Timing Jitter Education

Why is jitter important?

Timing jitter is a significant issue in modern high-speed digital design. Communications bus standards like USB, IEEE1394, SPI4.2, Fibre Channel, and PCI-Express contain jitter specifications that must be met for compliance certification. Engineers working with these systems need to understand what jitter is, how to characterize it, and how to find and mitigate its root causes so that their systems work properly.

What is jitter and what causes it?

Modern digital systems are comprised of large quantities of information. The smallest unit of information in such a system is the “bit.” Every bit can have a value of 0 or 1, and each value is defined by a distinct voltage level. As information is transferred in a digital system, the voltage at a given point can transition very rapidly between these two voltage levels (corresponding to 0 and 1 values). In order for the system to work correctly, these transitions must happen at very precise times. Jitter is the phenomenon that causes such transitions to occur either before or after their expected times.

There are two broad categories of timing jitter: random and deterministic. Random jitter (RJ), as its name implies, is caused by events that occur randomly, and causes varying amounts of jitter on the waveform transitions. Deterministic jitter (DJ) is caused by specific issues with a given system, and the designer often has much control over those issues. There are three basic types of DJ, each with its own underlying cause. Periodic jitter originates from electromagnetic interference. Intersymbol interference results from a bandwidth limitation in the transmitting medium. Duty cycle distortion is normally caused by an incorrect threshold voltage setting, although it is possible for it to be caused by asymmetric edge rates.

How can we measure and display jitter?

Figure 1: Jitter histogram

Perhaps the three most basic types of jitter measurements are period jitter, cycle-to-cycle jitter, and time interval error. Note the distinction between the two very similar terms, periodic jitter and period jitter. Periodic jitter is one of the types of jitter mentioned above. Period jitter is a jitter measurement; it is the difference between the minimum and maximum periods of a waveform. Cycle-to-cycle jitter can be defined as the period jitter measured between two adjacent clock cycles of a waveform. There is also a variation of this measurement called n-cycle jitter. N-cycle jitter is the period jitter measured between two cycles of a waveform separated by “n” clock cycles.

Figure 2: Jitter trend display

Time interval error is a more complex measurement. First, the entire captured waveform must be analyzed to reconstruct its “ideal clock.” Then, the actual waveform transitions are compared to the “ideal” transitions to determine the jitter, called time interval error.

There are several ways to display jitter, including the

histogram, trend display, spectrum, eye diagram, and bathtub plot. The histogram (see Figure 1 for an example) plots the number of occurrences of jitter as a function of the jitter magnitude. Trend displays (see Figure 2) plot jitter magnitude against “wall clock time.”

Figure 3: Jitter eye diagram

Waveform trends, such as spread spectrum clocking, show up very clearly on this display. The trend display is very helpful for correlating periodic jitter with the signal causing it. The spectrum display shows jitter magnitude as a function of frequency, and is also helpful for determining sources of electromagnetic interference. The eye diagram (Figure 3) overlays repetitive waveform captures and simultaneously shows voltage and time information in a very intuitive fashion. A less common and arguably less intuitive display is the bathtub plot (Figure 4). This plot is normally generated by an instrument called a bit error ratio tester (BERT). It maps the size of the data valid window to the ratio of bits transmitted.

Figure 4: Bathtub plot

Modern test instruments that can measure timing jitter include real-time digital storage oscilloscopes (DSOs), sampling oscilloscopes, logic analyzers, spectrum analyzers, bit error ratio testers (BERTs), and time interval analyzers (TIAs). The real-time DSO is by far the most common of these instruments.

How do we teach jitter?

In-depth jitter analysis is a fairly complex subject, but the basic concepts are straightforward. Students having a basic background in analog and digital circuits are prepared to learn about timing jitter. Professor Harding teaches the subject in his second-semester digital electronics course. He has authored a few papers on the subject in recent years, including:

A PDF of the lab instructions for the first-year lab can be downloaded here. If you need the lab instructions in MS Word format, have questions, or want to offer feedback/suggestions about any of this material, please email Professor Harding at glhardin@purdue.edu.

Need more info about timing jitter?

There are numerous white papers and application notes about timing jitter at www.agilent.com and www.tek.com.