Precision PCB Fabrication, High Frequency PCB, High Speed PCB, Standard PCB, Multilayer PCB and PCB Assembly.
The most reliable PCB custom service factory.
PCB News
About differential line transmission
PCB News
About differential line transmission

About differential line transmission

2021-11-09
View:37
Author:Kavie

I would like to ask you, differential line transmission requires tight coupling, will it cause crosstalk?


pcb


Will cause crosstalk, so the differential pair must be matched,
The tighter the coupling, the greater the requirement for matching.
While achieving good matching, the tighter the coupling, the stronger their anti-interference to the outside world
Because external interference affects them at the same time, and a differential pair focuses on the difference between the two signals, it is immune to external common-mode noise.

Crosstalk is coupled by the coupling capacitance between the lines when the signal jumps. What is the principle of the matching of the differential pair to eliminate the crosstalk? Are there any specific instructions?
Matching is to reduce the impact of crosstalk by eliminating reflections, but not to eliminate crosstalk
Since there is a coupling of capacitance and inductance between the differential pair, there is also crosstalk in theory,
The reflection of forward crosstalk and backward crosstalk is the main influence on the receiving end. Because of the opposite polarity of capacitive and inductive coupling, forward crosstalk can be mostly cancelled (large capacitive coupling between differential lines), while backward crosstalk After the reflection to the matched terminal, there is no total reflection superposition, and its amplitude is also very small. Regarding the specific mutual influence and restriction relationship, I am also thinking about it, please discuss it together

But "forward crosstalk can be mostly cancelled due to the opposite polarity of capacitive and inductive coupling." I don't quite understand, can I draw a schematic diagram?
In addition, in long-distance differential line transmission, such as 485, 422, etc., the reference design is at the receiving end
Connecting a 120 ohm resistor is a way to match the differential line to eliminate reflections?
That 120 ohm is the differential matching resistance, which is a simplified form of differential matching

There are many debates on whether differential pairs must be coupled, and many experts have different opinions on this issue.
The main purpose of differential pair coupling is to enhance the anti-interference ability to the outside world and suppress EMI.
If it can be ensured that all the surrounding traces are far away from the differential pair (for example, far greater than 3 times the line width), then the differential traces do not need to ensure tight coupling. The most important thing is to ensure that the trace lengths are equal. (You can refer to the explanation of differential wiring on Johnson's signal integrity website. He asked his layout engineer to keep the differential wiring far away so that the wiring can be routed). It's just that most of the current multi-layer high-speed PCB boards have very tight wiring space, and it is impossible to isolate the differential wiring from other wiring, so it is necessary to maintain a tight coupling at this time to increase the anti-interference ability.

In the design of high-speed (above 10G) chip package substrates, a more popular method is to isolate the two differential lines and the outside with copper areas, so that there is no coupling between the differential pairs. This approach can Meet the strict requirements of single trace impedance of 50 ohms and differential impedance of 100 ohms, and lower loss can be obtained at high frequencies (S11 return loss and S21 differential loss parameters).
I don’t think the differential pair must be coupled

Differentials can ignore crosstalk, because their crosstalk results will be canceled when they are finally accepted. In addition, differentials need to be balanced, and parallel is only part of the balance.

I think the coupling of the differential pair should still be needed. For single-line matching, although the theory is very mature, the actual PCB circuit still has an error of about 5% (I haven't done it myself). On the other hand, the differential line can be regarded as a self-loop system, or the signals on its two signal lines are related. Too loose coupling may cause different interference from elsewhere, and for some interface circuits, the equal length of the differential pair is an important factor in controlling the line delay. Therefore, I think the differential line should be tightly coupled.

For most current high-speed PCB boards, maintaining good coupling is beneficial
But I hope you don't mistakenly think that coupling is a necessary condition for differential pairs, as this sometimes limits the design ideas.
When doing high-speed design or analysis, not only do you need to know how most people do it, but also why others do it, and then understand and improve based on the experience of others, and constantly exercise your creative thinking ability

Matching is needed, but the reason for matching is not reflection, but to reduce the degree of cross-winding interference. If the reduction is related to the matching method, if the resistance is connected, it will not be effective, but if the termination matching method of grounding or power supply is adopted, Because the line impedance of the two lines is reduced, the crosstalk is reduced... (Everyone knows that the impedance line is easy to accept external signal interference, because it is easier to absorb)
You can do an experiment, cut the difference line, and then connect it with a flying line.
If the two wires are separated, signal instability will occur.
The performance is sometimes there is no signal at the port.
However, the signal is stable if two wires are used to fly together.
This is an experiment originally done, for reference only

If you put him in a shielded place, the result may be different again.

In order to ensure more reliable signal transmission, in the design, each signal between the two boards is connected by a two-point two-wire connection, that is, a pair of four signal wires. This will cause problems, and the receiving end cannot normally accept and form the correct signal. Only the difference signal will have this problem, the ttl level will not have this problem. how to explain?
1. I think that the so-called tight coupling of differential lines does not mean that the closer the line is, the better. Too close is useless, because differential transmission is also related to the ground.
2. The forward and backward crosstalk mentioned by fenix is roughly like this. Forward crosstalk is divided into capacitive and inductive. The impedances of the two are 1/jwc and jwl respectively. The phases are exactly opposite. The time to reach the end point is the same, and the magnitude is similar, so it can be considered as cancellation. The backward crosstalk has to last 2Tpcb and cannot be cancelled. You can consider using a resistor to absorb it at the terminal

The differential signal itself does not require single-line impedance, because as long as the + and-signals are symmetrical, the signal integrity problems caused by the single-line impedance mismatch can be eliminated at the terminal, but the single-line impedance is not controlled, which will interfere with other surrounding signals. Similarly, if the EMI interference of the surrounding signals is small, there is no need to consider parallel routing at all, but only care about the same length of the two signal lines, and the flying line does not matter. The tightly coupled wiring is just to make the common mode interference of the surrounding signals to the two signals basically the same, and to achieve a higher common mode rejection performance.

The above is the introduction of differential line transmission. Ipcb is also provided to PCB manufacturers and PCB manufacturing technology.