Analysis of relevant technologies for differential circuit design in communication applications

Successfully capturing signals with sufficient fidelity is a major challenge in communication system design. Stringent standards and specifications require the selection of appropriate interface topologies. This panel discusses the advantages of differential design techniques and how their performance advantages impact stringent system requirements in today's high-performance communication systems. In addition, the definition of RF will be reviewed, system budgets will be outlined, and different implementation methods will be compared.

What are the relevant technologies for differential circuit design in communication applications? First, a comparison of single-ended and differential signals will be made, and then some factors that need to be considered in the receiver signal chain and system performance will be briefly introduced, and then the advantages of differential applications will be discovered. From the perspective of driving ADCs, compared with single-ended applications, we will find that differential applications are more likely to achieve higher data rates. Finally, we will return to the system design level and summarize the benefits of differential applications.

Single-ended and differential signals

First, let's talk about the concepts of single-ended and differential signals, which everyone is familiar with. Here we use another way to express it. We can divide signals into unbalanced signals or balanced signals. Single-ended signals belong to unbalanced signals. Because it is a single-sided signal, it is relatively speaking, and there is no signal pair to balance it. Compared with balanced signals, unbalanced signals generally produce higher harmonic distortion.

Differential signals are balanced signals. The differential pairs generally have a common common-mode level and a differential-mode level with the same amplitude. When measuring differential signals or balanced signals, we are concerned with the difference between the positive and negative input signals. The harmonic distortion caused by this balanced signal is relatively small.

System-level design

On the other hand, when applied in communication systems, we see a relatively common signal chain of a superheterodyne receiver. Figure 1 shows the signal chain of a common superheterodyne receiver. A low-noise amplifier is connected after the antenna to amplify the signal and suppress noise. Then a two-stage mixer is used to down-convert the signal to a lower frequency. During this period, we will add appropriate filters to filter out noise and harmonics outside the useful signal band. After that, the buffer amplifier drives the ADC. This is the main issue we will discuss today. The main purpose of this level of op amp is to adjust the signal level range and improve the driving capability. Sometimes it is also used as a conversion between single-ended and differential. Before entering the ADC, we need to add an anti-aliasing filter, and finally use the ADC to perform analog-to-digital conversion on the baseband signal. We can see that if the system wants to achieve a higher dynamic range, it cannot introduce too much noise and harmonics in addition to the signal.

Figure 1. Generic superheterodyne receiver signal chain

Let's take a closer look at the performance and indicators that are worth noting in a communication system. Before we compare single-ended signals and differential signals, we need to understand some issues that need to be considered in system-level design.

So, what kind of design is a better RF system design? First of all, the signal sensitivity should be high, which means lower noise, and the phase noise introduced by the clock should also be low. The input signal should have sufficient driving capability and relevant indicators, such as high third-order intercept point and 1dB compression point. Then it is whether the performance of each module is good enough, whether it can better distinguish between signals and noise, whether the linearity is good enough, etc. In addition, there are considerations such as low power consumption and low cost.

We say that the differential signal chain has many advantages over the single-ended signal. Because it is a differential signal, the output is two differential signals. In fact, the amplitude of the output differential signal is relatively doubled. From another perspective, under the same output range, the operating voltage will be lower. In this way, in applications requiring low harmonic distortion, sufficient amplitude margin can be guaranteed. The odd-function-like characteristics of the differential system itself can eliminate the even-order harmonics in the system, that is, the 2nd, 4th, 6th harmonics, etc. The harmonics at these frequencies are relatively small or even invisible compared to the odd-order harmonics. Finally, since the return path of the signal is no longer the ground plane, the signal is not so sensitive to the ground plane or the power plane, thereby reducing the coupling introduction of noise and achieving better anti-electromagnetic interference effect.

As shown in Figure 2, the single-ended signal is more sensitive to common-mode noise, power supply noise and electromagnetic interference, and the op amp will amplify these noises to a certain extent. The differential signal, however, suppresses common-mode noise and interference because the signals on both sides form a current loop, and only effectively amplifies the differential signal.

Through derivation, we can also see the odd-order characteristics of differential amplification. Ideally, we can only see the fundamental and odd-order harmonics on the spectrum. Here we only give the conclusion. It is worth noting that the third harmonic and the third-order intercept point caused by it. IP3 is the theoretical input power at the intersection of the fundamental and third-order distortion output curves. It is an important indicator to describe the linearity of the amplifier:

in the design of communication systems, driving, extracting and loading useful signals to the ADC input is a critical issue. For high-precision system design, it is required to make appropriate selections of devices and interface methods. We will give you a few examples, but before that, please understand that as shown in Figure 3, we want to extract the useful signal in the blue part, which has very little energy and is also interfered by surrounding signals and noise. In order to capture it, we must pay attention to noise, dynamic range, and other ADC-related indicators, which will be explained in detail in the following slides. We can see that the main modules for functional implementation include buffer op amps, anti-aliasing filters, and ADCs.

Figure 2 The difference between single-ended and differential signals



Figure 3 Useful signals and noise

Figure 4 is an example of a single-ended op amp with single-ended input. You can see the signal gain, input and output third-order intercept power, and noise coefficient of the four stages of the intermediate frequency amplifier, anti-aliasing filter, transformer, and ADC. The single-ended signal is converted to a differential signal before the ADC using a passive transformer. It should be noted here that the terminal matching impedance of the ADC is assumed to be 200Ω, and since the previous stages are all 50Ω characteristic impedance, the impedance ratio of the transformer is set to 1:4.

If the transformer is moved forward and the signal is converted to a differential signal before the op amp, the single-ended op amp is replaced with a differential op amp, thus forming a fully differential structure. As shown in Figure 5.

Here we will talk about the equivalent calculation of the overall noise coefficient of the cascade system and the input and output third-order intercept point. When considering the overall noise coefficient, the first stage has the greatest impact; and when considering the intercept index, the last stage has the most obvious impact.

Considering the relationship between spurious free dynamic range and the third-order intercept point of the system, we know that as the input signal energy increases, the third-order intermodulation distortion and the noise floor are exactly equal, and the system reaches the maximum SFDR. At this time, it can be expressed by this formula: SFDR = (2/3)(IIP3-NF-10log(TERMAL NOISE).

So we can calculate the overall signal gain, third-order intercept point, noise figure and spurious-free dynamic range of the two single-ended to differential conversion methods mentioned above. From the indicators, they are not much different. The overall distortion and noise figure of the differential active drive structure are slightly higher, but the SFDR performance is also higher. In addition, it should be noted that in the single-ended passive conversion structure, if the intermediate frequency amplifier is removed, the full-scale reference input power is 6dBm, and the anti-aliasing filter design is an asymmetric structure. Moreover, more resistive matching devices must be added to the entire design, which requires the front-stage driver to have strong capabilities, that is, large current and power consumption. In addition, the even harmonics, common-mode rejection, and power supply rejection problems of the single-ended op amp will also affect the performance of the overall system to a certain extent.

On the other hand, when transmitting data, it can be transmitted bit by bit, or it can be divided into symbols for transmission, such as two bits per symbol, and then they are respectively mapped to 4 phases, and then acted on The signal is transmitted on the carrier. This is a very common modulation mode, namely QPSK.

Usually, we can use constellation diagrams to describe different modulation methods. We know that higher-order modulation can be used in transceivers with higher data rates, but at the same time, it requires lower local oscillator leakage, better power amplifier linearity, higher system bandwidth and demodulator signal-to-noise ratio. On the one hand, ADI is also developing higher-performance products to meet customer needs. On the other hand, we should also pay attention to discovering the principles of the problem during system design and use appropriate methods and techniques to solve it.

In Figure 6, we can see the impact of noise and harmonics in the receiving system on the error vector magnitude EVM. In other words, the demodulated signal will be offset from the ideal constellation position. Generally, we use the error vector magnitude to measure it. Too large an error vector magnitude will cause symbol errors and worsen the bit error rate. Especially in high-order modulation methods, the symbols are closer and the requirements for the error vector magnitude are more stringent.

Figure 4 Example of single-ended input and single-ended output


Figure 5 Example of fully differential structure


Figure 6 Effect of noise and harmonics in the receiving system on error vector magnitude (EVM)

From this we can conclude that higher-order modulation has a higher data rate and better EVM, and better EVM means a higher spurious-free dynamic range SFDR, which is related to the signal-to-noise ratio, intermodulation distortion and each harmonic term. Therefore, to improve the above performance indicators, the use of balanced signals and differential structures can be significantly improved.

Summary

Finally, for a good RF system, the main focus is on how to improve the sensitivity to useful signals, so as to better separate the signals from noise, harmonics and various interferences. The benefits of differential applications are better common-mode rejection, power supply rejection, anti-electromagnetic interference capability, better linearity and a larger dynamic range than single-ended signals under the same conditions. Undoubtedly, the differential structure has obvious advantages and is more suitable for high-performance RF systems.