Managing Reflections With Terminations

Have you heard recommendations to use a particular termination in particular situations for good signal integrity? Have you ever wondered how to incorporate terminations in your design? While there are typical use cases for various terminations, sometimes engineers use termination techniques based on a recommendation or assumption that may not work, or at least may not be optimal, for their particular use case.

An understanding of why terminations are used and how they work, coupled with a signal integrity simulator, gives an engineer a powerful combination of tools to tackle new signal termination challenges. In this article, we attempt to establish this understanding by briefly exploring the cause of signal reflections and some basic termination techniques. Equipped with this fundamental background, an engineer can use a signal integrity simulator to evaluate the appropriateness of a particular termination for their use case.

Figure 1 shows a digital waveform (green), which is the ideal signal we intend to transmit and receive. Overlaid with it is an analog waveform (red), which is an example of what a real signal might look like at a receiver. It has some overshoot and ringing caused by impedance discontinuities that can be managed with termination. The shape and extent of these kinds of signal integrity issues can vary greatly. If severe enough, they can cause bit errors or even component damage.

We will explore how these signal integrity issues arise and how termination can help manage them. It is important to remember that what works to manage the overshoot and ringing for one case might not work for another case, and we will see an example of that later. In general, signal integrity simulations are performed to evaluate particular situations.

Fig. 1: Comparing digital and analog waveforms.

Transmission lines and reflections

In signal integrity design and analysis, one of the fundamental elements we’re concerned with are the transmission lines carrying the signals. A transmission line is essentially a set of conductors separated by a dielectric and used to carry or guide electromagnetic energy from one place to another. In a transmission line, we consider at least one conductor as the return or reference, and another as the signal. Real transmission lines are not ideal, and they exhibit delay, loss, and coupling, which impacts the signal on a given connection and potentially other signals. Figure 2 shows two examples of printed circuit board transmission lines.

Fig. 2: Transmission line examples.

One fundamental, important characteristic of a transmission line is characteristic impedance. Figure 3 shows a representation of a transmission line with an impedance discontinuity, where two sections having different characteristic impedances (denoted as Z₀₁ and Z₀₂ in figure 3) are touching each other. When a signal propagating on a transmission line encounters an impedance discontinuity, some of the signal will be reflected back toward the source and some will be transmitted through. We can calculate the reflected voltage using the reflection coefficient equation shown in figure 3. The reflected voltage then is the incident voltage multiplied by the reflection coefficient. The total voltage at the boundary, and in section 1, is the sum of the incident and reflected voltages. The transmitted voltage moving into section 2 is the sum of the incident and reflected voltages. The reflected signal experiences the same transmission line in section 1 as the incident signal does, including delay. Therefore, it will take time for the reflected signal to propagate from the discontinuity back down the transmission line toward the source. Impedance discontinuities can be introduced if a transmission line’s cross section changes due to geometry or material changes. Changing a trace width would be an example of this. Other sources of impedance discontinuity could be vias, connectors, component pins and packages, transmission line branches, return path interruptions, and component input and output impedances.

Fig. 3: Reflection coefficient.

Consider the example shown in figure 4. We have a 3.3V CMOS driver and receiver at the ends of the line, and a 50 Ω transmission line with a 10 ns delay from driver to receiver. The CMOS driver has a low output impedance (approximately 4 Ω). The CMOS receiver has a very high input impedance (significantly higher than 50 Ω). In this case, we only have one transmission line with a consistent characteristic impedance, so we don’t expect any reflections from the interconnect between the driver and receiver. However, we still get reflections at the ends of the transmission line when the signal encounters the impedance of the buffers. In the figure, the red waveform is at the driver, and the green waveform is at the receiver. When the driver launches the rising edge into the transmission line, it rises to approximately 3 V. It doesn’t rise all the way to 3.3 V because of the small output impedance associated with the driver. This signal propagates toward the receiver on the 50 Ω transmission line and eventually encounters a very high impedance when it arrives at the receiver 10 ns later. The reflection coefficient there is approximately 1, producing a reflected voltage of approximately 3 V, and the voltage at the receiver rises to 6 V (3 V + 3 V). The reflected 3 V signal then propagates 10 ns back toward the driver on the 50 Ω transmission line when it eventually encounters a 4 Ω impedance at the driver. The reflection coefficient there is -0.85 so we get a reflected voltage of -85% of the 3 V signal that was reflected off the receiver. The voltage at the transmitter approximately 20 ns after the initial signal was launched is the sum of the original launched voltage (3 V) plus the reflected voltage from the receiver (3 V) plus the new reflected voltage at the transmitter (-2.5). This is approximately 3.5 V (3 V + 3 V – 2.5 V). The -2.5 V reflection from the transmitter then propagates toward the receiver where there will be another positive reflection, and so on. This is how the overshoot and ringing in figure 1 is caused as the reflected signal bounces back and forth along the transmission line.

Fig. 4: Reflections example.

As in the previous example, it’s not always possible to avoid impedance discontinuities, but we can often reduce their effects to an acceptable level by using terminations to manage them.

Using terminations to manage reflections

Consider the example shown in figure 5, which is the previous example with the addition of a resistor R_s. If we keep resistor R_s very close to the transmitter, we can effectively combine it with the output impedance of the buffer to get a new effective transmit impedance, Z_TX, which is the sum of R_s and the buffer’s output impedance (Z_source). This technique is called series termination. The idea is to take advantage of the first reflection from the receiver but then prevent any further reflections by not allowing any reflections at the transmitter. If we don’t want any reflections from the transmitter then the reflection coefficient there should be zero. To achieve this, we set R_s to the difference of the transmission line characteristic impedance and the transmit buffer output impedance, or R_s = Z₀ – Z_source.

To see how this works, first notice that the effective transmit impedance (Z_TX) and the characteristic impedance of the transmission line (Z₀) create a voltage divider. When they are equal, the voltage launched into the transmission line is half of the source voltage (seen in the red waveform in the figure). This voltage then propagates on the transmission line with characteristic impedance Z₀ to the receiver where it encounters a very large impedance. The reflection coefficient there is approximately 1. With approximately 100% reflection, the voltage at the receiver is approximately two times the arriving voltage. Because the incident voltage was half of the source voltage, when adding the incident and reflected voltages, we get the full source voltage at the receiver, which is our goal. This is seen in the green waveform in the figure. Then notice that the reflected voltage propagates back toward the transmitter along the transmission line with characteristic impedance Z₀. When arriving at the transmitter, the signal encounters the effective output impedance Z_TX, which equals Z₀, so there is no reflection at the transmitter, which also is our goal.

In summary, we have managed reflections with a series resistor (series termination) by allowing the reflection at the receiver and terminating the reflected signal at the transmitter to eliminate further reflections. We must keep R_s close to the transmitter. If we don’t, then the reflected wave doesn’t encounter a match to the transmission line characteristic impedance. Instead, it encounters R_s and then some other impedance of the connection between R_s and the transmitter before encountering Z_source.

Fig. 5: Managing reflections – series terminations.

Figure 6 shows some other termination possibilities, all using the same transmitter, receiver, and transmission line. A series termination example is also included for comparison (green waveform). All the other examples make use of a termination at the receiving end rather than at the transmitting end of the transmission line. The general intent of these approaches is to match the transmission line impedance at the receiving end and prevent reflections there. Signals arriving at the receiver have been traveling down the characteristic impedance of the transmission line and then encounter an impedance of the same value at the receiver so there are no reflections. The impedance at the receiving device is the parallel combination of the termination impedance and the input impedance of the receiving device. Since the buffer input impedance in this case is very large, the effective impedance at the receiver is approximately just that of the termination. Consider though that the devices and their packages can have a significant impact on signal integrity. Signals arriving at the pins of a receiving component, for example, may still have to continue through additional interconnect of the IC package before arriving at the actual input buffer. Some devices also provide terminations on the die which puts the termination as close as possible to the receiver. For now, however, we’re focusing on concepts and assume the IC package has minimal effect.

Fig. 6: Additional (parallel) termination examples.

For all the terminations in figure 6, we notice that if the termination impedance is a good match to the transmission line characteristic impedance, the signal at the receiver has minimal overshoot and ringing. While the various terminations may be effective at managing reflections, they have various advantages and disadvantages to consider. Some considerations include the number of required components, power dissipation, voltage levels, and physical placement. For example, the termination with a resistor and capacitor has larger signal swing than the resistor-only terminations and it doesn’t dissipate DC power, but it requires two termination components instead of one.

Consider now a multi-load example, as shown in figure 7, with no termination. The waveform at the load farthest from the driver (green) is perhaps the best-looking waveform. Moving toward the transmitter, the waveforms at the remaining loads look worse the closer we get to the transmitter.

Fig. 7: Multi-load example.

In figure 8, we try a series termination for the previous example and see waveforms similar to those we saw with no termination. A series termination does not really help in this case. There are still reflections from the loads and, even though they are terminated at the transmitter, they affect the other loads.

Fig. 8: Multi-load example – series termination.

Figure 9 shows the multi-load example using a parallel termination. The incident signal propagates to each load and is terminated at the end after the last load, so there is no reflection there. The static voltages are a bit less, due to the parallel termination, but the waveshapes look good at all of the loads. Note, we are assuming each of the branches, or stubs, to each load is very short. If they were not, we could get reflections from each of those branches.

Fig. 9: Multi-load example – parallel termination.

Conclusion

The specific examples we’ve examined are used to demonstrate concepts of reflections and terminations. They don’t cover all possibilities but provide a basis of understanding to be applied in other situations. With a foundational understanding, a signal integrity simulator can be used to evaluate and select the best option for interconnect topology and termination for a given situation.

To gain a fuller understanding of transmission lines, reflections, and termination techniques, please read the new eBook from Siemens DISW/EDA, Signal integrity basics.

In the book, the author also covers the following contributors to obtaining the optimal signal integrity in your designs:

• Signal integrity basics
• Crosstalk
• Differential pairs
• Vias and impairments
• Timing