LeCroy’s statistical view of jitter gives insight by providing a view of the distribution of jitter. As with any noise-based phenomena, the peak-to-peak value grows as more values are measured. Therefore, anyone interested in determining worst-case timing and jitter values needs to consider the number of measurements taken in making this determination. More is better.

One type of jitter commonly measured is the variation in a clock period as diagrammed below. Consider each arrow as one measurement of period and the difference between the shortest and longest durations as the peak-to-peak jitter. In most designs, it is the worst case jitter occurrences that cause problems. So finding them is crucial.

Hundreds of thousands — even millions — of timing values are necessary for accurate peak-to-peak measurements. More measurements mean more accuracy!

As more individual measurements are accumulated, their distribution can be displayed in a histogram. The peak-to-peak value grows statistically as more measurements are made.

Some measurements, such as n-cycle jitter measurements, can only be performed on a single acquisition. Acquisition duration and the ability to analyze the entire acquisition are the keys to peak-to-peak measurement accuracy. LeCroy scopes can capture and analyze hundreds of thousands — even millions — of clock cycles in a single acquisition with zero dead time!

160 kpoints Captured/Analyzed

Shown above is a frequency jitter measurement for a high-speed clock using a capture duration of 160 kpoints of acquisition memory (typically available on non-LeCroy scopes). Due to the short capture duration, the histogram has only 1,567 measurement values in its distribution. Note that the frequency jitter value range (D) is 21.25 MHz. Is this the peak-to-peak jitter your system actually sees?

64 Mpoints Captured/Analyzed

Shown above is the same measurement as on the left, using the 64 Mpoint capture duration available in LeCroy scopes. Note that the histogram distribution is now very clear. Also note that the peak-to-peak measurement range (D) is considerably higher and more accurate now because 395 times more measurement values (a total of 619,418) can be included to determine the peak-to-peak range of jitter. The more accurate range of frequency jitter is 49.5 MHz.