistance, RL is the load resistance and fc is the target center frequency.
With 1120-MHz center frequency, 50 Ω source, and 100 Ω balanced load, we find the ideal inductance to be 10.0 nH, and the ideal capacitance to be 2.0 pF. The differential balance is better than 0.5 dB over a 300-MHz frequency range. Next we consider the effects of a practical PCB layout, designed using FR-4 material (er =4.6) with a 9-mil thick dielectric, 50-Ω micro-strip transmission lines, and 30 mil x 30 mil component pads. Figure 2b illustrates how the PCB parasitics disrupt the amplitude and phase balance, moving the center frequency and limiting the bandwidth. The tuning capabilities available in many modern circuit simulation packages make it relatively easy to find a practical solution that accounts for PCB parasitics. Increasing the capacitance to 2.7 pF compensates the design for the PCB layout parasitics and yields a solution with better than ±1dB of amplitude imbalance over a ≈200-MHz frequency range.
Design simulation tools can be very helpful in providing accurate analysis of PCB effects on microwave networks. Determining the optimum network would involve a painful trial-and-error process if reasonable simulation tools capable of capturing PCB circuit trace effects were not available.
Broadband active balun solutions
In broadband applications, a solution that does not rely on reactive components such as lumped element baluns, electromagnetically coupled transformers and coupled transmission line baluns is desirable. The active balun implementations depicted in Figure 3 can provide broadband single-ended to differential conversion with the added benefits of signal gain and good reverse isolation.
Click to Enlarge Image
Figure 3: Active single-ended to differential circuit implementations.
a) Simple common emitter follower single-ended to differential buffer.
b) Differential-pair configured for single-ended to differential conversion.
c) Homebrew single-ended to differential amplifier using high-speed op-amps.
d) Integrated differential amplifier configured for single-ended to differential conversion.
One of simplest active balun circuit implementations is shown in Figure 3a , where a common-emitter amplifier is used to convert a single-ended signal to a differential signal.
Such a simplified topology is not very practical, however, because the common mode levels presented at the outputs differ by the collector to emitter voltage, and the available signal swing of each output is somewhat limited. As a result, the harmonic distortion performance will not be adequate for most analog signal balancing applications. The classic differential pair using a single-ended input drive, Figure 3b , offers better output headroom and equal common-mode output levels, but the circuit has limited output drive capabilities and may require a differential buffer stage to improve output linearity when attempting to drive reasonable output load impedances.
Figure 3c depicts a homebrew differential amplifier incorporating two op-amps. The first is configured as a non-inverting amplifier whose output is inverted by the second, which is configured as an inverting amplifier, thus providing a balanced output. This implementation provides good amplitude and phase balance with an accurate translation of input to output common mode levels, but it can be cumbersome to implement due to the large number of components.
A more integrated high performance broadband solution is presented in Figure 3d . The AD8352 differential amplifier is used to balance a single-ended input signal, while providing useful signal gain. The AD8352 has internal feedback networks connecting the outputs to the gain setting resistor nodes, RGP and RGN . By connecting a 200-Ω external-balancing resistor from the VIP input node to the RGN gain-setting node, the output signal can be balanced over a very broad frequency range. This can be especially useful in broadband digitizer designs where it is desirable to balance a single-ended signal and drive a differential high-performance ADC.
Figure 4 illustrates a fully differential, digitally-controlled variable-gain amplifier configured for a single-ended to differential-signal conversion.
Click to Enlarge Image
Figure 4: Variable gain active-balun solution using the AD8370 configured for single-ended to differential conversion, RL = 1 kΩ
The unused input is simply AC-coupled through a sufficiently large bypass capacitor. This allows the full difference voltage to be amplified and delivered to the output with good amplitude and phase balance. At the higher gain settings, the differential balance is much better than most practical passive balun solutions. The active balun solution also provides signal gain, good load-driving capability, and excellent output-to-input isolation. Such a configuration can be very useful in wireless communication receiver designs where maintaining a constant output power level is desirable, even though the received signal strength as measured at the antenna port varies.
About the author
Eric Newman is an applications systems engineer in the RF and Wireless Products Group at Analog Devices Inc., www.analog.com. He received his Masters of Science in Electrical Engineering in 1993, focusing on wireless communications, from the University of Massachusetts at Lowell.