Why does an eye diagram not correlate to a bathtub curve?
The eye diagram is probably the most well-known signal integrity tool because it combines numerous signal integrity characteristics such as rise/fall, overshoot/undershoot, and voltage/jitter into a simple visual of the signal. As additional layers of signal analysis and abstraction are added, there is often the need to perform a sanity check to ensure correlation to the physical layer waveforms. A case in point is understanding how a bathtub curve relates back to the eye diagram.
In the figure below, the familiar eye diagram is on the left; the bathtub is on the right for the same signal.
The first thing to note is the vertical scale of the bathtub. It has terms like 1E-8, 1E-10, and 1E-12. However, the scale is really a ratio/rate. So a ratio at 1E-12 is really the same as a population size of 1E+12. The population size or number of UIs increases as you go down the bathtub curve.
You will notice that the eye opening is very large compared with the bathtub at 1E-12. The reason is that the eye diagram closure/jitter is a directly measured Pk-Pk value where the UIs captured is approximately 1E+5. If you go to the bathtub curve, you will find a direct correlation with the eye opening at 1E-5.
If you wait long enough with the eye diagram and capture 1E+12 UIs, the eye diagram opening will match the bathtub opening (pink horizontal line) and the eye diagram Pk-Pk jitter will match the bathtub total jitter at 1E-12.
In an upcoming post, I will discuss the correlation between total jitter and bit error rate.