[Design View / Design Solution]
Characterizing High-Speed Serial Communications Links Requires Some Analog Savvy
A six-step process helps measure, identify, and eliminate clock and data jitter on those blazing serial signals.
To minimize the bit error rate (BER) you must properly time this phase shift with the data stream, and for that reason serial-communication standards now place a greater importance on high-accuracy measurement of jitter. Jitter is generally classified as deterministic jitter (Dj) or random jitter (Rj). Because each is created differently, they are characterized separately.
Two Fundamental Components Of Jitter Random jitter (Rj) represents timing noise with no discernable pattern. For the purpose of modeling, it is assumed to have a Gaussian probability distribution (Figure 4). Usually due to forces of nature, random jitter is statistical and unbounded. (It is characterized by its standard deviation value, expressed as an rms quantity.) Thus, providing an Rj spec without a sample size does not make much sense. Other than measuring the value of Rj in a system, however, most designers do little else with this parameter. Finding the cause of Rj is a difficult task, and beyond the scope of this article.
Deterministic jitter (Dj) is caused by events in the system, and appears as timing noise with somewhat discernable patterns. It is usually repeatable, persistent, and predictable. In addition, it is typically the result of faulty design in areas such as the circuit, the layout, and the transmission line. It is typically non-Gaussian, as is power-supply noise due to a bad reference plane.
Deterministic jitter is further classified into sub-components: periodic jitter (Pj in Figure 5), data-dependent jitter (DDj, also known as inter-symbol interference, or ISI), duty-cycle-distortion jitter (DCDj), and any other timing jitter that is uncorrelated and bounded to the data. Pj can be caused by cross-talk (from other signals and from semiconductor switching close to the serial-data signals), electromagnetic interference (EMI), and other unwanted modulation. DCDj results from unbalanced transitions in the data (differences in rise and fall times), and DDj is jitter correlated with bit sequences in the data stream (also affected by the channel’s frequency response)[1].
Total Jitter (Tj) As you might guess, total jitter is composed of random and deterministic components (Figure 6). There are several techniques for estimating Tj. Some find the total jitter by resolving it into Rj and Dj components, then adding them together using a multiplier in front of the Rj component. Other methods find total jitter by extrapolating the histogram of time interval error (TIE) measurements. Tj is usually a peak-to-peak value expressed in pico-seconds or fractions of a unit interval (UI). For example, 0.2UI means that jitter is 20% of the data eye.
To predict the overall performance of a system, you must understand the types of jitter and their effects. Because jitter causes timing errors, it has become increasingly important to characterize and qualify all jitter components in a system. Before that can be done, however, you must determine the sources of jitter. As mentioned earlier, the two types (random and deterministic) have different sources. A designer has little or no control over the sources of Rj in an existing system of embedded circuit boards[1], but he can use good design practices to greatly mitigate or even eliminate the sources of deterministic jitter. Each jitter component has a specific cause, as shown in (Table 1)[1].
Proposed Link-Characterization Framework The proposed link-characterization framework we will discuss helps to identify and measure the sources of clock and data jitter. The technique hinges on the designer’s ability to separate jitter sources, and to focus on the problem areas revealed by this testing framework. Jitter testing generally requires the observation of a repeating test pattern on the channel.
The data pattern to be used is important, because reflection and intersymbol interference (ISI) are both data-dependent sources of noise. The test patterns used to collect the majority of plots in this paper included a mixed-frequency repeating K28.5 sequence (also known as the comma character: K28.5 = 00111110101100000101), and a pseudo-random bit sequence (PRBS-23). PRBS patterns give a good spread of the different bit sequences that might be observed in actual data traffic. Other compliance test patterns for jitter evaluation are available, including the Jitter Test Pattern (JTPAT), Compliance Random Pattern (CRPAT), and Compliance JTPAT (CJTPAT), to name a few.
Please refresh the page if you have trouble reading this text.
Search Electronic Design
Email Newsletter
Sponsored By:
Electronic Design UPDATE provides readers with late-breaking news, opinions from industry experts, and timely technology stories. It's a unique opportunity to get your product message in front of engineers, engineering managers, and corporate managers while they're reading about critical information online.
Reprints
