Teledyne LeCroy SDA II - Manual (Page 16)

Serial Data Analysis II Software
l
A damping factor below 0.707 provides an under-damped response that reacts more slowly to sud-
den changes in frequency; however, does not over-correct.
The default value of 0.707 represents a critically damped response that provides the fastest reaction time
without over-correcting.
The second order PLL with a damping factor of 0.707 is specified in the serial ATA generation II document.
This type of PLL is also very useful for measuring signals with spread-spectrum clocking because it can
accurately track and remove the low-frequency clock spreading while allowing the signal jitter to be meas-
ured. The natural frequency is somewhat lower than the actual 3 dB cutoff frequency given by the equa-
tion:
The quantity is the damping factor, and is the natural frequency. For a damping factor of 0.707,
this relationship is f
c
= 2.06 f
n
. Settings for the Custom PLL: Touch inside the Poles field to select the
order of the PLL. The number of poles can be 1 or 2.
Clock Recovery Theory
Eye, jitter and vertical noise analysis all require a clock that defines the boundaries of each unit interval.
Additionally, the clock is used to determine the difference between actual and expected arrival times of
data edges (i.e., the time interval error, or TIE measurements). In many of today's high-speed serial stand-
ards, the clock is not a physical signal but is instead derived from the data signal via clock recovery hard-
ware or software.
You may use a PLL as part of the clock recovery algorithm in order to best emulate the PLL in a receiver.
The recovered clock is defined by a list of times that correspond to expected edge arrival times.
The first step in creating a clock signal that is tracked by a PLL is to create a digital phase detector. This is
simply a software component that measures the location in time when a signal crosses a given threshold
value. Even given the maximum sampling rate available), interpolation is necessary in order to accurately
determine crossing times. Interpolation is automatically performed by SDAII. Interpolation is not per-
formed on the entire waveform; rather, only the points surrounding the crossing level are interpolated. A
cubic interpolation is used, followed by a linear fit to the interpolated data to find the precise time that a
data signal edge traverses the crossing level (see figure below).
14 922968 Rev A