PCB Technical - Four of High-speed PCB Design Guidelines: Crosstalk Control of High-speed Digital Systems

Content: In high-frequency circuits,crosstalk may be the most difficult to understand and predict, but it can be controlled or even eliminated.

As the switching speed increases, modern digital systems have encountered a series of problems, such as signal reflection, delayed fading, crosstalk, and electromagnetic compatibility failures. When the switching time of the integrated circuit drops to 5 nanoseconds or 4 nanoseconds or less, the inherent characteristics of the printed circuit board itself begin to appear. Unfortunately, these features are harmful and should be avoided as much as possible during the design process. In high-frequency circuits, crosstalk may be the most difficult to understand and predict, but it can be controlled or even eliminated.

1. What causes crosstalk?

When the signal propagates along the wiring of the printed circuit board, its electromagnetic wave also propagates along the wiring, from one end of the integrated circuit chip to the other end of the line. In the process of propagation, electromagnetic waves cause transient voltages and currents due to electromagnetic induction.

Electromagnetic waves include electric and magnetic fields that change over time. In the printed circuit board, in fact, the electromagnetic field is not limited to various wiring, a considerable part of the electromagnetic field energy exists outside the wiring. Therefore, if there are other lines nearby, when the signal propagates along a wire, its electric and magnetic fields will affect other lines. According to Maxwell's equation, time-varying electricity and magnetic fields will cause adjacent conductors to generate voltages and currents. Therefore, the electromagnetic field accompanying the signal propagation process will cause adjacent lines to generate signals, which leads to crosstalk.

In printed circuit boards, the lines that cause crosstalk are often called "intruders." The line subject to crosstalk interference is usually called the "victim". The crosstalk signal in any "victim" can be divided into forward crosstalk signal and backward crosstalk signal, these two signals are partly caused by capacitive coupling and inductive coupling. The mathematical description of the crosstalk signal is very complicated, but, like a high-speed boat on the lake, some quantitative characteristics of the forward and backward crosstalk signals can still be understood by people.

High-speed boats have two effects on the water. First, the speedboat stirs up waves on the bow, and the arc-shaped ripples seem to move forward with the speedboat; secondly, when the speedboat travels for a period of time, it will leave long water trails behind it.

This is very similar to the reaction of the "victim" when the signal passes through the "intruder". There are two types of crosstalk signals in the "victim": the forward signal before the intrusion signal, like the water and ripples on the bow of the ship; the backward signal, which is behind the intrusion signal, like the water trail in the lake after the ship sails away. .

2. Capacitance characteristics of forward crosstalk

Forward crosstalk manifests as two interrelated characteristics: capacitive and perceptual. When the "invasion" signal advances, a voltage signal of the same phase is generated in the "victim". This signal has the same speed as the "invasion" signal, but it is always before the "invasion" signal. This means that the crosstalk signal will not propagate in advance, but will be coupled with more energy at the same speed as the "intrusion" signal.

Since the change of the "intrusion" signal causes the crosstalk signal, the forward crosstalk pulse is not unipolar, but has both positive and negative polarities. The pulse duration is equal to the switching time of the "intrusion" signal.

The coupling capacitance between the wires determines the amplitude of the forward crosstalk pulse, and the coupling capacitance is determined by many factors, such as the material of the printed circuit board, the geometric size, the position of the line crossing, and so on. The amplitude is proportional to the distance between parallel lines: the longer the distance, the greater the crosstalk pulse. However, there is an upper limit to the amplitude of the crosstalk pulse, because the "intrusion" signal gradually loses energy, and the "victim" in turn couples back to the "invader". Inductance characteristics of forward crosstalk

When the "intrusion" signal propagates, its time-varying magnetic field will also produce crosstalk: forward crosstalk with inductive characteristics. But perceptual crosstalk and capacitive crosstalk are obviously different: the polarity of forward perceptual crosstalk is opposite to that of forward capacitive crosstalk. This is because in the forward direction, the capacitive and perceptual parts of crosstalk are competing and canceling each other out. In fact, when the forward capacitive and perceptual crosstalk are equal, there is no forward crosstalk.

In many devices, the forward crosstalk is quite small, and the backward crosstalk becomes a major problem, especially for long strip circuit boards, because the capacitive coupling is enhanced. However, without simulation, it is practically impossible to know to what extent perceptual and capacitive crosstalk cancel out.

If you have measured forward crosstalk, you can determine whether your trace is capacitively coupled or inductively coupled based on its polarity. If the crosstalk polarity is the same as the "intrusion" signal, capacitive coupling will dominate, otherwise, inductive coupling will dominate. In printed circuit boards, inductive coupling is usually stronger.

The physical theory of backward crosstalk is the same as that of forward crosstalk: the time-varying electric and magnetic fields of the "intrusion" signal cause perceptual and capacitive signals in the "victim". But there are also differences between the two.

The biggest difference is the duration of the backward crosstalk signal. Because the propagation direction and speed of forward crosstalk and "intrusion" signals are the same, the duration of forward crosstalk is the same as that of "intrusion" signals. However, backward crosstalk and the "intrusion" signal propagate in the opposite direction, it lags behind the "invasion" signal and causes a long train of pulses.

Unlike the forward crosstalk, the amplitude of the backward crosstalk pulse has nothing to do with the line length, and its pulse duration is twice the delay time of the "intrusion" signal. why? Suppose you observe backward crosstalk from the starting point of the signal. When the "intrusion" signal is far away from the starting point, it is still producing backward pulses until another delayed signal appears. In this way, the entire duration of the backward crosstalk pulse is twice the delay time of the "intrusion" signal.

3. The reflection of backward crosstalk

You may not care about the crosstalk interference between the driver chip and the receiver chip. However, why should you care about backward pulses? Because the driver chip is generally low-impedance output, it reflects more crosstalk signals than it absorbs. When the backward crosstalk signal reaches the driver chip of the "victim", it will be reflected to the receiving chip. Because the output resistance of the driver chip is generally lower than the wire itself, it often causes the reflection of the crosstalk signal.

Unlike the forward crosstalk signal, which has two characteristics: inductive and capacitive, the backward crosstalk signal has only one polarity, so the backward crosstalk signal cannot cancel itself. The polarity of the backward crosstalk signal and the crosstalk signal after reflection is the same as the "intrusion" signal, and its amplitude is the sum of the two parts.

Remember, when you measure the backward crosstalk pulse at the receiving end of the "victim", this crosstalk signal has already been reflected by the "victim" drive chip. You can observe that the polarity of the backward crosstalk signal is opposite to the "intrusion" signal.

In digital design, you often care about some quantitative indicators. For example, no matter how and when crosstalk is generated, forward or backward, its maximum noise tolerance is 150mV. So, is there a simple way to accurately measure noise? The simple answer is "no", because the electromagnetic field effect is too complicated, involving a series of equations, the topology of the circuit board, the analog characteristics of the chip, and so on.

4. Line length

Many designers believe that shortening the line length is the key to reducing crosstalk. In fact, almost all circuit design software provides the maximum parallel line length control function. Unfortunately, it is difficult to reduce crosstalk only by changing the geometric value.

Because the forward crosstalk is affected by the coupling length, when you shorten the length of the line that has no coupling relationship, there is almost no reduction in crosstalk. Furthermore, if the coupling length exceeds the drop or rise time delay of the driver chip, the linear relationship between the coupling length and the forward crosstalk will reach a saturation value. At this time, shortening the already long coupling line has little effect on reducing crosstalk.

A reasonable method is to expand the distance between the coupled lines. In almost all cases, separating the coupled lines can greatly reduce crosstalk interference. Practice has proved that the backward crosstalk amplitude is roughly inversely proportional to the square of the distance between the coupled lines, that is, if you double the distance, the crosstalk will be reduced by three quarters. This effect is more pronounced when backward crosstalk is dominant.

5. Crosstalk cancellation

From a practical point of view, the most important issue is how to remove crosstalk. What should you do when crosstalk affects circuit characteristics?

You can adopt the following two strategies. One method is to change one or more geometric parameters that affect coupling, such as line length, distance between lines, and the layered position of the circuit board. Another method is to use the terminal to change the single line into a multi-channel coupled line. With a reasonable design, the multi-line terminal can cancel most of the crosstalk.

6. Difficulty of isolation

It is not easy to increase the distance between the coupled lines. If your wiring is very dense, you must spend a lot of effort to reduce the wiring density. If you are worried about crosstalk interference, you can add one or two isolation layers. If you have to expand the distance between lines or networks, then you'd better have a software that is easy to operate. The width and thickness of the circuit also affect the crosstalk interference, but its influence is much smaller than the distance factor of the circuit. Therefore, these two parameters are generally rarely adjusted.

Because the insulating material of the circuit board has a dielectric constant, it will also generate coupling capacitance between the lines, so reducing the dielectric constant can also reduce crosstalk interference. This effect is not very obvious, especially the microstrip circuit part of the dielectric is already air. More importantly, changing the dielectric constant is not so easy, especially in expensive equipment. A workaround is to use more expensive materials instead of FR-4.

The thickness of the dielectric material affects the crosstalk interference over a large length. Generally, making the wiring layer close to the power layer (Vcc or ground) can reduce crosstalk interference. The exact value of the improvement effect needs to be determined by simulation.

7. Stratification factors

Some printed circuit board designers still do not pay attention to the layering method, which is a major mistake in high-speed circuit design. Layering not only affects the performance of the transmission line, such as impedance, delay and coupling, but also the circuit operation is prone to malfunction or even change. For example, it is impossible to reduce crosstalk interference by reducing the dielectric thickness of 5mil, although it can be done in terms of cost and process.

Another factor that is easy to overlook is the choice of layers. In many cases, forward crosstalk is the main crosstalk interference in microstrip circuits. However, if the design is reasonable, the wiring layer is located between the two power layers, so that capacitive coupling and inductive coupling are well balanced, and backward crosstalk with a lower amplitude becomes the main factor. Therefore, you must pay attention to what kind of crosstalk interference dominates during simulation.

The positional relationship between the wiring and the chip also affects crosstalk. Because the backward crosstalk reaches the receiving chip and is reflected to the driver chip, the location and performance of the driver chip are very important. Because of the complexity of the topology, reflections and other factors, it is difficult to explain who is mainly affected by crosstalk. If there are multiple topological structures to choose from, it is best to use simulation to determine which structure has the least impact on crosstalk.

A non-geometric factor that may reduce crosstalk is the technical indicators of the driver chip itself. The general principle is to choose a driver chip with a long switching time to reduce crosstalk interference (the same is true for solving many other problems caused by high speed). Even if the crosstalk is not strictly proportional to the switching time, reducing the switching time will still have a significant impact. In many cases, you cannot choose the driver chip technology, you can only change the geometric parameters to achieve your goal. Reduce crosstalk through the terminal

As we all know, an independent, uncoupled transmission line terminal is connected to match the impedance, it will not produce reflection. Now consider a series of coupled transmission lines, for example, three transmission lines with crosstalk with each other, or a pair of coupled transmission lines. If you use circuit analysis software, you can derive a pair of matrices, representing the transmission line itself and the capacitance and inductance between each other. For example, three transmission lines may have the following C and L matrices:

In these matrices, the diagonal elements are the values of the transmission lines themselves, and the off-diagonal elements are the values between the transmission lines. (Note that they are expressed in pF and nH per unit length). A sophisticated electromagnetic field tester can be used to determine these values.

It can be seen that each group of transmission lines also has a characteristic impedance matrix. In this Z0 matrix, the diagonal element represents the impedance value of the transmission line to the ground, and the off-diagonal element is the coupling value of the transmission line.

For a group of transmission lines, similar to a single transmission line, if the terminal is an impedance matrix matched with Z0, its matrix is almost the same. The required impedance does not have to be the value in Z0, as long as the formed impedance network matches Z0. The impedance matrix includes not only the impedance of the transmission line to ground, but also the impedance between the transmission lines.

Such an impedance matrix has good properties. First, it can prevent the reflection of crosstalk in uncoupled lines. More importantly, it can eliminate the crosstalk that has been formed.

8. Lethal weapons

Unfortunately, such a terminal is expensive and impossible to achieve ideally, because the coupling impedance between some transmission lines is too small, which will cause a large current to flow into the driver chip. The impedance between the transmission line and ground cannot be too large to drive the chip. If these problems exist and you plan to use this type of terminal, try adding a few AC coupling capacitors.

Although there are some difficulties in implementation, the impedance array terminal is still a lethal weapon to deal with signal reflection and crosstalk, especially for harsh conditions. In other environments, it may or may not work, but it is still a recommended method.