PCB News - PCB time-domain crosstalk measurement method

Time-domain crosstalk measurement method for PCB quality verification
Analysis of time-domain crosstalk measurement method for PCB quality verification
This article discusses the composition of crosstalk and shows readers how to use Tektronix’s TDS8000B series sampling oscilloscope or CSA8000B series communication signal analyzer to measure crosstalk on a single-sided PCB.
With the increasing speed of digital systems in the fields of communication, video, networking and computer technology, the quality requirements for printed circuit boards (PCBs) in such systems are also getting higher and higher. The early PCB design has been unable to guarantee the system performance and working requirements in the face of increasing signal frequency and decreasing pulse rise time. In the current PCB design, we need to use transmission line theory to model the PCB and its components (edge connectors, microstrip lines, and component sockets). Only by fully understanding the forms, mechanisms and consequences of crosstalk on PCBs and using corresponding technologies to suppress them to the greatest extent can we help us improve the reliability of systems including PCBs. This article mainly focuses on PCB design, but I believe that the content discussed in the article will also help other applications such as the characterization of cables and connectors.

Possible consequences of crosstalk

The reason why PCB designers care about crosstalk is that crosstalk may cause the following performance problems:

>The noise level rises,

>Harmful spike burrs,

>Data edge jitter,

>Unexpected signal reflection.

Which of these issues will affect the PCB design depends on many factors, such as the characteristics of the logic circuit used on the board, the design of the circuit board, the mode of crosstalk (reverse or forward), and the interference line and the interfered line The termination conditions on both sides. The information provided below can help readers deepen their understanding and research on crosstalk, thereby reducing the impact of crosstalk on the design.

The method of studying crosstalk

In order to minimize the crosstalk in PCB design, we must find a balance between capacitive reactance and inductive reactance, and strive to achieve the rated impedance value, because the manufacturability of PCB requires that the transmission line impedance be well controlled. After the circuit board is designed, the components, connectors, and termination methods on the board determine which type of crosstalk will have much impact on circuit performance. Using time-domain measurement methods, by calculating the inflection point frequency and understanding the PCB crosstalk (Crosstalk-on-PCB) model, it can help designers to set the boundary range of crosstalk analysis.

Time domain measurement method

In order to measure and analyze crosstalk, frequency domain technology can be used to observe the relationship between the harmonic components of the clock in the frequency spectrum and the maximum EMI at these harmonic frequencies. However, time-domain measurement of the digital signal edge (the time it takes to rise from 10% of the signal level to 90%) is also a means of measuring and analyzing crosstalk, and time-domain measurement has the following advantages: Changes in digital signal edges Speed, or rise time, directly reflects how high each frequency component in the signal is. Therefore, the signal speed (ie rise time) defined by the signal edge can also help reveal the mechanism of crosstalk. The rise time can be directly used to calculate the inflection point frequency. This article will use the rise time measurement method to explain and measure crosstalk.

knee frequency

In order to ensure that a digital system can work reliably, designers must study and verify the performance of the circuit design below the inflection point frequency. The frequency domain analysis of digital signals shows that signals higher than the inflection point frequency will be attenuated, and therefore will not have a substantial impact on crosstalk, while the energy contained in the signals below the inflection point frequency is sufficient to affect the operation of the circuit. The inflection point frequency is calculated by the following formula:

Fknee = 0.5/ trise

PCB crosstalk model

The model given in this section provides a platform for the study of different forms of crosstalk and clarifies how the mutual impedance between two microstrip lines causes crosstalk on the PCB board.

The mutual impedance is uniformly distributed along the two traces. Crosstalk is generated when the digital gate circuit sends a rising edge to the crosstalk line, and it spreads along the trace:

1. Both the mutual capacitance Cm and the mutual inductance Lm couple or "crosstalk" a voltage to the adjacent interfered line.

2. The crosstalk voltage appears on the interfered line in the form of a narrow pulse with a width equal to the rise time of the pulse on the interference line.

3. On the interfered line, the crosstalk pulse splits into two, and then starts to propagate in two opposite directions. This divides the crosstalk into two parts: forward crosstalk that propagates along the original interference pulse propagation direction and reverse crosstalk that propagates along the opposite direction to the signal source.

Crosstalk type and coupling mechanism

According to the model discussed above, the coupling mechanism of crosstalk will be introduced below, and the two types of crosstalk, forward and reverse, will be discussed.

capacitive coupling mechanism

Interference mechanism caused by mutual capacitance in the circuit:

>When the pulse on the interference line reaches the capacitor, a narrow pulse will be coupled to the interfered line through the capacitor.

>The amplitude of the coupled pulse is determined by the size of the mutual capacitance.

>Then, the coupled pulse is divided into two and starts to propagate in two opposite directions along the interfered line.

Inductance or transformer coupling mechanism

The mutual inductance in the circuit will cause the following interference:

>The pulse propagating on the interference line will charge the next position where the current spike appears.

>This kind of current spike generates a magnetic field, and then induces a current spike on the interfered line.

>The transformer will produce two voltage spikes of opposite polarity on the interfered line: the negative spike propagates in the forward direction, and the positive spike propagates in the reverse direction.

reverse crosstalk

The capacitive and inductive coupling crosstalk voltage caused by the above model will produce an additive effect at the crosstalk position of the interfered line. The resulting reverse crosstalk includes the following characteristics:

> Reverse crosstalk is the sum of two pulses of the same polarity.

>Because the crosstalk position propagates along the edge of the interference pulse, the reverse interference appears as a low-level, wide pulse signal at the source end of the interfered line, and there is a corresponding relationship between its width and the trace length.

>The amplitude of the reflected crosstalk is independent of the pulse rise time of the interference line, but depends on the mutual impedance value.

forward crosstalk

It needs to be reiterated that the capacitive and inductive coupling crosstalk voltage will accumulate at the crosstalk position of the interfered line. Forward crosstalk includes the following characteristics:

> Forward crosstalk is the sum of two reverse polarity pulses. Because the polarity is opposite, the result depends on the relative value of the capacitance and inductance.

>Forward crosstalk appears at the end of the interfered line as a narrow spike with a width equal to the rise time of the interference pulse.

>Forward crosstalk depends on the rise time of the interference pulse. The faster the rising edge, the higher the amplitude and the narrower the width.

>The amplitude of the forward crosstalk also depends on the length of the pair: as the position of the crosstalk propagates along the edge of the interference pulse, the forward crosstalk pulse on the interfered line will gain more energy.

Characterization of crosstalk

This section will use several single-layer PCB measurement examples to study the generation mechanism of crosstalk and the several types of crosstalk introduced above.

Note: To familiarize yourself with the crosstalk problems and consequences on multilayer PCBs and their ground planes, please read the references or other resources at the end of this article.

Instruments and settings

In order to effectively measure crosstalk in the laboratory, a wideband oscilloscope with a measurement bandwidth of 20 GHz should be used, and a high-quality pulse generator should output a pulse with a rise time equal to the rise time of the oscilloscope to drive the circuit under test. At the same time, high-quality cables, termination resistors and adapters are used to connect the PCB under test.

The 80E04 electronic sampling module is installed in the Tektronix 8000B series of instruments, which is an ideal instrument combination for successfully measuring crosstalk. 80E04 is a dual-channel sampling module, including a TDR step voltage generator, which can generate a 250mv narrow pulse with a rise time of 17ps and output with a source impedance of 50 ohms. The tester only needs to connect the PCB to be tested.

Forward crosstalk measurement

If you are only measuring forward crosstalk, you need to terminate all traces to eliminate reflections. Forward crosstalk should be measured at the end of a well-terminated interfered wire.

If the mutual inductance is greater than the crosstalk of mutual capacitance coupling, then the crosstalk pulse should be negative at the rising edge of the interference pulse, and the width is equal to the rise time of the interference pulse. The instrument in the figure shows a negative pulse (C4) with an amplitude of 48.45 mV. The interference pulse amplitude is 250 mV, and the crosstalk amplitude is nearly 50 mV, so the fast edge of the interference pulse produces 20% of the crosstalk on the interfered line.

Because the input step voltage from 80E04 has a very fast edge during measurement, the crosstalk obtained is too large and cannot represent the driving signal in the actual logic circuit. For example, if the driving signal comes from a 1.5 ns CMOS gate, the generated crosstalk pulse is wider and has a smaller amplitude. To make the measurement reflect this situation, you can use the Define Math function of the instrument to add a low-pass filter after the signal is captured. The M1 waveform (white) in Figure 7 gives the measured results after filtering. It should be noted that M1 is 10 times more sensitive in the vertical direction than the unfiltered waveform.

Although mathematical analysis has proved that the effect of low-pass filtering after signal capture is the same as that of physical filtering of interference pulses connected to the line, the following steps are more convincing:

>Measure the crosstalk caused by two rising edges, one fast and one slow, and the same amplitude interference pulse,

>Then change the crosstalk caused by the interference pulse with the fast rising edge to the crosstalk caused by the interference pulse with the slow rising edge through the low-pass filter, and finally check the result.

>$waveform (R2) is the slow-edge interference pulse, and the red waveform (R3) is the crosstalk caused by it.

> The green waveform is the fast edge TDR pulse (R1), and the white waveform (R4) is the crosstalk caused by it.

> The blue waveform is the waveform obtained by slowing the rising edge of the pulse after filtering the white waveform, and it represents the result of post-filtering the crosstalk. The red and blue crosstalk waveforms shown in the figure are displayed on the same voltage scale.

When single measurement of reverse crosstalk, it is necessary to terminate the interference line and the interfered line with a 50 ohm resistor to eliminate the reflection. The measurement should be made at the left end of the interfered line. The amplitude of the reflected pulse is very low, and the width is twice the line length, because the crosstalk at the end of the trace must be transmitted back to the source end of the trace. In the measurement of reverse crosstalk, the crosstalk generated by the fast-edge interference pulse is about ?? mV, which is equivalent to 4% of the amplitude of the interference pulse. The magnitude of the reverse crosstalk has nothing to do with the rise time of the interference pulse. The following two waveforms are the crosstalk generated by the slow-edge pulse and the crosstalk generated by the fast-edge pulse after filtering. Their amplitudes are both 6.5 mV. The difference between the length of the trace and the rise time of the interference pulse makes the reverse crosstalk amplitude generated by the slow-edge pulse smaller.

Because the rise time of the interference pulse is longer than the line length of the trace at this time, the pulse edge has not reached the peak of the amplitude when it is transmitted back to the source end of the trace along the trace direction. Figure 11 shows the crosstalk measurement results obtained when a 200 ps rise time generator (DG2040) and the output of the 80E04 sampling module’s 17 ps generator are used as interference pulses. The three crosstalk waveforms shown in the figure all use a voltage scale of 5 mV/div.

Among them, the white waveform is the result of the crosstalk generated by the interference pulse with a rise time of 17 ps after being filtered (post filtering) to a rise time of 200 ps. These measurements have confirmed that unless the rise time of the interference pulse exceeds the length of the trace, the rise time does not affect the reverse crosstalk. And if the rise time of the interference pulse exceeds the length of the trace, the reverse crosstalk amplitude generated is smaller, because in this case the pulse edge cannot reach the peak amplitude even after the pulse edge traverses the entire trace.