Differential Circuit Design Techniques for Communication Applications



One of the major challenges in communications system design is the successful capture of signals with adequate fidelity. Strict standard specifications call for proper interface topology selection. Deployment of communication standards for cellular telephony, such as code division multiple access (CDMA) and wideband CDMA, require high dynamic range, high input linearity and low noise in order to avoid the effects of blockers, signal distortion, and sensitivity degradation. In the past, the performance advantages of fully differential signal chains were trumped by single-ended options due to practical application issues. However, recent advances in integrated RF circuit technology and expansions of available high-performance differential RF building blocks allow differential architectures to be applied to high-performance receiver designs.


Figure 1: Single-ended signal example. 

A single-ended signal is unbalanced by definition and is measured by the difference between the signal of interest and a constant reference point, normally ground, which serves as the return path for the signal. A problem can be encountered if an error source is introduced to the single-ended signal. Since the ground reference will be unaffected by the injected error, the error is carried forward through the signal. In a single-ended configuration, any variation introduced to the desired signal will be problematic to remove without involving overly complex cancellation techniques. For this reason, single-ended or unbalanced signals are more prone to noise and interference, such as electro-magnetic coupled interference. Also, as will be shown later, unbalanced configurations have higher distortion than balanced circuits.


Figure 2: Differential signal example. 

Differential signals are made up of pairs of balanced signals moving about a reference point at equal but opposite amplitudes. The composite differential signal corresponds to the difference between the positive and negative balanced signals. For example, from two 1 VP-P signals, the result is a composite signal of 2 VP-P. In this instance, if an error source is introduced to the differential signal path, it will likely be added to each of the two balanced signals equally. Since the return path is not a constant reference point, the differential signal will be unaffected once the difference of the two balanced signal components cancels out the error, which typically is equal in amplitude on each signal transition. For this reason the balanced signal is less prone to noise and interferences than the unbalanced signal would be. Also, as will be discussed, balanced signals have lower distortion than single-ended circuits.


Figure 3: Traditional receiver architecture. 

Shown here is the block diagram of a traditional super-heterodyne receiver. Regardless of the topology, single-ended or differential, the system goal is to successfully deliver a desired signal to an analog-to-digital converter for digitization. The signal path consists of several RF blocks: the antenna, filters, low noise amplifier (LNA), mixer, ADC driver amplifier, and the ADC.

The first block after the antenna is the LNA which is tasked with amplifying the signal above the thermal noise. Amplification at this stage is critical as it will determine the sensitivity of the system and will ensure that subsequent mixers and amplifiers after the LNA do not add significantly to the noise floor. Along the way there are band pass filters to suppress any out-of-band content and reduce distortion or noise that the receiver stages may add along the signal path. The next block, the mixer, which follows the LNA, frequency translates the signal of interest, down-converting the high-frequency RF signal to a lower, more manageable intermediate frequency (IF). The ADC driver amplifier and the anti-aliasing filter (AAF) prepare the signal to be digitized by the ADC. The driver offers gain and the AAF suppresses anything outside the first Nyquist zone of the ADC, including noise that would be delivered to the ADC input, and out-of -band spurious components still existing in the signal path. Finally, at the end of the analog signal path, the ADC performs the function of digitizing the baseband information.


Figure 4: Communication systems considerations. 

In order to contrast single-ended to differential, there are system-level performance metrics which must be observed to design a good overall system. Some of the key considerations that are prevalent in communication systems have been mentioned, but it is important to have a full view.

What makes for a good radio design? Depending on the application and the architecture, the performance specifications will vary. However, in general there are common considerations that are prevalent in communication systems, such as distortion, noise floor, and dynamic range. Additionally, good sensitivity requires a low noise floor and low clock phase noise. A high input, third-order intercept (IP3) and a high 1 dB compression point (P1dB) are vital for input-signal-level handling capability.

There are a lot of transmissions sharing the air waves. A robust system is needed to process the desired signal, which is generally small and in the presence of other interfering signals, which may be large. Therefore, high sensitivity, input linearity, good selectivity, and high noise immunity to large nearby signals are required in the design of a robust system. Other considerations include low-cost, low-power (particularly for portable devices), and compact size.


Figure 5: Differential advantages. 

There are several advantages to using differential versus single-ended signal chains; reviewed here are the most common. Differential signal chains have an advantage compared to single-ended chains with regard to the output transition. The lower signal voltage on each output means that higher overall signal voltages can be achieved. Therefore, the same overall signal swing can be achieved, compared to a single-ended signal, with lower power consumption. As a result of the greater available output swing: higher overall signal swings can be achieved; the same overall signal swing can be achieved but on a lower power supply; and power dissipation can be reduced.

There are also benefits to system linearity. In very low distortion applications, the headroom of the power supplies can be increased by a factor of two when compared to a single-ended signal. The inherent cancellation of even-order harmonics in a differential system, means that the 2nd, 4th, 6th, and so on harmonics will be quite low compared to the odd-order harmonics. It is important to note that a perfect cancellation cannot be achieved, but there is a clear advantage. The differential architecture also allows some pre-distortion techniques to help reduce odd harmonics. In addition, there is typically an approximate 6 dB improvement in the output 1 dB compression point (P1dB) and OIP3 for the same supply rails. Finally, because the signal's return path is no longer through ground, the signal will be less sensitive to ground noise and interference, which translates to a better common mode rejection ratio (CMRR) with an improved power supply rejection ratio (PSRR). Also, the differential approach provides improved immunity to coupled electro-magnetic interference (EMI).


Figure 6: Unbalanced versus balanced signal. 

Shown here are two block diagrams contrasting the single-ended and differential approaches. The first diagram represents the single-ended block diagram with a small, one-sided input signal. The blue signal illustrates any sort of common-mode interference coming into the system. Note that this blue interference signal is also amplified at the output – it is amplified just as much as the desired signal. It is difficult to separate the desired signal from the interference signal.

The differential block diagram shows the desired signal made up of two signals of opposite polarity, one is positive and the other is negative. Any interference introduced at the inputs would be the same polarity on both reference levels, as illustrated in blue. Although the interference signals are amplified at each output, when looking at the composite signal, the difference between the two differential signals, the desired signal is doubled, and the interference has been removed. While the single-ended approach is susceptible to common-mode noise, power supply noise, or EMI, the differential block has an immunity to those interferences by the virtue of cancellation.


Figure 7: Even-order cancellation. 

In addition to common-mode interference immunity, the differential approach also has an inherent cancellation of even-order harmonics. Shown here is a review of the single-ended approach. A non-linear device, in this case a single-ended amplifier, is described by a power-series expansion transfer function and is supplied a sinusoid at its input. The expansion of the power series (the equation at the bottom) reveals that a constant is tied to each frequency multiple, even and odd.


Figure 8: Output spectrum chart. 

Illustrating this equation makes it easier to visualize its various components. The portion of the equation representing the fundamental signal is highlighted in grey. The portions representing the second and third harmonics are highlighted in pink and green, respectively. The expansion of the power series reveals that a non-zero constant is tied to each frequency multiple, even and odd. It is clear that the single-ended, non-linear device has created harmonics throughout the spectrum, as would be expected.


Figure 9: Cancellation effect of the differential block. 

Taking the same mathematical approach to look at the differential approach there can be seen an inherent cancellation of even-order harmonics, a benefit similar to the previously discussed common-mode interference immunity. Again, a non-linear device, in this case a differential amplifier block, is described by a power-series expansion transfer function and is supplied with a pair of sinusoids of opposite polarities – these represent a differential signal at the device’s inputs. Through expansion, the cancellation effect of the differential block is shown.


Figure 10: Output spectrum of nonlinear device. 

Again, illustrating this equation makes it easier to visualize its various components. The expansion of the power series reveals that all even-order harmonics in the series are cancelled by their counterparts of opposite magnitude. Only the fundamental signal, highlighted in grey, and the third order harmonic, highlighted in green, have a non-zero contribution. In real life, non-ideal devices will not achieve perfect cancellation but they do benefit from lower even-order harmonics.


Figure 11: Challenge of driving the ADC. 

One of the major challenges in communications system design is successfully driving the desired signal into the analog-to-digital converter. The examples shown here help to illustrate the benefits of differential signal chains versus single-ended signal chains. The three major blocks shown here are the driver amplifier, the anti-aliasing filter, and the analog-to-digital converter. Adequate fidelity for signal detection requires proper component selection and implementation of the interface.

Reviewed here (Figure 11) are two examples, one-single ended and one differential. The goal is to capture the blue portion of the signal transmission on the left hand side. It is a small signal in the presence of other, larger, interfering signals. To capture it, it is necessary to consider noise, dynamic range, and other factors specific to the ADC requirements. This is all necessary to extract only the signal of interest and deliver it to the ADC. This receiver signal is shown at the right; it has been amplified and the blockers have been removed, leaving only the desired portion of the signal, shown in blue.


Figure 12: Comparison of the single-ended and differential signal chain performance of a communications system. 

Figure 12 compares the single-ended and differential signal chain performance in a real-world example of the receiver end of a communications system. The first example is the single-ended approach which begins with a single-ended IF driver amplifier, followed by a single-ended anti-aliasing filter and is then converted to a differential signal by a transformer at the input to the ADC. Note that in many cases, the single-ended approach is considered the passive approach because the transformer is used to translate the signal to the differential ADC.

The differential example begins with a differential signal at the output of a transformer, a differential ADC driver amplifier, followed by a differential anti-aliasing filter, and the input to the ADC. The differential approach is called active because an amplifier may be used to make the conversion. The performance metrics of each component are listed, however the following figure uses signal chain performance tables with these same metrics to help analyze the cascaded system performance and compare the single-ended and differential approaches.


Figure 13: Signal chain performance table. 

For the single-ended topology, using the equations for cascaded noise figure and IIP3, it is possible to calculate the input referred RF performance. For this example the input referred IIP3 is 18.8 dBm and the noise figure (NF) is 11.4 dB. This results in an SFDR (spurious free dynamic range) of 76 dB for a 5 MHz analysis bandwidth. The cascaded power gain is 14.7 dB with an input referred full-scale of -10.7 dBm.

Using the same equations to calculate the input-referred RF performance for the differential approach results in the following: the input referred IIP3 is 21.5 dBm with a noise figure of 13.7 dB. This results in an SFDR of 76.5 dB for 5 MHz analysis bandwidth, 14 dB of cascaded power gain, and an input-referred full-scale of -10 dBm.

The numbers for both approaches are quite similar. However, the active, differential approach has higher distortion performance, with the noise figure being slightly higher. Also, the spurious free dynamic range is higher with the active configuration. Keep in mind that the input-referred full-scale for the single-ended approach would be only 6 dBm without the IF amplifier. It should also be noted that the differential anti-aliasing filter would require twice as many series elements as the single-ended approach. Still, the passive interface often requires more resistive padding and needs higher output power from an upstream driver, which typically means higher supply current. Also consider that single-ended driver amps tend to have worse even-order distortion and CMRR and PSRR. Therefore by eliminating the ADC driver, amplification needs are being moved upstream. The differential approach would be the logical choice based on overall performance.


Figure 14: ADL5562 key specifications and features. 

The ADL5562 is an example of a device which takes advantage of the even-order cancellation of the differential approach to reduce distortion. It is a high-performance differential amplifier that is optimized for RF and IF applications. This amplifier offers low noise of 2.1 nV/√Hz and low distortion performance over a wide frequency range, making it an ideal driver for high-speed 8-bit to 16-bit ADCs. The ADL5562 provides three gain levels of 6 dB, 12 dB, and 15.5 dB through a pin-strappable configuration. If used in the single-ended input configuration, the gains are reduced to 5.6 dB, 11.1 dB, and 14.1 dB. The device is optimized for wideband, low distortion performance. These attributes, together with its adjustable gain capability, make this a good choice for general-purpose IF and broadband applications where low distortion, low noise, and low power consumption are critical. The ADL5562 is also optimized for a good combination of slew speed, bandwidth, and broadband distortion which allows it to drive a wide variety of ADCs and makes it ideally suited for driving mixers, pin diode attenuators, SAW filters, and multi-element discrete devices.


Figure 15: ADL5562 harmonic distortion graph. 

The plot on Figure 15 shows the resulting second and third-order harmonics when the ADL5562 is configured for single-ended and differential topologies. Though the distortion performance in single-ended mode is very low, there is a clear advantage in the even-order performance with differential operation. In the single-ended mode the ADL5562 has second-order harmonic values of -82 dBc at 100 MHz. In differential operation, the device has lower second-order harmonic values, less than -100 dBc at the same frequency. In addition, approximately 6 dB of improvement on the output 1 dB compression point and OIP3 can be expected from a differential topology on the same supply rails.


Figure 16: Analog Devices differential amplifiers portfolio. 

The ADL5562 is one of many in the Analog Devices broad portfolio of differential amplifiers. Whether the requirement is to drive a differential input ADC, or to send and receive signals over long cable lengths, ADI has a differential amplifier to meet the need. The amplifiers come in fixed gain, or three basic types of gain control: resistor set gain, serial and parallel digital control, or pin-strappable gain selection.


Figure 17: AD8375/6 key specifications and features. 

The AD8375 is a digitally controlled, variable gain wide bandwidth amplifier that provides precise gain control, high OIP3, and a low noise figure. The AD8376 is a dual-channel version of the AD8375; two VGAs in a single 5 mm x 5 mm package. Both offer excellent distortion performance and high-signal bandwidth, making them a good gain control device for a variety of receiver applications. These VGAs provide a broad 24 dB gain range with 1 dB resolution. Using an advanced high-speed SiGe process and incorporating proprietary distortion cancellation techniques, the AD8376 achieves 50 dBm output OIP3 at 200 MHz.


Figure 18: Evolving differential receiver architecture. 

Receivers are evolving and are increasingly using differential components. This evolution began at the ADC and is gradually moving up the signal chain. In the past, signal application issues and limited high performance differential RF building block availability lead to single-ended or partially differential signal chains. As discussed earlier, one example of a partial differential signal chain is the designer’s choice to omit a differential ADC driver and use single-ended devices up to a transformer to the ADC. While this offers a simple solution, the performance demands are simply pushed upstream. In addition to consuming higher power, single-ended driver amplifier solutions tend to have worse even-order distortion, CMRR and PSRR.

The receiver signal chain shown here is an architecture commonly used for receivers with single-ended RF input and differential output. The dividing line between single-ended and differential operation seems to have settled at the mixer, with RF components such as the LNA still being offered as single-ended components. Most SAW filters and mixer cores are also natural differential circuits, but they are converted to single-ended due to application limitations. For years doubly balanced mixer topologies have been adopted for cellular applications due to their highly linear performance. Unfortunately, the traditional transformer networks used to couple the signals to the mixing core, to keep the system differential, consume considerable board space and add significant cost to the design.


Figure 19: ADL5355 key specifications and features. 

Recent advances in integrated RF circuit technology have allowed the design of easy-to-use RF blocks with single-ended RF input to differential IF output. The Analog Devices ADL5355 mixer is an example of a device which mixes a single-ended RF input into a differential IF output. It allows for the inherent differential advantage to be maintained in a stubborn single-ended world. The ADL5355 block diagram shows that all three internal mixer ports are differential. For ease of use, the RF and LO ports are connected to the outside world using transformers, allowing for a single-ended interface. By contrast, the IF output port, which includes a driver amplifier is differential, with an output impedance of 200 ohms, to facilitate connection to a differential SAW filter. The integration of the LO and RF baluns (balanced to unbalanced transformers) limits the operating frequency range of the device, requiring a family of devices specified for operation across the cellular frequency ranges. As a result, the input frequency range of this device is limited to the range of 1200 MHz to 2500 MHz with a low-side LO; that is, an LO frequency which is always less than the RF frequency.


Figure 20: Analog Devices high linearity receiver portfolio. 

Since the integration of baluns limits the operating frequency of mixers, ADI provides a large family of devices which cover popular frequencies. Notice that some of the devices in the Figure 20 table have very broad frequency ranges. These devices do indeed have broad frequency ranges but they require external baluns on their LO and RF ports for optimum performance.


Figure 21: Radio design requirements and differential benefits. 

In summary, the differential performance advantage is a necessity for modern communications systems whose strict standard specifications call for uncompromised performance in the face of blockers, signal distortion, and sensitivity degradation. To meet this need, the evolving differential signal path offers high dynamic range, high input linearity and low noise, among other benefits. Analog Devices offers a broad portfolio of differential RF components brought forth by advances in integrated RF circuit technology. The increased availability of high-performance differential RF building blocks allows differential architectures to be applied for high-performance receiver designs in the most stringent wireless cellular applications.

References
  1. For a full audio version of this training please visit this link.
Supplier