Embedded Files
Boulder, CO (September 2024) - This experiment explored the impact of interconnected geometries on the amount of cross talk between aggressor and victim signal-return path pairs. The key takeaway is demonstrating why it is important to use a plane for the return path to reduce cross talk. The first experiment examined the impact from best-case cross talk with a continuous return path. Next the negative impact of using a shared return path was investigated
Figure 1: The printed circuit board used to test trace-to-trace cross talk. The bottom third of the board victim line uses a separate trace connecting ground to ground over a ground plane. The middle third does not have a ground plane and has a switch which can be used to separate or share the returns for the victim and aggressor lines.
To look at cross talk noise specifically, as opposed to switching noise, the victim line does not have a signal transmitted over it. This means the signal on the victim line is exclusively from cross talk noise. Figure 1 shows the hardware used to measure this effect. Figure 2 shows the noise on the victim line with a ground plane present from six LEDs being driven simultaneously from an Arduino Uno development board running at 16MHz. The signals are driven high for one clock cycle, then low for one clock cycle. The comparison between driving all of the LEDs (the green line) and just driving the LED on pin D8 (the orange line) shows that the proximity of the aggressor signal to the victim line contributes significantly to the amount of noise; in other words, the two signals are similar even though only one LED is being driven for the orange signal.
Next, the impact of sharing return paths was investigated using the portion of the hardware without a ground plane. Figure 3 shows approximately 3V of noise when all signals share a return path (depicted in orange). This is significantly higher than the 724mV of noise measured when the signals have separate returns.
Figure 2: Compares the cross talk on the victim line when only Pin 8 is being switched every clock cycle (depicted in orange) with when all the aggressor lines are being switched every clock cycle (depicted in green). The yellow line depicts the signal from Pin 13. The orange and green victim lines scale is 200mV per division and the yellow aggressor line scale is 1.0V per division.
Figure 3: Compares the cross talk on the victim line when the signal-return is shared between all lines (depicted in orange) and when the victim line has a separate return (depicted in green). The yellow line depicts the signal from Pin 13. Both the orange and green victim lines and the yellow aggressor line scale is 1.0V per division.
A significant amount of noise, approximately 300mV, was measured when using long return lines near other signals as shown by the green line in Figure 1. Removing the ground plane increased the noise to more than 700mV, as shown by the green line in Figure 2. This shows the importance of using a plane for the return path as well as keeping signal lines separated. The most cross talk noise was found when lines shared a return without a ground plane which caused approximately 3V of noise as shown by the orange line in Figure 2. This not only reinforces using a plane for the return path but shows the significant problems that can be caused by sharing return paths making it crucial to avoid in practice. To summarize, this experiment shows why it is important to use a plane for the return path, to keep critical signals separated, and to avoid sharing return paths across traces.
Page updated
Google Sites
Report abuse