[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.
The key to getting accurate measurements lies in selecting the right measurement equipment for your application (oscilloscopes and probes, for instance). For step 1 of this framework (and for the remaining steps as well), the signal is measured after it has propagated through a channel of 50-Ω transmission line that also includes the cable, connector, and FR-4 PCB. You should solder a differential, high-performance probe with high bandwidth and low capacitive loading to the PCB trace as close as possible to the receiver IC.
STEP I: Quantify Random And Deterministic Jitter (Rj And Dj) The first step is to observe the signaling level. Then, you collect link measurements and compare them to the standard. (Table 2 gives an example of measurements versus the XAUI specification, which is measurement of the PHY’s input characteristics.) The SI engineer can create a similar matrix for the standard against which his system is being tested.
An eye diagram is one of the most important measurement techniques used in assessing high-speed signal integrity. It overlays waveforms from multiple unit intervals (UI) using either the real clock or a reconstructed clock as the timing reference. Because the eye diagram helps you to visualize amplitude behavior as well as the timing behavior of a waveform, it represents one of the most useful presentations of jitter. Figure 7) shows an eye diagram measurement taken from a XAUI channel.
Using timing-analysis software loaded on the scope (TDSJIT3 from Tektronix, for instance) and with the scope set for “golden PLL,” the SI engineer can set the parameters shown in Table 2 and capture an eye diagram of the channel traffic. Then, he can complete the matrix shown in Table 2 for the particular standard being used. (Golden PLL is a method for filtering out jitter on the scope trigger, thereby ensuring that any jitter represented in the measured jitter amplitude and histograms is actually present on the linkC.)
STEP II: Amplitude Noise Or Voltage Error Histograms This step measures amplitude noise, which can cause error in the design. We are looking to see if the probability density functions (PDFs) for amplitude have a normal distribution for both the “1” and “0” levels. (Figure 8 shows the PDFs for a XAUI link.) The random-amplitude noise shown in blue in the histograms (circled in red) can be considered as normal distributions. The SI engineer can also use this plot as a graphic aid in determining whether other signaling issues are present, such as overshoot and undershoot. If amplitude noise is an issue (if the amplitude histograms are bi-modal, for instance), then we likely have a power-distribution problem on the board.
STEP III: Eye Diagram Versus “Far-End” Mask Analysis In Step III, the SI engineer estimates jitter quality for the received signal over a long sequence of data. Many jitter application packages include standard masks, whose minimum-closure dimension allows you to rate the quality of the measured channel. By comparing the eye diagram to the receive masks, you can view the amount of eye closure in a given configuration. The eye should be clear of the masks (Figure 9).
At this stage, the tester also analyzes the eye plot’s rising edges separately from the falling edges. In the example of Figure 10, you can clearly observe that the rising and falling edges are not aligned in the middle at the eye crossing point (the bi-modal histogram circled at mid-top of the figure). This bi-modal histogram is an indication of the presence of cycle-to-cycle or periodic jitter on the channel. The histogram could also represent DCD or ISI jitters.
Designers often limit their testing to a measurment of total jitter and only view the histogram, which represents the total jitter (Dj and Rj mixed together). To understand the root cause of jitter and eliminate its contributing components, it’s essential to separate and identify each component. Since the eye diagram is a general tool that gives insight only into the amplitude and timing behavior of the signals, other means are needed to separate the jitter components. In the next step, we separate total jitter into its components by analyzing the jitter histogram and bathtub plots.
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
