Which jitter measurement is correct? Part 2

Last week I shared a recent discussion around jitter measurement correlation. As a follow-on to that discussion, my colleague Mark Guenther shared some additional insights. Let’s return to the original question of why don’t the Rj/DJ measurements match the TJ(BER=1×10-12) expected result, given the ~14 multiplier for RJ. Jitter estimation based on RJ/DJ separation depends in part on the specific jitter components modeled.

For the purposes of analyzing jitter and identifying root cause, it is very useful to identify components as specifically as possible. But for the purposes of determining compliance (PCI Express, USB, etc), it has been found that a simplified jitter model yields results that are more consistent across different measurement instruments and different vendors; hence, the long history of using the Dual-Dirac-based model (Rj-dd and DJ-dd).

Why are the values of Rj and Rj-dd sometimes the same?

There are two slightly different Dual-Dirac models. The first one was documented in the MJSQ document published by the Fibre Channel working group in 2004. It was commonly referred to as “effective jitter,” or “equivalent jitter,” and was described in Chapter 8 of the MJSQ document. In this method, the red curve was fitted to the blue (measured) curve by picking two BER levels, which intersect the blue curve to define four reference points. Sigma and DJ-dd values are then chosen to minimize deviation of the red curve from these reference points. The process is described graphically thus:

As the diagram suggests, the 1e-6 and 1e-12 BER levels are used for the fit. When this model is used, the measured RJ and the Rj-dd can be different.

The other model, which is in far more common use today, is the “Q-scale fit” approach suggested by the image below. In that approach, the bathtub is plotted using the Q-scale for the vertical axis, and the Rj-dd is obtained by fitting sloping lines to the asymptotes (deep parts) of the blue curve. If the blue curve came from a BERT scan, then the blue curve might have bumps and wiggles even down in the low parts. But if the blue curve was derived from spectral analysis (as we do on real-time scopes), the lower parts of the blue curve have actually come from extrapolating based on the measured jitter components. In this case, it is natural for the measured Rj and the fitted Rj-dd to be identical.

In my world, jitter is one topic we discuss a lot. But it’s not the only thing. What else is on your mind? Don’t hesitate to add your comments below, or drop me a note at scope-guru@tektronix.com.

Randy White is the Serial Applications Technical Marketing Manager at Tektronix. Randy has worked with various aspects of test and measurement solutions at Tektronix over the past few years. He has given seminars on high-speed serial measurements and is actively involved in many working groups for high-speed serial standards. He holds a BSEE from Oregon State University in Corvallis.