AN1934 Freescale Semiconductor / Motorola, AN1934 Datasheet - Page 4

• Manufacturer Part Number: AN1934
• Description: Effects of Skew and Jitter on Clock Tree Design
• Manufacturer: Freescale Semiconductor / Motorola

Jitter Values and Data Sheets
calculating the values are necessary for an understanding of
jitter and how it relates to a clock design. However, another
important factor in understanding clock drivers is the metrics
and methods used of the published specifications. These
metrics may vary from manufacturer to manufacturer and also
within a manufacturer’s clock driver offering. The values may
be given as RMS values or as a peak–to–peak value. They may
be listed as typicals or as actual maximum values.
Understanding of the background of the jitter values on a data
sheet are necessary for circuit design as well as comparing
clock driver devices.
phase, are measured over some large number of samples. The
data for a typical device, when plotted, represents a classic
distribution Gausian or bell shaped curve where most of the
clock cycles are close to the ideal frequency (in the case of
period jitter) with fewer and fewer devices having increasing
deviation from the ideal period.
Gausian curve is defined in values of standard deviation or
sigma; and the higher the sigma multiplier, the higher the
confidence level is that a device will not exhibit a jitter value
greater than a predefined amount.
sigma deviation above the mean and 1 sigma deviation below
the mean or a total of 2 sigma confidence level. Table 1 lists
these confidence factors for ±1 sigma through ±6 sigma. If data
sheet values are specified as RMS values and higher levels of
confidence are desired, then the data sheet values for jitter
must be multiplied by the desired confidence factor.
points for a device with an output frequency of 400 MHz (period
of 2.5 ns). A value of ±3 sigma or 6 sigma gives a confidence
level or a probability that the clock edge is within the distribution
of 0.9970007%. The ±3 sigma limits define the upper period
limit of approximately 2.52 ns and the lower period limit of
approximately 2.48 ns.
jitter of a 400 MHz clock device.
Definitions of the various types of jitter and the equations for
Jitter measurements, whether cycle–to–cycle period or
In classical statistics, the distance from the center of the
Data sheet specifications that list RMS values imply a 1
Figure 9 shows the Gausian distribution curve and sigma
Figure 10 shows actual measurements made for the period
4
∅
∅
Figure 8. Phase Jitter
∅
∅
Freescale Semiconductor, Inc.
For More Information On This Product,
Go to: www.freescale.com
(approximately 2.5 ns) with the standard deviation being 6.48
ps and the peak–to–peak jitter being 57 ps. This data was
captured with a sample size of 13100 samples.
peak–to–peak. However, peak–to–peak values may be
needed for clock jitter analysis. If RMS values are specified,
peak–to–peak values may be derived based upon the required
system reliability for the specific applications. If the other cases
where peak values are specified, these values may be used
directly. However, the question that must be asked in order to
use the manufacturer’s peak–to–peak values is what level of
uncertainty is being specified. These parameters may be
specified differently on each manufacturer data sheet.
Therefore, care must be taken to insure that when one is
comparing values, the values are specified and measured in a
similar fashion.
have the classic Gausian curve or statistical distribution. This
would be the case if the jitter is completely random. However,
clock driver devices may have internal mechanisms that
produce jitter that deviates from the classic bell curve. In this
case, the total jitter is composed of the addition of a series of
bell curves providing a more complex distribution. With care,
the terms of RMS and the various sigma levels may still apply.
Table 1. Confidence Factor
Sigma
In this example, the mean period is 2.49921 ns
RMS values for jitter look better on the data sheet than
The above discussion assumes the jitter measurements
±1
±2
±3
±4
±5
±6
Figure 9. Classic Gausian Distribution Curve
(10 sigma)
(12 sigma)
(2 sigma)
(3 sigma)
(6 sigma)
(8 sigma)
Value
0.68268948
0.95449988
0.99730007
0.99993663
0.99999943
0.99999999
Confidence Factor
MOTOROLA