5 Starting Point: 5-Meter System with 2 Nodes
Understanding the effect of the inherent noise of a complex mixing segments begins with understanding a simple, 2 node example. The base line example for this application note will use 5 meters of cable betweeen the nodes. The nodes are built using the min BIN configuration from the BIN Reference Design Application Note4. The cables join directly at the MDI of the node, so that there are no stubs. All of these systems work under laboratory conditions (ideal power supplies, room temperature, no alien noise).
First, examine the eye diagrams of this simple system when using different cables. The eye patterns are different between each transmitter and each receiver. To study inherent noise, it is necessary to look at all of the eye patterns. The left column shows the eye when node 0 is transmitting, while the right column shows the eye when node 1 is transmitting. The first row of results is measured with the receiver at node0 and the 2nd row is measured with the receiver at node 1. The plots shown in red are made at the transmitter, and the plots in blue are at nodes that are receiving from the transmitter.
The results for cable 1 are shown in Figure 5-1, while those for cable 4 are shown in Figure 5-2.
Cable 1 is a good quality cable. The signal at the receivers (blue plots) show a small amount of attenuation over the 5 meters, and no significant increase in signal jitter. Cables 2 and 3 produce similar results.
Cable 4 is a poor quality cable. Even at the transmitters (the red plots), the eye patterns are showing signal distortion, with visible jitter, shown as multiple zero crossing and reduced noise margin as the maximum signal amplitude has significant variation. At the receivers, significant cable attenuation causes further signal distortion. Jitter is large, and the smaller eye opening reduces the noise margin. Under additional stress, such as more nodes or longer cable length, a system based on this cable will not have enough noise margin to properly function.
As the number of nodes in a system increases, the number of eye diagrams increases by the number of nodes squared. To make the analysis easier, the eye diagrams can be combined as shown in Figure 5-3. This plot is created by dividing each diagram into 4 quadrants, rotating the quadrants and superimposing them on each other, and repeating the process for each of the N x N eye diagrams. The minimum height of the opening for any jitter value is used to create the red line, which is a plot of jitter tolerance versus maximum alien noise level. The 30mV offset of the eye patterns represents the 30mV differential input required by the receiver. The amplitude of the red curve above 0 is then the alien noise margin that remains after the worst case inherent noise. In the Figure, plot A shows the results for cable 1 and plot B shows the results for cable 4. In an automotive environment, a rule of thumb is to allow for at least 200mVpp of alien noise, so the poor quality cable 4 is already at its limit in this simple configuration for a jitter tolerance of less than 10ns.
